Sim Studio v0.5.103 adds OpenAPI docs and three new integrations//Sim Studio releases v0.5.103 with enhanced API documentation via OpenAPI spec and auto-generated endpoint pages, plus new integrations for Amplitude, Google PageSpeed Insights, and PagerDuty. The release also includes a monitoring fix to improve production visibility for memory telemetry.
Magic Patterns launches usage-based pricing with on-demand billing and feature-tiered plans//Magic Patterns is restructuring its pricing model effective April 2, 2026, moving from flat-rate credits to usage-based pricing that scales with request complexity. The new plan tiers emphasize features rather than credit limits, with all paying customers migrated to equivalent plans and existing pricing honored through June.
pricingfeatureplatform
Ollama v0.17.5 adds GGUF model compatibility support//Ollama releases v0.17.5 with improved compatibility for imported GGUF models. The update enables better integration with external model formats, expanding the range of models users can run locally.
Sim v0.5.102 adds 10 new integrations including X API v2, Databricks, and Greenhouse recruiting tools//Sim Studio released v0.5.102 with major expansion to its integration ecosystem, adding support for Google Contacts, X (Twitter), Resend, Loops, Ashby, Greenhouse, Gamma, Luma, and Databricks. The release also introduces MCP server discovery mode for agents and includes numerous performance improvements and bug fixes.
GitHub Copilot usage metrics now track CLI activity for enterprise deployments//GitHub has expanded its Copilot enterprise usage metrics API to include telemetry from the Copilot CLI, enabling organizations to track daily active users, request counts, and token consumption across command-line environments. This addition gives enterprises comprehensive visibility into how developers use Copilot across all interfaces and helps inform rollout planning and support allocation.
apifeaturerelease
NVIDIA Launches Free GPU-Accelerated Endpoints for Alibaba's Qwen3.5 397B Vision-Language Model//Alibaba's Qwen3.5 is now available on NVIDIA's free GPU-accelerated endpoints and through API integration, bringing a 397B-parameter multimodal model with UI navigation capabilities to developers. The release includes support for containerized deployment via NVIDIA NIM and fine-tuning tools through the NeMo framework for domain-specific customization.
GitHub Copilot metrics now generally available with enterprise dashboards and APIs//GitHub Copilot usage metrics is now generally available, providing enterprises with dashboards and APIs to track adoption and usage trends across their organization. The release includes enterprise, organization, and user-level visibility alongside fine-grained access controls and data residency support.
Vercel Queues enters public beta with durable event streaming and at-least-once delivery//Vercel Queues, a durable event streaming system built on Fluid compute, is now available in public beta for all teams. The service enables developers to reliably defer work, process messages asynchronously with automatic retries, and guarantee task completion even during function crashes or deployments.
Cloudflare Radar adds ASPA monitoring to detect BGP route leaks in real time//Cloudflare has introduced ASPA (Autonomous System Provider Authorization) deployment monitoring to Cloudflare Radar, enabling operators to track and verify the path Internet traffic takes across networks. ASPA builds on RPKI to detect route leaks—misdirected traffic caused by configuration errors or malicious actions—by cryptographically validating the chain of authorized providers in the AS path.
Cloudflare Radar adds post-quantum monitoring for origin connections, key transparency dashboards//Cloudflare expands Radar with new security datasets including post-quantum encryption support tracking for origin servers (now at ~10% adoption), a real-time Key Transparency dashboard for end-to-end messaging services like WhatsApp, and expanded BGP routing security metrics via ASPA deployment tracking.
OpenAI and Amazon launch Stateful Runtime Environment on AWS Bedrock; $50B investment announced//OpenAI and Amazon have announced a multi-year strategic partnership that includes a $50 billion investment and several new product launches. Key deliverables include a Stateful Runtime Environment for AI agents on Amazon Bedrock, exclusive AWS distribution of OpenAI Frontier for enterprise AI deployment, and a $100 billion compute agreement expanding on their existing partnership.
OpenAI and Amazon launch Stateful Runtime Environment for production-grade agents in Bedrock//OpenAI and Amazon have jointly built a Stateful Runtime Environment that runs natively in Amazon Bedrock, enabling developers to deploy multi-step AI agents with persistent state, context, and governance controls. The runtime eliminates the need for manual orchestration by handling state management, tool invocations, error handling, and long-running task resumption automatically.
featureapiplatformintegrationrelease
Ollama v0.17.3 fixes tool calling bug in Qwen 3 and 3.5 models//Ollama released v0.17.3, a patch version that addresses a critical parsing issue with tool calls in the Qwen 3 and Qwen 3.5 model families. The bug occurred when tool calls were emitted during the model's thinking phase, preventing proper execution of function calls.
releasebugfix
Ollama v0.17.2 fixes Windows app crash on update downloads//Ollama releases v0.17.2, a patch release addressing a critical stability issue affecting Windows users. The update resolves a crash that occurred when the application detected and downloaded new updates.
releasebugfix
Notion API adds Markdown Content endpoints; SDK gains auto-retry and MCP enhancements//Notion released v5.10.0 and v5.11.0 of its JavaScript/TypeScript SDK with three new Markdown Content API endpoints for creating, reading, and updating page content using enhanced markdown instead of block-based operations. The update also includes automatic retry with exponential backoff, improved MCP tools with ~91% context token reduction, and new capabilities for managing comments, meeting transcripts, and data sources.
Cloudflare Radar adds post-quantum encryption tracking and Key Transparency audit dashboard//Cloudflare Radar now provides visibility into post-quantum cryptography adoption with new APIs and dashboards tracking X25519MLKEM768 support across origins. A Key Transparency dashboard was also added to monitor audit logs for end-to-end encrypted messaging services like WhatsApp and Facebook Messenger.
Vercel Chat SDK adds Telegram adapter, now supports five messaging platforms//The Chat SDK now supports building Telegram bots with the same single-codebase approach used for Slack, Discord, GitHub, and Teams. The new adapter enables developers to create interactive workflows with mentions, reactions, direct messages, and keyboard buttons directly in Telegram.
featuresdkintegration
Deepgram Flux adds Configure control message for mid-stream speech recognition updates//Deepgram's Flux speech-to-text API now supports dynamic configuration changes without disconnecting, allowing voice agents to inject context-specific vocabulary, adjust turn detection sensitivity, and modify timeout behavior on the fly. The feature is currently available via WebSocket connections with SDK support coming soon.
Sim releases v0.5.101 with Devin integration and expanded Google Workspace support//Sim Studio v0.5.101 adds autonomous coding session integration with Devin alongside new integrations for Google Tasks, BigQuery, and Translate. The release also expands Google Workspace tool coverage and includes workflow management improvements with lock/unlock functionality.
GitHub Actions now supports non-zipped artifacts, eliminating double compression friction//GitHub Actions has added support for uploading and downloading unzipped artifacts through updated versions of `actions/upload-artifact` (v7) and `actions/download-artifact` (v8). Users can now set the `archive` parameter to `false` to skip automatic compression, enabling direct browser viewing of compatible file types and eliminating the "double zip problem."
GitHub Copilot metrics API now served from new CDN domains//GitHub has migrated the download URLs returned by the Copilot usage metrics API to a new endpoint hosted on different CDN domains. The API contract and response schema remain unchanged, but organizations with firewall allowlist rules will need to update their domain patterns.
Linear adds deeplinks to AI coding tools, enables prefilled prompts with issue context//Linear now lets developers launch AI coding tools directly from issues with pre-populated context including descriptions, comments, and linked references. The feature supports nine tools including Claude Code, Cursor, and v0, with customizable prompt templates for standing instructions.
Google launches Nano Banana 2 image model combining Pro capabilities with Flash-speed generation//Google DeepMind has released Nano Banana 2, a new image generation model that combines advanced features like world knowledge and text rendering with Gemini Flash's speed. The model is now rolling out across Google products including Gemini, Search, and Ads, enabling faster creative iteration and editing workflows.
GitHub launches Copilot Content Exclusion REST API in public preview//Admins can now programmatically manage Copilot content exclusion rules via a new REST API supporting GET and SET operations at organization and enterprise levels. This enables automation and governance at scale for controlling what content is excluded from Copilot's context.
Vercel rolls out redesigned dashboard with sidebar navigation and mobile optimization//Vercel's new dashboard design is now the default experience for all users following a successful beta launch in January. The redesign introduces a resizable sidebar, consistent navigation tabs, and mobile-optimized controls to streamline developer workflows.
Vercel AI Gateway adds Gemini 3.1 Flash Image Preview with web search and configurable reasoning//Nano Banana 2 (Gemini 3.1 Flash Image Preview) is now available on Vercel's AI Gateway, bringing improved visual generation quality while maintaining flash-tier performance. The model supports Google Image Search integration for grounding outputs in real-world imagery, configurable thinking levels for complex prompts, and new resolutions including 512p with 1:4 and 1:8 aspect ratios.
GitHub Mobile adds live notifications for Copilot coding agent sessions//GitHub Mobile now supports Live Activities to track coding agent progress in real time, showing session status, pull request details, and supporting iOS 17.2+ and Android 16+. The feature works with Copilot coding agents and third-party agents like Claude and Codex, allowing developers to monitor cloud-run sessions from anywhere.
Bun v1.3.10 launches native REPL, adds ES decorators support, enables self-contained HTML compilation//Bun v1.3.10 introduces a completely rewritten REPL built in Zig with instant startup and full terminal UI, adds support for TC39 standard ES decorators, and enables compiling TypeScript/JSX to self-contained HTML files with all assets inlined. The release addresses three major developer pain points with feature completeness and improved developer experience.
OpenAI Codex integrates with Figma, enabling seamless code-to-design workflows//OpenAI and Figma have launched a new integration that allows developers to generate and iterate on designs bidirectionally—converting code to editable Figma designs and vice versa. The integration uses Figma's MCP Server to connect Codex directly to Figma's design tools, supporting a unified workflow where teams can start from code, design, or a prompt.
integrationfeatureapi
vLLM v0.16.0 brings async scheduling with pipeline parallelism, 30.8% throughput gains//vLLM's latest release introduces async scheduling combined with pipeline parallelism, delivering significant performance improvements and new WebSocket-based realtime audio streaming capabilities. The update adds support for 12+ new model architectures and major enhancements to speculative decoding, RLHF workflows, and Intel XPU platforms.
GitHub expands Copilot Usage Metrics API with org-level pull request metrics parity//GitHub has extended its organization-level Copilot Usage Metrics API to include comprehensive pull request activity metrics, bringing feature parity with enterprise-level endpoints. Organizations can now track pull request creation, throughput, and Copilot's impact on the development lifecycle through a unified API.
apifeatureplatform
Databricks AI/BI adds agent mode, pivot table hierarchies, and chart faceting in February 2026 update//Databricks released a comprehensive update to its AI/BI platform with dozens of dashboard and Genie enhancements. New capabilities include Agent mode (Beta) with parallel queries, pivot table hierarchies, chart faceting, improved table visualization with aggregations, and expanded dashboard grid layouts supporting 12 columns.
releasefeatureplatform
Cursor's Bugbot Autofix now proposes pull request fixes automatically; 35% merge rate//Cursor has launched Bugbot Autofix, a feature that automatically identifies issues in pull requests and proposes fixes using cloud agents. The feature has already achieved a 35% merge rate for proposed changes and can be configured to either post comments for manual approval or push changes directly to branches.
GitHub's Enterprise AI Controls reaches general availability with agent governance tools//GitHub has released Enterprise AI Controls and the agent control plane as generally available features, giving enterprise administrators deeper oversight and auditability of AI agent usage across their organizations. The release includes new capabilities for discovering agent activity, configuring enterprise agent policies via API, and managing custom agent standards through fine-grained permissions.
Vercel expands Developer role to Pro teams, enabling granular access control//Vercel now allows Pro team members to be assigned the Developer role, previously exclusive to Enterprise plans. This enables more granular access control for deployment and reduces team-wide configuration visibility for non-admin developers.
Cloudflare adds Markdown error responses for 1xxx errors; reduces payload size by 98%//Cloudflare now supports structured Markdown responses for its 1xxx error codes when clients request `Accept: text/markdown`. The new format includes YAML frontmatter with actionable metadata and guidance sections, dramatically reducing payload size and token footprint compared to HTML error pages.
featureapiperformance
Databricks adds Python unit testing, ISMAP compliance, and maintenance updates across 8 runtime versions//Databricks released a broad set of February 2026 updates including integrated Python unit testing tools, ISMAP compliance support for Japanese government requirements, and maintenance patches across Databricks Runtime 12.2 through 18.0. The release also introduces HITRUST compliance controls in preview, ADBC as the default Power BI driver, and JAR tasks for serverless compute.
GitHub expands Claude and Codex availability to Copilot Business and Pro users//Claude by Anthropic and OpenAI Codex are now available as coding agents for Copilot Business and Pro subscribers, expanding beyond Enterprise and Pro+ tiers. Both agents run natively within GitHub.com, GitHub Mobile, and VS Code with shared history and context, requiring no additional subscription fees.
GitHub Actions macOS 26 runners now generally available with Apple Silicon and Intel support//GitHub has promoted the macOS 26 runner image from public preview to general availability for GitHub Actions. The update provides fully supported environments for building and testing apps with the latest macOS and Xcode tooling, running natively on both Apple Silicon (arm64) and Intel (x64) architectures.
Cloudflare Cache now revalidates asynchronously, reducing latency and error exposure//Cloudflare has made stale-while-revalidate fully asynchronous across its cache service, allowing expired assets to be served immediately while fresh copies are fetched in the background. This eliminates visitor wait times during revalidation and provides more consistent performance across all requests.
GitHub launches GPT-5.3-Codex in Copilot across web, mobile, and VS//GitHub's new GPT-5.3-Codex model is now generally available in GitHub Copilot Chat for Enterprise, Business, Pro, and Pro+ subscribers. The model is accessible across github.com, GitHub Mobile, Visual Studio Code, and Visual Studio via the chat model picker, with org admins needing to enable the policy to roll out access.
GitHub Copilot gains native web search capabilities for real-time answers//GitHub has enabled model-native web search in GitHub Copilot Chat on github.com for specific GPT models, allowing faster and more accurate answers to time-sensitive questions. The feature is available to paid Copilot subscribers as part of the public preview program and can be toggled on or off in Copilot settings.
featurereleaseplatform
Intercom launches 12 major Procedures and Simulations updates for complex AI customer support workflows//Intercom announced significant upgrades to Fin's Procedures and Simulations capabilities, enabling AI agents to handle complex, multi-step customer queries with greater control and reliability. The updates include AI-assisted Procedure drafting, deterministic controls, improved agentic behavior, and enhanced testing tools to help teams deploy confidently at scale.
GitHub Copilot CLI reaches general availability with autonomous agentic capabilities//GitHub Copilot CLI, which debuted in public preview last September, is now generally available to all paid Copilot subscribers. The tool has evolved from a terminal assistant into a full autonomous coding agent with plan mode, autopilot capabilities, and cross-session memory, supporting multiple models from Anthropic, OpenAI, and Google.
Sim v0.5.100 adds Attio and Gong integrations, credential management overhaul//Sim releases v0.5.100 with new integrations for Attio CRM (40 tools) and Gong (18 API tools), plus support for multiple credentials per provider. The release also introduces enhanced security controls, webhook triggers for Confluence, and performance improvements across the platform.
Ollama v0.17.1-rc1 adds support for Qwen3.5 architecture//Ollama releases v0.17.1-rc1, a release candidate that introduces support for the Qwen3.5 model architecture. This expansion allows users to run Qwen3.5 models locally through Ollama's platform.
releasefeaturemodel
Devin 2.2 Launch: 3x Faster Startup, Desktop Testing, v3 API Exit Beta//Devin releases version 2.2 with significant performance improvements, new desktop testing capabilities, and official v3 API launch. The update includes a redesigned UI, faster integrations with Slack and Linear, and support for end-to-end testing across Linux desktop applications.
Cloudflare extends Python Workers to Windows with Pywrangler support//Pywrangler, the CLI tool for managing Cloudflare Python Workers, now supports Windows environments alongside macOS and Linux. Developers can develop and deploy Python Workers on Windows using the same commands and workflows as other platforms with no additional configuration.
Stripe adds tax collection support for Sri Lanka digital services//Stripe has expanded its tax capabilities to support tax collection for digital services in Sri Lanka, including support for Sri Lanka VAT tax IDs. This enables merchants operating in Sri Lanka to properly handle tax compliance for their digital service offerings.
apifeaturerelease
Databricks Releases Serverless Environment Version 5 with Enhanced SQL and Geospatial Support//Databricks has released serverless environment version 5, available for both CPU and GPU configurations, bringing new SQL capabilities and spatial functions to notebooks and jobs. The latest version introduces support for recursive CTEs with LIMIT ALL, dynamic SQL with constant expressions, and expanded geospatial functions for polygon operations.
Cloudflare WAF adds three new detection rules for SmarterMail flaws and command injection//Cloudflare is releasing scheduled WAF updates on March 2, 2026, introducing two new detections for SmarterMail vulnerabilities (CVE-2025-52691 and CVE-2026-23760) plus a beta rule for command injection via Nslookup. All rules will initially log violations without blocking traffic.
Vercel activity log now tracks 100% of team and project changes with 88 new events//Vercel has expanded its activity log to capture every change made to team and project settings, with 88 new events added to achieve complete coverage. The feature provides full visibility into who changed what and when across your organization.
Cloudflare Containers now support 15x higher resource limits, up to 6TB memory and 30TB disk//Cloudflare has significantly expanded the resource limits for its Containers platform, enabling users to run concurrent workloads at much larger scale. The new limits support up to 6TB of memory (from 400GB), 1,500 vCPUs (from 100), and 30TB of disk (from 2TB) across concurrent live instances.
Agents SDK v0.6.0 adds RPC transport for MCP, optional OAuth, and schema hardening//Cloudflare's Agents SDK v0.6.0 introduces RPC transport for MCP connections within Workers, eliminating network overhead and enabling direct Durable Object bindings. The release also makes OAuth optional for simple MCP connections and hardens the JSON Schema converter with circular reference guards and better error handling.
Wrangler 4.68.0 brings automatic configuration; projects deploy without config files//Cloudflare Workers' Wrangler CLI now automatically detects project frameworks and deploys without requiring manual configuration. The feature, which graduated from experimental status, simplifies onboarding by auto-generating wrangler.jsonc files and installing required adapters.
Cloudflare Workers Observability adds structured query language for logs and traces//Workers Observability now includes a query language that lets developers write structured queries to filter logs and traces directly in the search bar. The feature supports free text search, field-level queries, comparison operators, functions like regex and startsWith, and boolean logic, with two-way sync between the search bar and Query Builder sidebar.
Cloudflare Radar adds RPKI ASPA deployment tracking with new API endpoints//Cloudflare Radar now provides visibility into Autonomous System Provider Authorization (ASPA) adoption across the global routing ecosystem. The update includes three new API endpoints for retrieving ASPA snapshots, changes, and timeseries data, plus new dashboard widgets for tracking ASPA deployment trends globally and by region.
featureapiplatform
Deepgram Voice Agent adds UpdateThink for mid-conversation LLM provider switching//Deepgram has released UpdateThink, a new client message that allows developers to dynamically replace the entire Think provider configuration during live Voice Agent conversations. This enables switching LLM providers, models, prompts, and functions in real-time without interrupting the conversation.
featureapiplatform
Sourcegraph 7.0 debuts MCP integration for AI agents; adds semantic search across enterprise codebases//Sourcegraph 7.0 repositions the platform as a shared intelligence layer for both human developers and AI coding agents, introducing deep semantic search capabilities via the Model Context Protocol (MCP). The release enables AI agents to reason about cross-repository dependencies, architectural patterns, and code history with the same accuracy developers rely on.
CodeQL 2.24.2 adds Go 1.26 and Kotlin 2.3.10 support, improves query accuracy//CodeQL's latest release extends language support to Go 1.26 and Kotlin 2.3.10 while refining query accuracy across multiple languages. Improvements include better false positive elimination in C# CSRF detection and enhanced sanitizer recognition for Java/Kotlin injection queries.
Dependabot now consolidates cross-directory dependency updates into single pull requests//Dependabot now supports grouping dependency updates by name across multiple directories in a single pull request. This feature is especially valuable for monorepos where a single dependency bump can affect numerous directories, significantly reducing pull request volume and simplifying upgrade workflows.
Replit launches Pro tier at $100/month; Core drops to $20/month with team collaboration//Replit is restructuring its pricing with a new Pro plan ($100/month) for serious builders offering Turbo Mode for Agent, tiered credits up to $4,000/month, and support for up to 15 team members. The Core plan is being simplified and reduced from $25 to $20/month while gaining collaboration features for up to 5 users, making it more accessible for newcomers.
GitHub Code Quality adds organization-level dashboard for cross-repository health tracking//GitHub Code Quality now offers an organization-level dashboard in public preview, enabling teams to view code health metrics across multiple repositories in one place. The dashboard allows sorting and filtering by reliability, maintainability, and other quality attributes while respecting repository permissions.
Vercel launches Slack Agent Skill to simplify agent deployment with one-click setup//Vercel's new Slack Agent Skill enables developers to build and deploy Slack agents directly from their coding assistant in a single session, automating OAuth configuration, webhook setup, and deployment to Vercel. The wizard-based tool handles infrastructure complexity so developers can focus on agent logic.
GPT 5.3 Codex arrives on Vercel AI Gateway with 25% faster execution//Vercel's AI Gateway now offers GPT 5.3 Codex, a new model combining GPT-5.2's coding strengths with enhanced reasoning capabilities. The model is optimized for multi-step agentic work and delivers 25% faster performance with improved token efficiency.
Vercel doubles Python Functions bundle size limit to 500MB//Vercel has increased the maximum uncompressed deployment bundle size for Python Functions from 250MB to 500MB. This change enables developers to deploy more complex Python applications with larger dependencies on Vercel's serverless platform.
Cloudflare WARP Client for Windows reaches GA with multi-user mode improvements//Cloudflare has released version 2026.1.150.0 of the WARP client for Windows as a stable GA release. The update includes fixes for multi-user mode MDM configuration loss, a new NetBIOS over TCP/IP management feature, and several bug fixes addressing DNS failures and local network exclusion issues.
Cloudflare releases WARP macOS client v2026.1.150.0 with DNS, tunnel fixes//Cloudflare has released a new GA version of its WARP client for macOS, addressing several critical bugs affecting local network exclusion, DNS configuration, and tunnel connectivity. The update focuses on reliability improvements for enterprise users managing Zero Trust device security.
releasebugfixplatform
Ollama v0.17.1-rc0 updates mlx-c bindings to 0.5.0//Ollama releases v0.17.1-rc0, a release candidate that updates mlx-c bindings to version 0.5.0 and switches Linux builds to use GCC 11. This incremental update improves compatibility and compiler support for users running Ollama on Linux systems.
Cloudflare releases WARP client v2026.1.150.0 for Linux with DNS and local network fixes//Cloudflare has released a new general availability version of the WARP client for Linux, addressing DNS failures, local network exclusion issues, and improving device reporting in the Cloudflare One dashboard. Users must update their Linux package public key if they installed before September 12, 2025 to maintain repository access beyond December 4, 2025.
Cloudflare Containers raises resource limits 15x; now supports up to 6TB memory and 30TB disk//Cloudflare has significantly expanded the capacity of its Containers platform, increasing memory limits to 6TiB (from 400GiB), vCPU to 1,500 (from 100), and disk space to 30TB (from 2TB). This 15-fold increase enables developers to run larger-scale workloads concurrently, supporting up to 1,500+ standard instances simultaneously.
Cloudflare's deleteAll() now clears Durable Object alarms, reducing API calls//Cloudflare's `deleteAll()` method for Durable Objects now automatically deletes alarms alongside stored data, consolidating cleanup operations. This change applies to both KV-backed and SQLite-backed Durable Objects with a compatibility date of 2026-02-24 or later.
Cloudflare rolls out asynchronous stale-while-revalidate, eliminating cache revalidation latency//Cloudflare has made its stale-while-revalidate caching feature fully asynchronous, eliminating the need for end users to wait for origin round-trips during cache revalidation. The change is live for all Free, Pro, and Business tiers, with Enterprise zones rolling out throughout the quarter.
Cloudflare Workers: deleteAll() now clears Durable Object alarms automatically//Cloudflare Workers' deleteAll() method now removes both stored data and alarms in a single API call for Durable Objects with compatibility date 2026-02-24 or later. This change eliminates the need for separate deleteAlarm() calls and applies to both KV-backed and SQLite-backed storage.
Cloudflare Stream adds ability to disable and enable live inputs//Cloudflare Stream now allows developers to disable live inputs to reject incoming RTMPS and SRT connections, giving more control over broadcasts. The feature can be toggled via API or Dashboard, enabling temporary pauses, programmatic broadcast termination, and prevention of new streams on specific inputs.
Cloudflare Pipelines adds dropped event metrics, typed bindings, and streamlined setup//Cloudflare has shipped three improvements to Pipelines: new dashboard and GraphQL metrics for tracking dropped events with detailed error messages, schema-specific TypeScript types for Pipeline bindings to catch data quality issues at compile time, and a simplified setup command that auto-configures R2 buckets and sensible defaults.
GitHub Enterprise Server 3.20 RC adds immutable releases, enhanced secret scanning, and backup service//GitHub Enterprise Server 3.20 release candidate introduces several major features including immutable releases for supply chain protection, enhanced secret scanning with validity checks and enterprise-level bypass controls, and a new backup service that replaces the need for separate backup utilities. The release also adds enterprise team management capabilities and new security roles for simplified governance.
Cloudflare Pipelines adds dropped event metrics, typed bindings, and simplified setup//Cloudflare Pipelines now surfaces dropped event metrics with detailed error categorization via dashboard and GraphQL, generates schema-specific TypeScript types for Pipeline bindings to catch schema mismatches at compile time, and streamlines initial setup with automated R2 bucket creation and sensible defaults.
featureapiplatformrelease
Cursor Adds Computer Use to Cloud Agents, Enabling Self-Testing and Demo Recording//Cursor's cloud agents can now interact with software they create using isolated VMs, test changes automatically, and generate video/screenshot artifacts during the development process. The feature is available across Cursor's web, desktop, mobile, Slack, and GitHub integrations.
Claude Code 2.1.52 fixes extension crash on Windows, ships with 50+ improvements in recent versions//Claude Code 2.1.52 addresses a critical VS Code extension crash on Windows. The latest releases (2.1.50-2.1.52) deliver significant improvements including worktree isolation for agents, optimized memory usage in long sessions, enhanced MCP OAuth authentication, and support for custom npm registries.
releasebugfixfeatureperformancesecurity
NVIDIA Achieves 1.6x Training Throughput Gains with NVFP4 Low-Precision Format on Blackwell GPUs//NVIDIA has published results demonstrating that NVFP4, a 4-bit low-precision training format, delivers up to 1.6x higher throughput and substantial memory savings while maintaining accuracy comparable to BF16 on large-scale model training. The results come from experiments training Llama 3 8B and an internal research model on 1 trillion tokens using the open-source NeMo Megatron Bridge framework on B200 GPUs.
Vercel deprecating now.json config files, migration deadline set for March 31, 2026//Vercel is sunsetting support for the legacy now.json configuration file format, with removal scheduled for March 31, 2026. Developers need to rename their now.json files to vercel.json, a straightforward migration with no content changes required.
Vercel opens Chat SDK in public beta, unifies bot development across Slack, Teams, Discord, and five other platforms//Vercel has open-sourced the Chat SDK, a TypeScript library that allows developers to write chatbot logic once and deploy across Slack, Microsoft Teams, Google Chat, Discord, GitHub, and Linear. The SDK features event-driven architecture with type-safe handlers, JSX-based UI components that render natively on each platform, and pluggable state management adapters.
Sim v0.5.99 adds expandable workflow blocks and fixes parallel subflow display//Sim v0.5.99 introduces expandable child workflow blocks in the console terminal, improving workflow visibility and debugging. The release also fixes active state pulsing for parallel subflow blocks and resolves a security issue with localhost HTTP connections.
Cloudflare One becomes first SASE platform with post-quantum encryption across all components//Cloudflare One now offers post-quantum hybrid ML-KEM encryption across its entire Secure Access Service Edge platform, including Secure Web Gateway, Zero Trust, and Wide Area Network services. The expansion covers Cloudflare IPsec (in closed beta) and Cloudflare One Appliance (generally available), enabling organizations to secure their enterprise network traffic against future quantum threats ahead of NIST's 2030 cryptographic transition deadline.
GitHub adds enterprise team support to organization APIs, unifying team discovery//Organization owners in GitHub Enterprise can now query both enterprise and organization teams through a single set of API endpoints, eliminating the need for separate queries and permission management. This update simplifies team management for enterprise customers by providing unified visibility and filtering capabilities across team types.
Hyperdrive stops caching queries with PostgreSQL STABLE functions to prevent stale data//Hyperdrive now treats PostgreSQL STABLE functions (like NOW(), CURRENT_TIMESTAMP) as uncacheable, preventing incorrect cached results from being served across transactions. This aligns with PostgreSQL's function volatility semantics and may require application code changes for affected queries.
Cloudflare Sandbox SDK adds backup and restore API for persistent environments//Cloudflare Sandbox now supports `createBackup()` and `restoreBackup()` methods to create point-in-time snapshots of directories, eliminating the need to re-run expensive setup steps like git clones and npm installs. Backups are stored in R2 with support for TTLs and can be reused across multiple sandbox sessions.
Cloudflare Sandbox SDK adds backup and restore API for point-in-time snapshots//Cloudflare's Sandbox SDK now supports `createBackup()` and `restoreBackup()` methods for creating and restoring point-in-time snapshots of directories, eliminating the need to repeat expensive setup steps like git clones and npm installs across sandbox sessions. Backups are stored in R2 with configurable TTLs and can be forked across multiple sandbox instances.
GitHub extends IP allow list coverage to Enterprise Managed Users namespaces in public preview//GitHub Enterprise Cloud now supports IP allow list enforcement for Enterprise Managed User (EMU) namespaces in public preview, extending network access controls to user-owned repositories. This feature applies filtering across web UI, git protocol, and API access, with support for all credential types including personal access tokens, app tokens, and SSH keys.
Vercel Sandbox adds secure header injection for API credentials in untrusted code//Vercel Sandbox now supports automatic injection of HTTP headers containing API credentials into outbound requests, keeping secrets outside the sandbox VM boundary. This feature is designed for AI agent workflows where prompt injection poses security risks, allowing sandboxed code to call authenticated services without ever accessing the credentials directly.
Cloudflare Security Center adds saved views for Threat Events dashboard//Cloudflare Security Center now lets users save custom dashboard configurations for Threat Events, enabling analysts to instantly access pre-filtered views without manual reconfiguration. The feature supports complex filtering scenarios like industry-specific attacks and geographic filtering, improving workflow efficiency for Cloudforce One subscribers.
Cloudflare adds saved views to Security Center's Threat Events dashboard//Cloudflare Security Center now lets users create and save custom filtered views of the Threat Events dashboard, enabling analysts to instantly return to specific configurations like industry-specific attack patterns or regional data flows. The feature eliminates repetitive manual filter reapplication and helps teams maintain consistent threat intelligence views.
GitHub enterprise administrators can now create custom organization roles available across all organizations//Enterprise administrators can now create a standardized set of custom organization roles available across all organizations within their enterprise, supporting up to 20 roles at once. These roles help enforce compliance requirements and provide consistency when users move between organizations, with assignment available through organization settings or the organization role assignment API.
releasefeatureapi
Docker Desktop 4.62.0 adds MCP server profiles and custom catalogs//Docker Desktop 4.62.0 introduces the ability to organize MCP servers into named profiles and create custom team catalogs. The release also includes kernel updates, bug fixes across all platforms, vLLM Metal support for Mac, and a security patch for a grpcfuse kernel module vulnerability.
releasefeaturesecuritybugfixplatform
Databricks SQL 2025.40 introduces SQL scripting GA and expanded parameter markers//Databricks SQL version 2025.40 is now rolling out to the Current channel with SQL scripting moving to general availability. The release adds expanded support for parameter markers and IDENTIFIER clauses across SQL contexts, plus new approximate computation functions for distinct counts and quantiles.
Sim Studio v0.5.98 adds hex tool support and OAuth client metadata discovery//Sim Studio releases v0.5.98 with new hex tooling and improved OAuth support via CIMD client metadata discovery. The update also addresses critical stability issues including Redis outage handling and removes deprecated Claude models.
Sim Studio v0.5.97 fixes OAuth metadata URL discovery in MCP//Sim Studio released v0.5.97 with a targeted bug fix addressing OAuth discovery metadata URLs in the Model Context Protocol (MCP) integration. The patch ensures proper base URL handling for OAuth-related endpoints.
releasebugfixintegration
Ollama v0.17.0-rc1 removes noisy error output from MLX library loading//Ollama's v0.17.0-rc1 release candidate addresses misleading error messages that appeared during dynamic library loading for MLX on systems without rpath set. The fix silences expected fallback failures while preserving proper error logging for actual failures.
Sim v0.5.96 adds Slack ephemeral messages and OAuth 2.1 support//Sim Studio released v0.5.96 with new Slack ephemeral messaging capabilities and OAuth 2.1 provider integration for MCP connectors. The release also includes bug fixes for Slack event handling, copilot operation conditions, and log data management.
releasefeaturebugfixintegrationapi
Granola launches native HubSpot integration for automatic CRM syncing from meeting notes//Granola now offers a direct HubSpot integration (available on Business and Enterprise plans) that syncs enhanced meeting notes directly to HubSpot Contact records with a single click. For fully automated syncing without manual steps, Zapier provides additional routing capabilities and multi-object workflows.
Sim Studio v0.5.95 adds Gemini 3.1 Pro support and 4 new integrations//Sim Studio v0.5.95 brings new model support with Gemini 3.1 Pro Preview and expanded integrations including Redis, Upstash, Algolia, and RevenueCat. The release also introduces a new tables feature and fixes for build stability and workflow management.
GitHub Copilot organization-level usage metrics dashboard enters public preview//GitHub is expanding Copilot's analytics capabilities beyond enterprise-only access, making usage metrics dashboards available to organization owners. The new dashboard provides organization-scoped insights into adoption and usage trends without requiring enterprise-wide visibility.
Cloudflare launches MCP server using Code Mode, reducing API context requirements by 99.9%//Cloudflare introduced a new Model Context Protocol (MCP) server that provides access to the entire Cloudflare API using just two tools and consuming only ~1,000 tokens. The approach, called Code Mode, allows AI agents to write JavaScript code against a typed SDK to explore and execute API operations, reducing token usage from 1.17M to 1K compared to traditional MCP implementations.
Anthropic launches Claude Code Security in limited preview; found 500+ zero-day vulnerabilities in open-source code//Claude Code Security, a new AI-powered capability within Claude Code, automatically scans codebases for complex vulnerabilities and suggests patches—finding security issues that traditional static analysis tools miss. Available in limited research preview for Enterprise and Team customers, the tool leverages recent improvements in Claude's cybersecurity abilities demonstrated through over 500 zero-day discoveries in production open-source repositories.
Cloudflare Tunnel now available in main Dashboard with full lifecycle management//Cloudflare Tunnel management is now integrated into the main Cloudflare Dashboard, eliminating the need to switch between interfaces. The new experience provides complete tunnel lifecycle management, native integrations with DNS and Workers VPC, real-time health monitoring, and a unified routing map for all ingress routes.
featureplatformintegration
Railway adds default DDoS protection via Fastly, launches domain registration//Railway has partnered with Fastly to provide automatic DDoS protection on all public services, with traffic filtered across 100+ global points of presence. The platform also now offers integrated domain registration and management, plus per-environment canvas layouts to prevent conflicting service arrangements.
Cloudflare Tunnel management now available in core dashboard//Cloudflare Tunnel is now accessible directly from the main Cloudflare Dashboard under Networking > Tunnels, enabling developers to create, configure, and monitor tunnels in one place. The new interface includes full lifecycle management, native integrations with DNS and Workers VPC, and a unified routing map for managing all ingress routes.
Cloudflare releases @cloudflare/codemode v0.1.0 with runtime-agnostic SDK architecture//Cloudflare has rewritten its codemode package into a modular, runtime-agnostic SDK that enables LLMs to write and execute code for tool orchestration. The new version introduces a standard `Executor` interface and includes a prebuilt `DynamicWorkerExecutor` for running generated code in Dynamic Workers with network isolation and console capture.
Cloudflare CASB now includes AI-powered Cloudy Summaries for instant security findings//Cloudflare has launched Cloudy Summaries for CASB, integrating AI-powered plain-language explanations of security misconfigurations and data exposures. The feature reduces triage time by providing contextual explanations, risk assessments, and actionable remediation guidance across all major SaaS integrations.
Cloudflare releases @cloudflare/codemode v0.1.0 with runtime-agnostic architecture//Cloudflare has rewritten its Code Mode SDK into a modular, runtime-agnostic package that allows LLMs to generate and execute code for orchestrating tools. The new version removes the built-in LLM integration, introduces a minimal `Executor` interface for code sandboxing, and includes a `DynamicWorkerExecutor` for running code in Cloudflare Workers with network isolation and console capture.
GitHub Copilot support in Zed editor now generally available//GitHub Copilot is now officially supported in Zed, a high-performance Rust-based code editor created by the makers of Atom. All Copilot Pro, Pro+, Business, and Enterprise subscribers can authenticate directly in Zed to access AI-powered coding assistance without any additional licensing.
GitHub Copilot coding agent gains model picker for Business and Enterprise users//GitHub has expanded model selection capabilities to Copilot Business and Enterprise users, allowing them to choose between eight different AI models for their coding agent tasks. Previously available only to Pro users, this feature lets developers optimize for speed, cost, or specific capabilities when delegating work to Copilot's autonomous background agent.
Gemini 3.1 Pro now available in GitHub Copilot with improved tool precision and fewer calls//Google's latest agentic coding model, Gemini 3.1 Pro, is rolling out in public preview to GitHub Copilot users. The model demonstrates strong performance in edit-then-test loops with high tool precision, enabling developers to accomplish more with fewer tool invocations.
Google releases Gemini 3.1 Pro with 77.1% ARC-AGI score, doubling reasoning performance//Google has released Gemini 3.1 Pro, an upgraded AI model with significantly improved reasoning capabilities now available across consumer and developer platforms. The model achieves a verified 77.1% score on ARC-AGI-2 benchmarks—more than double the performance of its predecessor—and is designed for complex problem-solving tasks requiring advanced reasoning.
Vercel AI Gateway adds Grok Imagine Video for text-to-video and video editing//xAI's Grok Imagine Video model is now available in Vercel's AI Gateway, enabling developers to generate videos in seconds with advanced features like video editing, audio generation, and lip-sync. Accessible via v0 Creative Studio, AI SDK 6, or the AI Gateway playground for Pro and Enterprise plan users.
Vercel adds Google Veo video models to AI Gateway with native audio synthesis//Google's Veo video generation models are now available through Vercel's AI Gateway, supporting text-to-video and image-to-video generation with synchronized audio at up to 1080p resolution. Developers can access the models via AI SDK 6 or the AI Gateway playground, with both standard and fast generation modes available.
Vercel AI Gateway adds Kling video models with multishot and audio generation//Kling video models, including the latest Kling 3.0, are now available through Vercel's AI Gateway and AI SDK. The integration enables developers to generate cinematic videos from text, images, or motion references with features like multishot narratives, synchronized audio generation, and first/last frame control.
Streamdown 2.3 adds interactive playground and unified design system for markdown rendering//Streamdown 2.3 introduces a refreshed design applying consistent nested-card styling across tables, code blocks, and diagrams, along with a new interactive playground for real-time testing. The release includes several developer experience improvements like sticky action buttons, new hooks for streaming detection, and Markdown copying for tables.
Vercel AI Gateway now supports Google's Gemini 3.1 Pro with medium thinking level//Google's Gemini 3.1 Pro Preview is now available through Vercel's AI Gateway with support for the medium thinking level. The model offers improved performance for software engineering and agentic workflows while reducing token consumption compared to earlier versions.
Vercel adds Alibaba's Wan video generation models to AI Gateway//Alibaba's Wan models for text-to-video, image-to-video, and character reference video generation are now available through Vercel's AI Gateway. Developers can generate stylized videos up to 1080p using AI SDK 6 or the Gateway playground, with beta access for Pro and Enterprise plans.
Vercel Blob adds private storage in public beta for authenticated file access//Vercel Blob now supports private storage requiring authentication for sensitive files like contracts and invoices, complementing its existing public storage option. The feature is available on all Vercel plans via CLI and SDK, with simple `access: 'private'` configuration for upload and retrieval operations.
Vercel adds billing API with FOCUS-compliant cost data access//Vercel now offers programmatic access to billing usage and cost data through a new `/billing/charges` API endpoint that returns data in the FOCUS v1.3 open-standard format. Teams can integrate cost data directly into FinOps tools, with support for day-level granularity up to one year of historical data via SDK, CLI, or REST API.
apifeatureintegration
Ollama v0.16.3-rc2 prevents partial script execution during installation//Ollama's installation process now wraps the download script in a main function to prevent truncated partial downloads from executing incomplete code. This security-focused fix ensures that interrupted installations won't leave the system in a compromised state.
Cloudflare improves AI Gateway and Workers AI dashboard with better navigation and onboarding//Cloudflare has rolled out a series of dashboard improvements for AI Gateway and Workers AI, including a dedicated top-level AI section in the sidebar, streamlined onboarding with OpenAI-compatible endpoints, and enhancements to dynamic routing and observability. The update also brings accessibility improvements to keyboard navigation and filtering components.
featureplatformintegration
Harvey launches Shared Spaces, enabling secure cross-organizational collaboration in legal AI platform//Harvey has introduced Shared Spaces, a new feature enabling secure, multiplayer collaboration across organizational boundaries for law firms and professional service networks. The capability allows legal teams to bring clients, advisors, and cross-functional teams into shared intelligent workspaces while maintaining security and confidentiality, built in partnership with design partners.
Discord Open-Sources Osprey Rule Engine for Real-Time Safety Detection//Discord has open-sourced Osprey, its safety rules engine developed in partnership with ROOST and internet.dev, enabling platforms to investigate activities and deploy dynamic rules in real-time with minimal engineering overhead. The engine processes thousands of events per second and uses a Python-inspired rule language (SML) to help teams detect and respond to emerging threats.
Claude launches automatic prompt caching and retires Sonnet 3.7, Haiku 3.5 models//Anthropic has introduced automatic prompt caching for the Claude Messages API, eliminating manual cache breakpoint management. The company is retiring older model versions (Sonnet 3.7 and Haiku 3.5) while recommending users upgrade to Sonnet 4.6 and Haiku 4.5.
Cloudflare adds threat event visualizations to Security Center dashboard//Cloudforce One customers can now view dynamic Sankey diagrams and attack distribution graphs in the Threat Events dashboard to identify emerging attack patterns and regional threat infrastructure. The new visualizations enable filtering and drill-down analysis directly from the charts.
featureplatform
NVIDIA and SGLang optimize DeepSeek for GB300 NVL72, achieving 226 tokens per second in 128K-token inference//NVIDIA and the SGLang team have published optimizations for running DeepSeek R1 on the GB300 NVL72 GPU, leveraging prefill-decode disaggregation, pipeline parallelism, and expert parallelism to achieve 226 tokens per second per GPU on long-context workloads. The optimization demonstrates a 1.53x throughput advantage over GB200 under identical conditions, with further gains possible through multi-token prediction.
Cloudflare adds threat visualization to Cloudforce One dashboard with Sankey diagrams//Cloudflare Security Center now includes dynamic visualizations for Cloudforce One Threat Events, featuring Sankey diagrams to trace attack flows geographically and charts showing campaign distribution over time. Subscribers can explore these new analytics views to better understand emerging threat patterns and drill down into specific attack vectors.
GitHub optimizes test merge commit generation for pull requests, reducing delays//GitHub has changed how and when test merge commits are generated for pull requests to improve system performance and reliability. Test merge commits are now only generated when code changes, merge bases change, or the existing commit is older than 12 hours—eliminating unnecessary generation when viewing the PR page.
GitHub Projects adds search-based imports and hierarchy view improvements//GitHub Projects now supports importing items via advanced search queries with the same filters available on the Issues page. The hierarchy view receives multiple improvements including inline sub-issue creation, drag-and-drop reordering, and synchronized ordering across projects and issues.
GitHub brings pull request comments to Files changed page, eliminating tab switching//GitHub's updated Files changed page now displays all pull request comments in a dedicated panel, allowing developers to review code and follow discussions without switching between tabs. The update also streamlines the review process by making comment summaries optional when requesting changes.
Cloudflare DEX now complies with EU Customer Metadata Boundary settings//Digital Experience Monitoring (DEX) now fully supports Cloudflare's Customer Metadata Boundary (CMB) for the EU, ensuring DEX logs remain within EU data residency requirements when enabled. Users can export DEX data via LogPush to build custom analytics dashboards while maintaining compliance.
GitHub Actions workflow dispatch API now returns run IDs and URLs//GitHub's workflow dispatch API now includes an optional `return_run_details` parameter that returns workflow metadata (ID, API URL, workflow URL) in a 200 OK response, enabling developers to track triggered runs without polling. The feature is also available in GitHub CLI v2.87.0+ with `gh workflow run` commands defaulting to return run details.
GitHub Copilot Usage Metrics API now tracks pull request throughput and merge time//GitHub has expanded its Copilot usage metrics API with new enterprise-level endpoints for measuring pull request outcomes. Organizations can now track PR review suggestions, merge rates, and cycle time to quantify how Copilot impacts development velocity.
GitHub Copilot coding agent now supports Windows development environments//GitHub's autonomous Copilot coding agent can now run on Windows development environments, powered by GitHub Actions, allowing developers working on Windows projects to get AI-assisted pull requests with verified builds and passing tests. The feature requires configuration via a `copilot-setup-steps.yml` file and is recommended for use with self-hosted or larger runners due to firewall compatibility constraints.
GitHub Copilot coding agent now supports code referencing with public repository tracking//The Copilot coding agent now integrates code referencing capabilities, automatically highlighting and linking to matching public code found in GitHub repositories. This helps developers understand code provenance and applicable licenses while working with the autonomous agent.
Sim releases v0.5.93 with audit logging and enterprise access controls//Sim v0.5.93 introduces a comprehensive audit logging system for enterprise environments, along with new access control features for self-hosted deployments. The release adds domain allowlisting for MCP integrations and restricts visibility of usage limits to non-admin enterprise members.
Claude Opus 4.6 now available in GitHub Copilot across Visual Studio, JetBrains, Xcode, and Eclipse//Anthropic's Claude Opus 4.6 model is now generally available to GitHub Copilot users across multiple IDEs and platforms. The model is accessible to Copilot Enterprise, Business, Pro, and Pro+ subscribers through the chat model picker, with enterprise administrators able to opt in via Copilot settings.
Gemini adds music generation with Lyria 3 model; creates 30-second tracks from text or images//Google DeepMind's Lyria 3 music generation model is now available in the Gemini app, enabling users to create 30-second custom tracks from text descriptions or uploaded photos. The model generates lyrics automatically and includes SynthID watermarking to identify AI-generated content.
featuremodelrelease
LangSmith Agent Builder adds unified chat interface, file uploads, and tool registry//LangSmith Agent Builder now features a central Chat agent with access to all workspace tools, enabling ad hoc task execution without dedicated agent setup. New capabilities include file uploads for data analysis and image processing, the ability to convert conversations directly into reusable agents, and a unified tool management interface with governance controls.
featureplatformapi
Duolingo launches free Practice tab for on-demand skill targeting across all language courses//Duolingo has introduced a new Practice tab that allows learners to choose specific skills and lesson types for targeted practice, now available free to all users. The feature includes five main practice options—Mistakes, Words, Speak, Listen, and Focused Listening—plus access to Stories, Radio, and Adventures content.
Vercel redesigns runtime logs search with structured filters and smart suggestions//Vercel has redesigned the runtime logs search interface in its project dashboard with improved filtering capabilities. The update introduces structured filter pills, intelligent query suggestions based on log data, and better input validation with error handling.
npm CLI v11.10.0 adds bulk OIDC publishing and git dependency security controls//npm CLI v11.10.0 introduces two security-focused features: bulk OIDC trusted publishing configuration for managing multiple packages at once, and a new `--allow-git` flag that lets developers restrict git dependencies during installation. The git flag defaults to `all` for backward compatibility but is expected to become `none` by default in npm v12.
releasesecurityapideprecation
Zed ships split diff view as default in v0.224; maintains alignment at scale//Zed's text editor now defaults to a side-by-side split diff view, showing base files on the left and working copy on the right with synchronized scrolling. Built on Zed's multibuffer architecture, the implementation solves alignment challenges at scale while maintaining the editor's performance standards.
GitHub Desktop 3.5.5 brings native Git hooks support with environment variable access//GitHub Desktop 3.5.5 finally resolves long-standing issues with Git hooks by granting them access to shell environment variables, version managers (nvm), and proper terminal formatting. The release also includes support for Warp terminal on Windows, branch management improvements, and fixes for crashes related to emoji and multibyte characters.
GitHub secret scanning adds extended metadata checks configurable at organization level//GitHub secret scanning now supports extended metadata checks through security configurations, making it easier to manage the feature at scale across organizations and enterprises. Repositories with validity checks enabled will automatically see extended metadata checks activated, displaying additional context like secret owner details, creation/expiry dates, and organizational information from secret providers.
Cloudflare adds cfWorker metric to Server-Timing header for performance debugging//Cloudflare has introduced a new `cfWorker` metric in the Server-Timing header that separately measures time spent executing Workers and their subrequests. This enables developers to more accurately diagnose whether latency originates from Worker code, external API calls, or slow origin servers.
featureapiperformancesdk
Weaviate launches Agent Skills library to guide AI coding assistants on vector database integration//Weaviate has released an open-source Agent Skills library that helps popular AI coding assistants like Claude Code, Cursor, and GitHub Copilot correctly implement Weaviate functionality without hallucinations or legacy syntax errors. The library includes both granular Weaviate-specific scripts and end-to-end project cookbooks for building RAG applications, chatbots, and multimodal retrieval systems.
featureintegrationsdkopen-sourceplatform
Notion adds Library: full-page view to organize workspace content as sidebar grows//Notion introduces Library, a new full-page view designed to help users manage cluttered sidebars by organizing Teamspaces, Recents, Favorites, Private pages, and Shared pages into dedicated tabs. Users can now customize their sidebar to show only daily-use items while keeping less-frequently accessed content readily accessible in Library.
featureplatform
Cursor CLI gains plan handoff to cloud and inline Mermaid diagram rendering//Cursor's CLI now supports handing off generated plans to the cloud for execution, alongside new inline ASCII rendering of Mermaid diagrams. The release also includes quality-of-life improvements and keyboard shortcuts for streamlined navigation.
releasefeatureplatform
Harvey launches BigLaw Bench: Global, expanding AI benchmarking to UK, Australia, and Spain//Harvey has released BigLaw Bench (BLB): Global, a new dataset that extends its AI benchmarking framework to measure model performance on legal tasks across multiple jurisdictions. The initial release more than doubles the original benchmark's size by adding localized benchmarks for the UK, Australia, and Spain, designed in collaboration with expert data company Mercor.
releasemodelplatformfeature
LlamaIndex 0.14.15 adds multimodal support and recursive LLM types//LlamaIndex core 0.14.15 introduces multimodal type operations, recursive LLM type support, and enhanced prompt templating for multimodal content. The release also adds retry/error handling to BaseExtractor and includes multiple bug fixes across embeddings, LLM integrations, and memory components.
NVIDIA releases Nemotron-Nano-9B-v2-Japanese, tops sub-10B category on Nejumi Leaderboard//NVIDIA has released a Japanese-optimized version of its Nemotron-Nano-9B-v2 model, achieving state-of-the-art performance in the under-10B parameter category on Japan's Nejumi Leaderboard 4. The model combines advanced Japanese language understanding with proven agent capabilities in a lightweight, deployable package designed for enterprise on-premises deployment.
Neon launches Cursor plugin for AI-assisted Postgres workflows//Neon has released an official plugin for Cursor's new marketplace, giving the AI code editor live access to Neon organizations and structured database management capabilities. The plugin bundles Neon's MCP Server and skills, allowing developers to leverage Cursor's AI assistance for database tasks directly within the editor.
Vercel Agent now suggests automatic build fixes in pull requests//Vercel Agent can now automatically detect build failures and suggest code fixes directly in GitHub pull request reviews and the Vercel Dashboard. The AI-powered feature analyzes build logs and your codebase to recommend specific changes for resolving errors before merging.
Vercel streaming runtime logs exports now enable background downloads without blocking dashboard//Vercel has improved its runtime logs export feature to stream downloads directly to browsers, allowing users to continue using the dashboard while exports process in the background. The update adds new filtering options and increases export limits for Observability Plus subscribers to 100,000 requests.
Vercel adds automated security audits to skills.sh for 60,000+ packages//Vercel has launched automated security audits for skills.sh in partnership with Gen, Socket, and Snyk. The audits now run across over 60,000 skills and provide transparent security reports visible on each skill's detail page, with malicious skills automatically hidden from search and leaderboards.
Claude Sonnet 4.6 now available on Vercel AI Gateway with 1M token context//Anthropic's Claude Sonnet 4.6 is now accessible through Vercel's AI Gateway, bringing advanced agentic capabilities and extended thinking to developers. The model supports a 1M token context window and integrates seamlessly with the Vercel AI SDK using the `anthropic/claude-sonnet-4.6` model identifier.
Recraft V4 image model now available on Vercel AI Gateway//Recraft V4, a text-to-image model built for professional design work, is now accessible via Vercel's AI Gateway with two variants: the standard V4 for faster iteration and V4 Pro for high-resolution print-ready assets. The model features improved photorealism, natural textures, and cleaner composition compared to previous versions.
releasefeaturemodelintegrationapi
Claude Sonnet 4.6 launches in Windsurf with 2-3x promotional pricing//Anthropic's Claude Sonnet 4.6 is now available within Windsurf, Exafunction's AI code editor, with limited-time promotional pricing for self-serve users. The model delivers frontier-level reasoning capabilities at a more cost-effective form factor than Opus, closing the gap on bug detection and code review tasks.
featuremodelreleaseintegration
Ollama v0.16.2 adds web search for Claude cloud models, fixes image generation//Ollama's latest patch release enhances Claude cloud model integration with web search capabilities and fixes a critical issue preventing experimental image generation models from running. The update also adds privacy controls to disable cloud models entirely for sensitive workloads.
GitHub's required reviewer rule exits preview with negation pattern support//GitHub's required reviewer rule for repository rulesets is now generally available, enabling organizations to enforce granular approval policies across branches and files at scale. The feature now includes support for negation patterns using `!` to exclude specific paths, making it easier to define precise policies without overly broad rules.
Cloudflare Access streamlines clientless app access with new policy setting//Cloudflare has introduced an "Allow clientless access" setting that simplifies how organizations grant browser-based access to private self-hosted applications without requiring a device client. Users can now manage this capability directly within their Access application policies rather than creating separate bookmark applications.
Anthropic releases Claude Sonnet 4.6 with 70% preference over predecessor in coding tasks//Claude Sonnet 4.6 is now available as the default model on Claude.ai and Claude Cowork, bringing major improvements in coding, computer use, and long-context reasoning. The model features a 1M token context window in beta and performs on par with or better than the frontier Opus 4.5 model at Sonnet pricing ($3/$15 per million tokens).
releasemodelfeatureapiperformance
Slack launches MCP server and Real-time Search API for AI agent integration//Slack has released two major AI-focused components: the Model Context Protocol (MCP) server designed for LLM consumption, and the Real-time Search API (evolved from Data Access API) for secure, server-side data access. The Real-time Search API now uses granular scopes for fine-grained permission control over channels, direct messages, and multi-party conversations.
apifeatureintegrationmodel
Cursor 2.5 launches plugin marketplace and async subagents for multi-file workflows//Cursor's latest release introduces a plugin marketplace with pre-built integrations from partners like AWS, Figma, and Stripe, plus asynchronous subagents that allow the parent agent to continue working while background tasks execute. The update also adds fine-grained sandbox network access controls for enterprise security policies.
GitHub Copilot coding agent now assignable from Raycast launcher//GitHub's Copilot coding agent can now be invoked directly from Raycast, a free launcher for macOS and Windows. Users can assign issues to Copilot, create new tasks, and track progress—all without leaving the Raycast interface—via the newly updated GitHub Copilot extension.
GitHub Enterprise adds credential management tools for incident response//Enterprise owners can now revoke credentials, block SSO, and review authorized credentials across their organization to respond to security incidents. These new capabilities include the ability to delete tokens and SSH keys across the enterprise, with support for EMU accounts and enterprises with SSO enabled.
Cloudflare Access now supports policies for bookmark applications//Cloudflare has added Access policy support to bookmark applications, allowing administrators to control which users see specific bookmarks in the App Launcher based on identity, device posture, and group membership. Previously, all bookmarks were visible to all users; now they can be restricted to specific audiences while maintaining backward compatibility for unpolicied bookmarks.
Cloudflare adds Docker-in-Docker support to Containers and Sandboxes//Cloudflare's Containers and Sandboxes now support Docker-in-Docker configurations, enabling developers to run full containerized development environments within sandboxed contexts. This capability is useful for testing, CI/CD workflows, and running arbitrary images at runtime.
Claude Sonnet 4.6 now available in GitHub Copilot for all subscription tiers//Anthropic's latest agentic coding model, Claude Sonnet 4.6, is rolling out to GitHub Copilot users across Pro, Pro+, Business, and Enterprise plans. The model excels at agentic coding tasks and search operations, and can be selected via the model picker across all Copilot interfaces including VS Code, VS, github.com, and GitHub Mobile.
GitHub enhances custom properties and rule insights with mandatory value enforcement and SHA filtering//GitHub has shipped two improvements to repository governance: custom properties now support required explicit values (preventing reliance on defaults), and rule insights now feature enhanced filtering with commit SHA support. These updates help organizations enforce stricter repository policies and gain better visibility into how rulesets are applied across their infrastructure.
featureplatform
Harvey Deploys Claude Sonnet 4.6; Scores 87.6% on Legal Tasks//Harvey has integrated Anthropic's Claude Sonnet 4.6 model into its legal AI platform, with the new model achieving an 87.6% score on Harvey's BigLaw Bench evaluation suite. Early testing shows particular strength in numerical precision, litigation strategy ideation, and code-based root cause analysis across transactional and litigation work.
GitHub releases billing usage reports REST API in public preview//Enterprise administrators can now programmatically request and download billing usage reports via a new REST API, enabling automation of the same reports previously available only in the GitHub UI. The API exports data in CSV format and covers all existing billing report types.
Codex CLI 0.106.0 ships with native install script, experimental realtime thread API//Codex CLI version 0.106.0 introduces a direct install script for macOS and Linux, expands app-server v2 with thread-scoped realtime endpoints, and promotes JavaScript REPL to experimental status. The release also enables user input requests in default collaboration mode and improves memory management through diff-based forgetting.
Cloudflare Access now supports policies for bookmark applications//Cloudflare has extended Access policies to bookmark applications, allowing administrators to control which users see specific bookmarks in the App Launcher based on identity, device posture, and other policy rules. This replaces the previous behavior where all bookmarks were visible to everyone in the organization.
Cloudflare Agents SDK v0.5.0 adds retry utilities, protocol message control, and data parts//The Agents SDK v0.5.0 release introduces built-in retry utilities with exponential backoff, per-connection protocol message control for non-standard clients, and a major rewrite of @cloudflare/ai-chat with data parts support and tool approval persistence—all without breaking changes. Synchronous queue and schedule operations no longer require async/await, and improvements address stream resumption race conditions and TypeScript depth errors.
GitHub Copilot in Eclipse gains MCP Registry, todo management, and UI improvements//GitHub Copilot's Eclipse extension has received a major update adding an MCP (Model Context Protocol) server registry for easy integration, a structured todo list tool for complex task management, and several UX improvements including font controls and dark theme enhancements. Developers can now automatically include code selection context in chat and customize agent behavior with new preferences for request limits and commit messages.
Cloudflare Containers and Sandboxes gain Docker-in-Docker support//Cloudflare's Containers and Sandboxes services now support Docker-in-Docker functionality, enabling developers to run full containerized environments within sandboxes. This feature is particularly useful for CI/CD workflows, testing container images, and running isolated development environments.
Cloudflare Access adds streamlined clientless access for private applications//Cloudflare has introduced a new "Allow clientless access" setting that simplifies how organizations grant access to private self-hosted applications without requiring a device client. Users who pass Access policies can now directly access these applications through their App Launcher, with automatic handling of the prefixed Clientless Web Isolation URL.
GitHub Copilot Coding Agent now available in Visual Studio with new task delegation UI//Developers can now delegate coding tasks to GitHub Copilot's autonomous agent directly from Visual Studio, which creates draft pull requests and works asynchronously in the background. The feature requires Visual Studio 2026 (December Update 18.1.0 or later) and is available to Copilot Pro, Pro+, Business, and Enterprise subscribers.
Qwen 3.5 Plus now available on Vercel's AI Gateway with 1M context window//Alibaba's Qwen 3.5 Plus model is now integrated into Vercel's AI Gateway, offering a 1 million token context window and built-in adaptive tool use. The model excels at agentic workflows and multimodal tasks, making it particularly useful for web development and code generation.
Alibaba releases Qwen3.5 model family; 397B variant matches Gemini 3 Pro and Claude Opus 4.5//Alibaba's Qwen3.5 is a new multimodal model family ranging from 27B to 397B parameters with mixture-of-experts architecture. The flagship 397B-A17B variant supports 256K context (expandable to 1M), hybrid thinking/non-thinking modes, and excels in coding, vision, and long-context reasoning, with performance comparable to leading proprietary models.
Cloudflare boosts Markdown for Agents with content encoding support, doubles response limit//Cloudflare has enhanced its Markdown for Agents feature, which automatically converts HTML to Markdown for AI systems. The update raises the origin response limit from 1 MB to 2 MB, removes the requirement for content-length headers, and adds support for content-encoded responses.
Cloudflare WAF adds protections for Zimbra LFI and Vite path traversal CVEs//Cloudflare's Web Application Firewall now includes new detection rules for two critical vulnerabilities: CVE-2025-68645 affecting Zimbra Collaboration Suite and CVE-2025-31125 in Vite's development server. Both new rules are set to block attacks by default.
securitybugfix
ElevenLabs expands ElevenAgents with versioning, RAG tools, and content guardrails//ElevenLabs released significant updates to its ElevenAgents API, introducing agent versioning, a new documentation search tool for RAG, MCP tool support, and configurable content moderation guardrails. The update also includes new endpoints for tracking conversation users and expanded SDK support across Python, JavaScript, and widget packages.
releasefeatureapisdk
SGLang-Diffusion v2 Adds Token-Level Sharding and Parallel VAE for Video Generation//SGLang-Diffusion releases advanced optimizations for production-ready video generation, including token-level sequence sharding, distributed VAE encoding/decoding, and fixed Cache-DiT serving stability. The updates focus on eliminating memory bottlenecks and reducing communication overhead across multi-GPU setups.
Cloudflare WAF adds detections for Zimbra LFI and Vite path traversal vulnerabilities//Cloudflare's Web Application Firewall has released new rule detections for CVE-2025-68645 affecting Zimbra Collaboration Suite and CVE-2025-31125 affecting Vite's development server. Both vulnerabilities now trigger automatic blocking by default in the Cloudflare Managed Ruleset.
securityfeatureplatform
Ollama v0.16.2-rc0 fixes mlxrunner model loading and image generation//This release candidate addresses several issues with the mlxrunner component, including fixes for loading GLM4 MOE Lite models, diffusion model loading, and the --imagegen flag. The update resolves functionality gaps that were preventing proper model execution in the mlxrunner backend.
Cloudflare open-sources ecdysis, a Rust library for zero-downtime service restarts//After five years of production use, Cloudflare has open-sourced ecdysis, a Rust library that enables graceful restarts of network services without dropping connections or refusing requests. The library implements a fork-based approach pioneered by NGINX, allowing parent and child processes to seamlessly share socket file descriptors during upgrades.
Vercel CDN adds stale-if-error cache directive for resilient error handling//Vercel's CDN now supports the stale-if-error Cache-Control directive, allowing developers to specify how long stale cached responses can be served when the origin fails. This enables applications to maintain availability during upstream service outages or network errors.
OpenAI introduces Lockdown Mode and Elevated Risk labels to mitigate prompt injection attacks//OpenAI is rolling out two new security protections to help users and organizations defend against prompt injection attacks: Lockdown Mode, an optional advanced security setting that restricts how ChatGPT interacts with external systems, and standardized "Elevated Risk" labels for features that may introduce additional security concerns. Lockdown Mode is now available for enterprise customers and will be rolled out to consumers in the coming months.
OpenAI introduces credit fallback system for Codex and Sora; users can exceed rate limits by purchasing credits//OpenAI has launched a hybrid access model combining real-time rate limits with a credit purchase system, allowing users to seamlessly continue using Codex and Sora once they hit rate limits. The system uses a decision waterfall that evaluates access from multiple layers—rate limits, free tiers, credits, and enterprise entitlements—while maintaining provably correct, auditable billing with idempotent transactions.
OpenAI releases GABRIEL, an open-source toolkit for quantifying qualitative research data//OpenAI's Economic Research Team has released GABRIEL, an open-source Python library that uses GPT to automatically convert unstructured text and images into quantitative measurements. The toolkit enables researchers to analyze qualitative data at scale, reducing manual labeling work and allowing social scientists to study large datasets of syllabi, interviews, social media, and other rich qualitative sources.
Sim Studio v0.5.89 adds Google Books integration and Opus 4.5 support//Sim Studio v0.5.89 brings new integrations and platform improvements, including Google Books support and a switch to Opus 4.5 as the default model. The release also includes refinements to the tool input system, better S3 handling, and fixes for terminal reconnection and change detection.
releasefeatureintegrationbugfixplatform
Olmix framework enables efficient data mixing for language model development//AI2 releases Olmix, an open-source framework designed to optimize how training data from multiple sources—web text, code, PDFs, math—are combined during language model development. The framework provides empirically grounded configuration defaults and mixture reuse techniques that reduce compute costs while achieving 12% better downstream performance.
open-sourcefeatureplatform
Ollama v0.16.1 improves installer UX and image generation timeouts//Ollama releases v0.16.1 with quality-of-life improvements to installation and model configuration. Changes include smarter password prompts for macOS curl installations, installation progress visibility on Windows, and respect for the OLLAMA_LOAD_TIMEOUT variable in image generation models.
GitHub adds configurable profile name visibility for enterprise and organization admins//Enterprise and organization admins can now control whether profile names (first and last names) appear alongside usernames across GitHub, applying the setting at either the enterprise or organization level. The feature includes three enterprise-wide options: enable everywhere, disable everywhere, or let organizations decide independently.
Workers VPC now supports Cloudflare Origin CA certificates for HTTPS connections//Workers VPC has added support for Cloudflare Origin CA certificates when connecting to private services over HTTPS. Previously limited to publicly trusted certificate authorities, developers can now use free Origin CA certificates on origin servers within private networks.
Cloudflare adds GLM-4.7-Flash model and TanStack AI support to Workers AI//Cloudflare has launched GLM-4.7-Flash, a multilingual AI model with 131K token context and multi-turn tool calling, alongside new TanStack AI adapters and enhanced Vercel AI SDK support. The updates enable developers to build and run AI agents entirely at the edge with expanded capabilities including transcription, speech synthesis, and document reranking.
GitHub Copilot in JetBrains IDEs gains Agent Skills in public preview//GitHub Copilot for JetBrains IDEs now supports Agent Skills in public preview, allowing developers to customize Copilot for specific workflows and reduce repeated setup. The update also includes inline chat improvements, refined settings management, and quality enhancements for better stability and reliability.
GitHub launches Agentic Workflows in technical preview, enabling AI-driven repository automation in Markdown//GitHub Agentic Workflows let developers automate repository tasks using AI agents within GitHub Actions by writing workflows in plain Markdown instead of YAML. The feature, available via the `gh aw` CLI extension, supports natural language automation for issue triage, PR reviews, CI failure analysis, and repository maintenance with security-first defaults including read-only permissions and sandboxed execution.
releasefeatureapiplatformintegration
Railway releases AI agent for canvas, Postgres metrics, and network flow visualization//Railway is shipping three major features: a conversational AI agent for infrastructure management, dedicated Postgres database metrics with query statistics, and network flow visualization showing real-time traffic between services. The AI agent debuts in Priority Boarding, while Postgres metrics and network flows graduate to general availability.
GitHub Copilot coding agent switches to plan-based network routing on Feb 27//GitHub is updating the Copilot coding agent's network configuration to use subscription-based routing, with different API endpoints for Business, Enterprise, and Pro users. Teams running the agent on self-hosted or Azure private networked runners must update their firewall rules by February 27, 2026 to avoid task failures.
featureapibreaking-change
Hugging Face releases CUDA kernel skill for Claude and Codex agents//Hugging Face has released an agent skill that enables AI coding agents like Claude and Codex to automatically generate production-ready CUDA kernels for deep learning models. The skill packages domain expertise about GPU architecture, PyTorch integration patterns, and optimization strategies into a reusable module that agents can apply to real transformers and diffusers pipelines.
Linear adds advanced filters and private team issue sharing in February release//Linear unveiled advanced filtering capabilities with AND/OR logic for refining searches and views, alongside private team issue sharing for Enterprise customers. The update also adds mobile project creation, Gemini Enterprise integration, and numerous performance improvements and bug fixes.
Cloudflare Python SDK v5.0.0-beta.1 introduces major breaking changes and 40+ new API resources//Cloudflare released the first beta version of Python SDK v5.0.0, featuring significant breaking changes driven by OpenAPI schema improvements and code generation updates. The release adds over 40 new API resources including AI-powered features, brand protection tools, D1 database management, and Real-time Kit integrations, alongside general fixes for type inference, request handling, and response parsing.
Cloudflare launches GLM-4.7-Flash on Workers AI with multi-turn tool calling for edge agents//Cloudflare has added GLM-4.7-Flash, a multilingual model with 131k token context, to Workers AI alongside new TanStack AI adapters and expanded workers-ai-provider capabilities. Developers can now build and deploy AI agents entirely at the edge with support for transcription, text-to-speech, and reranking.
Cloudflare Workers VPC now supports Origin CA certificates for private HTTPS connections//Workers VPC has expanded its TLS certificate support to include Cloudflare Origin CA certificates, enabling developers to encrypt traffic between VPC tunnels and private origin servers without relying on public certificate authorities. This builds on Workers VPC's existing support for publicly trusted CAs like Let's Encrypt and DigiCert.
Cloudflare Python SDK v5.0.0-beta.1 adds 40+ new API resources with breaking changes//Cloudflare released Python SDK v5.0.0-beta.1 featuring 40+ new API resources including AI Gateway, R2 Data Catalog, and Real-time Kit integrations. The beta includes significant breaking changes across 15+ existing resources due to OpenAPI schema improvements, requiring developers to review the migration guide before upgrading.
releasesdkbreaking-changeapi
Ollama v0.16.0 adds GLM-5 and MiniMax-M2.5 models, introduces app launcher//Ollama releases v0.16.0 with support for two new state-of-the-art models: GLM-5, a 40B-parameter reasoning model designed for complex systems tasks, and MiniMax-M2.5, optimized for productivity and coding. The release also introduces a new `ollama launch` command for seamlessly integrating models with applications.
Google upgrades Gemini 3 Deep Think; achieves 84.6% on ARC-AGI-2 and gold-medal math performance//Google has released a major upgrade to Gemini 3 Deep Think, a specialized reasoning mode designed for science, research, and engineering challenges. The updated model is now available to Google AI Ultra subscribers in the Gemini app and available via API for select researchers, engineers, and enterprises through early access.
Browserbase joins Vercel Marketplace with one-click browser automation for AI agents//Browserbase is now integrated into the Vercel Marketplace, allowing developers to easily connect AI agents to remote browsers via Chrome DevTools Protocol without managing infrastructure. The integration includes unified billing, support for Vercel Sandbox and AI Gateway, and Web Bot Auth capabilities for seamless agent interaction with Vercel-hosted deployments.
Vercel blocks deployments with vulnerable next-mdx-remote versions by default//Vercel now automatically rejects new deployments containing vulnerable versions of the next-mdx-remote package (CVE-2026-0969). The blocking behavior can be disabled via environment variable if needed, though upgrading to patched versions is strongly recommended.
Vercel AI Gateway adds MiniMax M2.5 model support for full-stack development//MiniMax M2.5 is now available through Vercel's AI Gateway, offering improved planning and code generation capabilities for full-stack projects. The model can be accessed via the AI SDK with automatic retry, failover, and observability features built-in.
featureapimodelintegration
OpenAI's GPT-5.3-Codex-Spark now available in Windsurf via Arena Mode//OpenAI's GPT-5.3-Codex-Spark model is now accessible to Windsurf users through the Fast Arena and Hybrid Arena battle groups. The model is currently exclusive to these arena features and not available in the standard model selector.
OpenAI releases GPT-5.3-Codex-Spark, optimized for real-time coding at 1000+ tokens per second//OpenAI has released GPT-5.3-Codex-Spark, a smaller, faster variant of GPT-5.3-Codex designed for real-time interactive coding tasks. Running on Cerebras' specialized hardware, the model delivers over 1000 tokens per second while maintaining strong performance on coding benchmarks, now available as a research preview to ChatGPT Pro users.
releasefeaturemodelapiperformance
AI2 launches AutoDiscovery, an AI tool that autonomously explores datasets and surfaces novel research hypotheses//AI2 has released AutoDiscovery as an experimental feature in AstaLabs, enabling researchers to automatically explore datasets without predefined hypotheses. The system generates novel research directions with reproducible code and statistical validation, already showing promise in oncology, marine ecology, and social science applications.
featureplatformrelease
Allen AI releases AutoDiscovery, automated hypothesis generation tool now available in AstaLabs platform//AutoDiscovery is an AI-powered system that autonomously generates and tests scientific hypotheses on structured datasets, surfacing novel research directions without requiring researchers to specify questions in advance. The tool uses Bayesian surprise and Monte Carlo Tree Search to prioritize high-value experiments, and is now available as an experimental feature in the Asta platform.
releasefeatureplatformapi
vLLM v0.16.0rc3 fixes MTP accuracy issue with GLM-5 model//vLLM releases a release candidate patch addressing an accuracy bug in Multi-Token Prediction (MTP) functionality for the GLM-5 model. This fix ensures more reliable inference results for users running GLM-5 on vLLM infrastructure.
Sim v0.5.88 adds Gemini Deep Research integration and Confluence enhancements//Sim Studio releases v0.5.88 with support for Google's Gemini Deep Research via the Interactions API and expanded Confluence integration capabilities. The update also includes performance improvements and bug fixes for variables and hotkey handling.
Cloudflare Terraform provider v5.17.0 adds AI Search support and Workers placement controls//Cloudflare's Terraform v5 provider reaches v5.17.0 with new capabilities for AI Search, enhanced Workers Script placement controls, and 25+ bug fixes. The provider continues its 2-3 week release cadence with an expanding list of stable resources and a migration tool planned for March 2026.
releasesdkapifeature
Modal SDK 1.3.3 adds billing reports, async warnings, and object ID methods//Modal's Python SDK v1.3.3 introduces a new `modal billing report` CLI and general availability billing report API for Team and Enterprise workspaces, along with convenience methods for referencing Queues and Dicts by object ID. The release enables async usage warnings by default to catch performance issues in async contexts.
Cloudflare Radar adds content type insights for AI bot traffic via new API dimension//Cloudflare Radar now provides visibility into the content types served to AI bots and crawlers through a new `content_type` dimension and filter. The feature categorizes MIME types across 16 content categories and is available through two new API endpoints and the Radar dashboard.
featureapiplatform
Deepgram releases self-hosted February 2026 container images with engine v3.107.0//Deepgram has released updated container images for its self-hosted speech recognition platform, including engine version 3.107.0 and updated API, license proxy, and billing components. The release requires NVIDIA driver version 570.172.08 or higher.
Cloudflare launches Markdown for Agents, auto-converts HTML via content negotiation//Cloudflare's network now automatically converts HTML to Markdown for AI systems that request it via Accept headers, reducing token consumption and improving AI processing efficiency. The feature is available for enabled zones and uses standard HTTP content negotiation.
Cloudflare WAN dashboard now displays Anycast IPs for tunnels//Cloudflare WAN customers can now view their Anycast IP addresses directly in the dashboard when configuring IPsec or GRE tunnels. This replaces the previous workflow requiring manual API calls or onboarding retrieval, streamlining tunnel setup and reducing configuration errors.
featureplatform
Cursor launches long-running agents to handle complex multi-step tasks autonomously//Cursor's new long-running agents can work autonomously over extended periods to complete larger, more complex development tasks without human intervention. Available now in research preview for Ultra, Teams, and Enterprise users, the agents reportedly complete work that was previously too difficult for standard agents, resulting in larger and more complete pull requests.
featureplatformrelease
Zed launches Theme Builder, a visual editor for crafting code editor themes without JSON//Zed has released Theme Builder, a web-based visual theme editor that eliminates the need to manually edit JSON files when customizing the editor's 200+ color tokens. The tool features a live preview, an inspector to identify which tokens control specific elements, color linking between related tokens, and import/export functionality.
Cloudflare Radar adds content type dimension for AI bot traffic analysis//Cloudflare Radar now provides content type insights for AI crawler traffic through new API endpoints and dashboard views. The update shows the distribution of MIME types being served to AI bots, helping developers understand what content types are being accessed by AI crawlers.
Cloudflare enhances logo matching with configurable thresholds starting at 75%//Cloudflare's Security Center now offers enhanced logo matching capabilities for brand protection, enabling detection of subtle brand asset variations and impersonations. The update introduces configurable match thresholds, visual match scores with color-coded severity indicators, and direct logo previews in the dashboard.
Cloudflare enhances logo matching for brand protection with configurable detection thresholds//Cloudflare's Security Center now offers configurable logo matching thresholds starting at 75% similarity, allowing users to detect subtle brand impersonations and variations. The update includes visual match scores, color-coded severity indicators, and direct logo previews to streamline threat triage.
Vercel AI Gateway now supports GLM-5 model with native agentic capabilities//GLM-5 from Z.AI is now available through Vercel's AI Gateway without requiring additional provider accounts. The model features multiple thinking modes, improved long-range planning, and enhanced multi-step task handling optimized for agentic workflows and structured data extraction.
Vercel Sandbox adds SNI-based egress firewall for controlling outbound network access//Vercel Sandbox now supports advanced egress firewall policies using Server Name Indication (SNI) filtering and CIDR blocks to restrict outbound connections. The feature allows developers to lock down network access to specific hosts, preventing unauthorized data exfiltration or unintended API calls when running untrusted or AI-generated code.
Vercel Flags moves to public beta with native Next.js and SvelteKit support//Vercel has launched Vercel Flags, a built-in feature flag service for the Vercel platform that enables developers to create, manage, and deploy feature flags with targeting rules directly from the dashboard. The service integrates natively with Next.js and SvelteKit, supports the OpenFeature standard for multi-framework use, and is priced at $30 per million flag requests.
AI2 launches MolmoSpaces, an open simulation platform for embodied AI with 230,000 scenes and 42 million grasps//MolmoSpaces is a large-scale, open-source ecosystem for training and evaluating embodied AI systems, unifying over 230,000 indoor scenes, 130,000+ object models, and 42 million annotated robotic grasps. The platform features physics-grounded simulation, a systematic benchmark for measuring generalization across multiple axes, and compatibility with major simulators like MuJoCo, ManiSkill, and NVIDIA Isaac.
Sim releases v0.5.87 with authentication workflow fix//Sim has released v0.5.87, a minor patch addressing a bug in the authentication workflow system handler. This update ensures more reliable auth handling for users relying on Sim's workflow system.
Sim releases v0.5.86 with copilot MCP server-side execution and enterprise configuration//Sim v0.5.86 introduces server-side copilot execution with MCP support and enterprise configuration options. The release also adds skill icons to trace spans, increases workflow log limits from 1k to 5k, and includes support for copilot-only Helm deployments alongside numerous bug fixes and integrations.
Cloudflare Workers lifts 1000-subrequest limit to 10,000 by default; configurable to 10 million//Cloudflare Workers no longer enforces a hard 1000-subrequest limit per invocation, enabling long-running workloads like websockets and Workflows to operate without hitting the barrier. Paid plans now default to 10,000 subrequests and can be configured up to 10 million via Wrangler configuration.
featureapiplatform
Slack introduces task cards and plan blocks for displaying AI app workflows//Slack has released three new Block Kit components that enable developers to display app and agent actions directly to users. The updates include task card blocks, URL source elements, and plan blocks, alongside new streaming API parameters for controlling how tasks are presented.
Cloudflare Workers lifts 1,000 subrequest limit; paid plans now support up to 10 million per invocation//Cloudflare has removed the hard 1,000 subrequest limit that previously constrained Workers invocations, with paid plans now defaulting to 10,000 subrequests and allowing configuration up to 10 million. This change particularly benefits long-running operations like WebSocket connections on Durable Objects and Workflows that frequently exceeded the prior constraint.
GitHub Mobile gains Copilot model picker for on-the-go coding sessions//Copilot Pro and Pro+ subscribers can now select from multiple AI models directly in GitHub Mobile when starting a coding agent session, without switching to desktop. The feature supports auto-selection or manual choice between Claude Opus/Sonnet and OpenAI GPT models.
Cloudflare Vite plugin adds seamless React Server Components support via childEnvironments//The Cloudflare Vite plugin now integrates the official Vite React Server Components plugin (@vitejs/plugin-rsc), enabling developers to build RSC applications on Cloudflare Workers. A new childEnvironments option allows using multiple module graphs within a single Worker environment.
Cloudflare Vite plugin adds React Server Components support with child environments//The Cloudflare Vite plugin now integrates the official Vite React Server Components plugin, enabling developers to use multiple environments within a single Worker. A new `childEnvironments` option allows the parent environment to import modules from child environments, supporting typical RSC patterns used by frameworks like React Router.
featuresdkintegration
Neon releases add-mcp CLI to install MCP servers across coding agents and editors//Neon has launched add-mcp, a CLI tool that simplifies MCP server installation across multiple coding environments like Claude Code, Cursor, and VS Code. The tool detects which agents are installed and automatically configures MCP servers in the appropriate config files, eliminating manual setup repetition across different tools.
Vercel's MCP server adds runtime log access for AI agents//Vercel has expanded its Model Context Protocol (MCP) server with a new `get_runtime_logs` tool that allows AI agents to access runtime logs from Vercel Functions. The feature enables agents to debug failing requests, inspect function output, and search logs for specific errors across preview and production deployments.
Vercel Adds Sign in with Apple Support//Vercel's login experience now supports Sign in with Apple, allowing users with Apple accounts to authenticate faster. Accounts are automatically linked if the Apple email matches the Vercel account email, with manual connection available in account settings for mismatched emails.
PostHog joins Vercel Marketplace with feature flags and A/B testing integration//PostHog is now available as an official integration in the Vercel Marketplace, enabling developers to declare and manage feature flags directly in their Next.js applications. The integration leverages the Flags SDK and provides real-time feature toggles, percentage-based rollouts, and A/B testing capabilities within Vercel workflows.
Vercel's logs CLI command gains historical querying and agent-friendly filters//The `vercel logs` command has been rebuilt with powerful querying capabilities designed for agent workflows. It now supports filtering by project, deployment ID, request ID, and arbitrary strings, with automatic git context detection for simplified debugging.
featureapiplatform
Replicate's MCP server now auto-discoverable via official registry//Replicate has implemented auto-discovery for its Model Context Protocol (MCP) server by publishing metadata to the official MCP Registry. This enables AI developers to find and install the Replicate MCP server directly through client applications like VS Code and Claude Desktop.
Supabase acquires Hydra team to build open data warehouse for Postgres//Supabase is welcoming Joe Sciarrino, co-creator of Hydra, to lead the development of Supabase Warehouse and an Open Warehouse Architecture initiative. The team will leverage pg_duckdb, an open-source Postgres extension that accelerates analytics queries by 600x, to enable serverless analytics workflows on Postgres with object storage integration.
releasefeatureplatformintegrationopen-source
LangSmith now available in Google Cloud Marketplace with consolidated billing//LangChain's LangSmith agent engineering platform is now available for procurement through Google Cloud Marketplace, enabling seamless billing integration and draw-down of existing Google Cloud commitments. The platform provides observability, evaluation, deployment, and no-code agent building capabilities with native integrations across Vertex AI, Gemini, and GCP databases.
GitHub Dependabot audit logs now track configuration changes for security and compliance//GitHub has added two new audit log event types to track Dependabot configuration changes at the organization and enterprise level. The new events log when Dependabot vulnerability updates or self-hosted runner configurations are enabled or disabled, capturing the actor and timestamp of each change.
featureapisecurityplatform
Unsloth ships MoE training kernels with 12x speedup and 35% lower VRAM usage//Unsloth introduced custom Triton kernels and optimizations for training Mixture of Experts (MoE) language models, delivering 12x faster training speeds with over 35% reduction in VRAM consumption and support for 6x longer context windows. The update supports popular MoE models including Qwen3, DeepSeek R1/V3, and GPT-OSS, working across data-center and consumer GPUs.
Cloudflare WAF adds detection rules for Zimbra and Vite vulnerabilities//Cloudflare is deploying two new security detection rules to its Web Application Firewall on February 16, 2026. The new rules target vulnerabilities in Zimbra (local file inclusion) and Vite (WASM import path traversal), launching initially in log mode to detect threats without blocking traffic.
Cloudflare disables fake Google Bot detection rule in WAF//Cloudflare has disabled the "Anomaly:Header:User-Agent - Fake Google Bot" rule in its Managed Ruleset, changing it from a blocking action to disabled status. This prevents the WAF from blocking requests that appear to be spoofed Google Bot traffic.
OpenAI updates GPT-5.2 Instant, launches GPT-5.3-Codex with 25% faster performance//OpenAI has released multiple model updates including an improved GPT-5.2 Instant with more measured responses and clearer output on advice-seeking questions. The company also introduced GPT-5.3-Codex, a unified coding model combining code generation with general-purpose reasoning, delivering 25% faster performance and new benchmark highs.
OpenAI adds Batch API support for GPT image models//OpenAI has extended its Batch API to support GPT image generation models, enabling developers to process multiple image generation requests asynchronously at reduced costs. This addition complements existing batch capabilities for text and other specialized models, allowing for more efficient large-scale image generation workloads.
Sim Studio v0.5.85 improves deploy modal UI and fixes trigger ID resolution//Sim Studio releases v0.5.85 with UI improvements to the deployment modal and terminal, along with a bug fix for ID resolution in tools using trigger mode. The minor patch addresses both user experience and underlying tool functionality.
releasebugfixfeature
NVIDIA TensorRT LLM adds AutoDeploy beta, automating model-to-inference optimization pipeline//NVIDIA announced AutoDeploy, a new beta feature in TensorRT LLM that automatically compiles PyTorch models into inference-optimized graphs without manual reimplementation. The compiler-driven approach enables deployment of new architectures at launch while supporting over 100 text-to-text LLMs, VLMs, and state space models.
Vercel introduces prefixed token formats and automatic secret scanning for leaked credentials//Vercel now automatically revokes API credentials accidentally committed to public repositories through GitHub secret scanning. All token and API key formats have been updated with identifying prefixes (vcp, vci, vca, vcr, vck) to improve security awareness and credential management.
OpenAI deploys ChatGPT to Pentagon's GenAI.mil platform for 3 million personnel//OpenAI is bringing a custom version of ChatGPT to GenAI.mil, the Department of Defense's secure enterprise AI platform used by military and civilian personnel. The deployment includes built-in safety controls and data isolation safeguards to protect sensitive government information while enabling service members to access AI capabilities for operational and administrative tasks.
OpenAI begins testing ads in ChatGPT's free tier, keeping conversations private from advertisers//OpenAI is starting a controlled test of advertisements in ChatGPT for logged-in free and Go tier users in the U.S., with paid tiers remaining ad-free. The company emphasizes that ads won't influence ChatGPT's answers, user conversations remain private from advertisers, and users have full control over ad preferences.
ElevenLabs reorganizes into platform structure, launches custom guardrails and WhatsApp outbound API//ElevenLabs restructured its product portfolio into three distinct platforms—ElevenAgents, ElevenCreative, and the new ElevenAPI—while graduating global server routing out of beta. The Agents Platform now supports custom guardrails, WhatsApp outbound messaging, improved TTS models, and enhanced error handling for webhook tools.
Transformers.js v4 Preview Debuts on NPM with New WebGPU Runtime and 10x Build Speed Gains//Transformers.js v4 preview is now available on NPM under the `@next` tag, bringing a complete rewrite with a new WebGPU runtime and major performance improvements. The release includes support for ~200 model architectures, cross-runtime compatibility (browsers, Node, Bun, Deno), and architectural optimizations that deliver 4x speedups for embedding models and 10x faster builds.
Cloudflare Sandbox SDK adds PTY terminal support for browser-based shell access//Cloudflare's Sandbox SDK now supports pseudo-terminal (PTY) passthrough via WebSocket, enabling interactive browser-based terminals to connect to sandbox shells. The update includes a new xterm.js addon with automatic reconnection and output buffering, plus support for multiple isolated terminals per sandbox.
Cloudflare AI Crawl Control adds path grouping, referral analytics, and data transfer metrics//Cloudflare has enhanced its AI Crawl Control analytics with new visualization and filtering capabilities to help developers understand crawler behavior. Updates include path pattern grouping, referral traffic trends, bandwidth consumption metrics, and image export functionality.
GitHub Copilot launches GPT-5.3-Codex with 25% faster performance on coding tasks//OpenAI's latest agentic coding model, GPT-5.3-Codex, is now generally available across all GitHub Copilot tiers. The model delivers improved reasoning and execution for complex workflows, with benchmarks showing significant performance gains over its predecessor.
GitHub Apps gain fine-grained permissions for Enterprise Teams API management//GitHub Apps can now access Enterprise Teams API endpoints using fine-grained permissions with read and write access levels, replacing the previous requirement for personal access tokens. This enables GitHub Enterprise administrators to programmatically manage teams at scale more securely through app integrations.
Cloudflare Agents SDK v0.4.0 adds readonly connections, MCP OAuth customization, and x402 v2 support//The Agents SDK v0.4.0 introduces readonly WebSocket connections for dashboards and spectator views, custom OAuth providers for MCP server authentication, and upgrades the MCP SDK to prevent cross-client response leakage. The release also completes migration to x402 v2 with new network identifiers and lazy server initialization.
releasefeatureapisecuritybreaking-changesdk
Bun v1.3.9 adds parallel/sequential script running, HTTP/2 upgrades, and 28% faster React markdown rendering//Bun 1.3.9 introduces `bun run --parallel` and `--sequential` for managing multiple package.json scripts with workspace support, fixes HTTP/2 connection upgrade patterns, and improves markdown rendering performance. The release also includes Symbol.dispose support for mocks, NO_PROXY environment variable fixes, and improved ARM64 CPU compatibility.
releasefeatureperformancebugfix
Ollama v0.15.6 fixes context limits and adds automatic model downloads//Ollama releases v0.15.6 with three key bugfixes addressing context handling and model management. The update improves reliability when launching AI models and enhances the user experience by automatically downloading missing models instead of failing.
Sim v0.5.84 released with request sanitization security fix//Sim Studio has released v0.5.84, a patch update focused on security improvements. The release includes a critical request sanitization fix in the models component to address potential vulnerabilities.
Sim v0.5.83 adds skills to agent blocks, airweave integration, and improved Azure support//Sim Studio released v0.5.83 with new agent skills functionality, an Airweave block integration, and enhanced Azure-Anthropic support. The update also includes improvements to model configuration, workflow preview rendering, and various bug fixes for authentication and worker pool performance.
Claude Opus 4.6 Fast mode launches in GitHub Copilot preview with 2.5x faster output speeds//GitHub has released a preview of fast mode for Claude Opus 4.6 in Copilot, delivering output token speeds up to 2.5x faster while maintaining the same intelligence. The feature is rolling out gradually to Copilot Pro+ and Enterprise users with a promotional period ending February 16, 2025, after which a 30x premium request multiplier will apply.
releasefeaturemodelperformance
Vercel's v0 Moves from Prototyping to Production with Real Codebases and Git Workflows//Vercel has rebuilt v0 into a platform for shipping production applications rather than quick prototypes, adding support for real codebases, Git workflows, security controls, and integration with live databases. The update positions v0 as infrastructure for both traditional apps and future AI agent workflows.
Sanity CMS now available on Vercel Marketplace with native integration//Sanity is now available as a native integration on the Vercel Marketplace, allowing developers to install, configure, and manage the headless CMS directly from the Vercel dashboard. The integration eliminates manual API token setup and environment variable configuration, keeping content management within the existing Vercel workflow.
Vercel Sandbox SDK adds file retrieval methods for simplified data extraction//The Vercel Sandbox SDK now includes `downloadFile()` and `readFileToBuffer()` methods that streamline retrieving generated files from isolated sandbox environments. This eliminates the need for manual stream handling and custom piping when working with files created inside sandbox VMs.
Sim v0.5.82 adds Slack file attachments, fixes resolver and integrations//Sim releases v0.5.82 with support for file attachments in Slack webhook triggers. The update includes fixes for response formatting in deactivated branches, alignment of Linear API integrations, and executor loop handling.
Supabase Auth adds X / Twitter OAuth 2.0 provider with PKCE support//Supabase Auth now supports X / Twitter authentication via OAuth 2.0, replacing the legacy OAuth 1.0a flow. The new provider includes modern security features like PKCE support and provides developers with a straightforward setup process through the Supabase dashboard.
featureapiintegrationsecurity
Ollama v0.15.5 adds two new models and improves agentic coding support//Ollama releases v0.15.5 with two new models for coding and document understanding, plus improved `ollama launch` support for sub-agents and context length auto-tuning based on available VRAM. The release also fixes token handling bugs in the API.
Cloudflare Workers Observability dashboard adds data visualization, exports, and shareable links//Cloudflare has released major updates to the Workers Observability dashboard, making it easier for developers to debug applications and collaborate with teams. The new features include data visualization tools, JSON/CSV exports, shareable event links, and improved table customization with keyboard shortcuts.
featureplatformrelease
Deepgram adds OpenAI GPT-5.2 and Google Gemini 3 support to Voice Agent API//Deepgram's Voice Agent API now supports three new LLM models: OpenAI's GPT-5.2 Instant and Thinking variants, plus Google's Gemini 3 Flash. Developers can configure these models in their Voice Agent settings via a simple provider configuration.
featureapimodelintegration
Suno Studio 1.2 adds audio warping, time signature support, and effect removal//Suno Studio, the generative audio workstation for Suno Premier subscribers, now includes tools for advanced audio editing. The update introduces the ability to remove effects from audio clips, warp and quantize timing, create alternate variations, and support non-4/4 time signatures.
CodeQL 2.24.1 improves Maven registry support and query accuracy across seven languages//CodeQL 2.24.1 extends language support including Kotlin 2.3.0, C23, and C++26, while adding Maven private registry support for plugins. The release includes accuracy improvements to static analysis queries, a new prompt injection detection query, and fixes for false positives in buffer overflow detection.
Sim releases v0.5.81 with Azure Anthropic support and Confluence integrations//Sim Studio v0.5.81 adds Azure Anthropic integration with backwards-compatible chat completions API support and Claude Opus 4.6 model access. The release also expands Confluence endpoint coverage and fixes workflow state handling and parameter validation issues.
Linear expands MCP server with initiatives and milestones for Claude, Cursor integration//Linear has significantly expanded its Model Context Protocol (MCP) server to support initiatives, project milestones, and updates—allowing product managers to keep plans synchronized across tools like Claude and Cursor. The update also introduces nested sub-issues display, team retirement functionality, and deprecates the older SSE endpoint in favor of HTTP streams.
releaseapifeatureintegrationdeprecationplatform
SyGra Studio v2.0.0 debuts visual no-code editor for synthetic data generation workflows//ServiceNow has released SyGra Studio, a visual interface for building and executing synthetic data generation pipelines without writing YAML or scripts. The platform enables drag-and-drop workflow composition with live execution monitoring, inline debugging, and real-time cost tracking.
Vercel AI Gateway adds Claude Opus 4.6 with adaptive thinking and 1M context window//Anthropic's Claude Opus 4.6 is now available on Vercel's AI Gateway, bringing the latest flagship model to the unified API platform. The model introduces adaptive thinking, allowing it to intelligently decide when and how much to reason, while supporting an extended 1M token context window for the first time in the Opus line.
OpenAI Launches Trusted Access for Cyber, Commits $10M in API Credits for Defensive Security Work//OpenAI is introducing Trusted Access for Cyber, an identity-based framework that grants priority access to GPT-5.3-Codex and other frontier models for cybersecurity professionals and researchers. The initiative includes $10 million in API credits through the scaled Cybersecurity Grant Program to accelerate vulnerability discovery and remediation in open source and critical infrastructure.
OpenAI Launches Frontier Platform for Enterprise AI Agent Deployment and Management//OpenAI has introduced Frontier, a comprehensive platform designed to help enterprises build, deploy, and manage AI agents at scale. The platform provides AI coworkers with shared business context, integrated execution environments, and governance capabilities—enabling organizations to move beyond isolated agent pilots to production-grade AI systems that work across multiple applications and data sources.
Sim v0.5.80 adds async job execution, execution timeouts, and block locking//Sim Studio v0.5.80 introduces async job queue support for background execution and configurable execution timeouts, alongside UI improvements like block locking and note block preview formatting. The release also includes numerous bug fixes and improvements across OpenAI API integration, MCP workflows, and provider tool calling compatibility.
releasefeatureapibugfixperformance
Ollama v0.15.5-rc4 fixes off-by-one error in token prediction limits//An off-by-one bug in Ollama's numPredict parameter caused users to receive one fewer token than requested and incorrect token statistics. The fix ensures token limits are properly enforced at prediction time rather than batch setup.
releasebugfix
vLLM v0.15.1 ships security patches, RTX Blackwell GPU fixes, and 4x faster torch.compile startup//vLLM's v0.15.1 patch release addresses two critical security vulnerabilities in dependencies (aiohttp and Protobuf) while fixing GPU support for RTX Blackwell workstations. The release also delivers significant performance improvements, including a 4x speedup in torch.compile cold-start times for large models like Llama3-70B.
releasebugfixsecurityperformancemodel
vLLM v0.15.2rc0 fixes TRTLLM attention conflict with KV cache transfer//vLLM releases v0.15.2rc0, a release candidate containing a critical bugfix that disables TRTLLM attention mechanisms when KV cache transfer is enabled. This patch resolves compatibility issues that could arise when both features are active simultaneously.
releasebugfix
Slack CLI v3.12.0 adds hot reloading and pyproject.toml support for Python projects//The Slack CLI now includes file watching and live reloading for Bolt for JavaScript and Python projects, automatically restarting the development server when files change. Python projects gain support for pyproject.toml configuration alongside requirements.txt.
Supabase adds PrivateLink, Ethereum integration, and Claude connector//Supabase released multiple major features in February 2026 including PrivateLink for private AWS connectivity, direct Ethereum blockchain querying via SQL, and official Claude integration. The update also includes a breaking change disabling pg_graphql by default on new projects for improved security posture.
GitHub Actions launches runner scale set client for custom autoscaling across any infrastructure//GitHub introduced a new open-source runner scale set client, a standalone Go-based module that enables custom autoscaling solutions for GitHub Actions runners without requiring Kubernetes. The client works across containers, VMs, and bare metal servers, supporting multiple labels and agentic workflows.
GitHub Actions extends self-hosted runner upgrade deadline to March 16, 2026//GitHub has extended the enforcement timeline for self-hosted runners to upgrade to v2.329.0 by one week. Between February 16 and March 16, 2026, brownout periods will temporarily block configurations from older runners to help identify those needing updates before permanent enforcement begins.
releasebreaking-changeplatform
Deepgram updates Nova-3 multilingual model with improved code-switching accuracy//Deepgram has released an updated Nova-3 multilingual speech-to-text model with lower word error rates across supported languages and significantly improved handling of code-switching scenarios. The improvements are available immediately without requiring API or configuration changes.
modelfeatureperformance
Slack adds delayed events replay feature to recover from outages//Slack has released a new Delayed Events feature that allows developers to replay events their apps may have missed during outages. This helps ensure applications don't lose critical event data when experiencing connectivity issues.
Anthropic releases Claude Opus 4.6 with 1M token context, outperforms GPT-5.2 by 144 Elo points//Anthropic introduced Claude Opus 4.6, its most capable flagship model to date, featuring a 1M token context window in beta and state-of-the-art performance on coding, reasoning, and knowledge work benchmarks. The model is now available on claude.ai, the API, and major cloud platforms with pricing unchanged at $5/$25 per million tokens.
Claude Opus 4.6 now available in GitHub Copilot for all plan tiers//Anthropic's Claude Opus 4.6 is rolling out to GitHub Copilot Pro, Pro+, Business, and Enterprise users. The model excels at agentic coding tasks requiring planning and tool calling, with availability across VS Code, Visual Studio, github.com, GitHub Mobile, and CLI.
GitHub adds pinned comments and anti-spam nudges to Issues//GitHub Issues now supports pinning comments to the top of threads to highlight key decisions and updates, while a new smart nudge encourages users to react or subscribe instead of posting low-quality comments like "+1". These changes aim to reduce noise and improve issue navigation for development teams.
OpenAI releases GPT-5.3-Codex with agentic coding capabilities; achieves new SWE-Bench Pro high and 25% faster performance//GPT-5.3-Codex is OpenAI's most capable coding model to date, combining frontier coding performance with advanced reasoning and agentic capabilities. The model sets new benchmarks on SWE-Bench Pro and Terminal-Bench 2.0, while operating 25% faster than its predecessor and enabling developers to delegate complex, long-running tasks without losing context during iteration.
releasefeaturemodelapi
Harvey integrates Anthropic's Opus 4.6, achieving 90.2% on legal reasoning benchmarks//Anthropic's latest Claude Opus 4.6 model is now available in Harvey, the legal AI platform. Early evaluations show the model scoring 90.2% on Harvey's BigLaw Bench suite with improvements in substantive accuracy, source grounding, and multi-step legal reasoning across litigation and transactional workflows.
GitHub improves pull request Files changed tab with CODEOWNERS validation and 67% faster performance on large diffs//GitHub has shipped significant improvements to its pull request "Files changed" experience, adding CODEOWNERS validation and addressing long-standing performance issues. The update includes optimizations that make large pull requests respond 67% faster to interactions, faster tab navigation (down from 10+ seconds to a few seconds), and better mobile responsiveness.
releasefeatureperformancebugfixplatform
Ollama v0.15.5-rc3 fixes Qwen3 delta net tensor broadcasting issue//This release candidate fixes a critical broadcasting issue in the Qwen3 model's delta network computation where gradient tensors were being multiplied along the wrong axis. The fix reshapes the gradient difference tensor to the correct dimensions for proper multiplication with the key tensor.
bugfixrelease
Kimi K2.5 multimodal VLM now available on NVIDIA GPU-accelerated endpoints with free prototyping access//NVIDIA has made Kimi K2.5, a 1-trillion parameter multimodal vision language model, available for free prototyping on its GPU-accelerated endpoints via build.nvidia.com. The model supports text, image, and video inputs and can be deployed via vLLM or fine-tuned using the NVIDIA NeMo Framework for production use cases.
GitHub Copilot Chat adds visible tool calls and chat export to web interface//GitHub has rolled out three improvements to Copilot Chat on the web: users can now see real-time tool call details as Copilot processes requests, export conversations as JSON or Markdown files, and benefit from enhanced repository search when attaching repos to chat. These updates aim to provide better transparency and usability for developers using Copilot in the browser.
GitHub Copilot v1.109 adds Claude agent support and multi-agent orchestration in VS Code//GitHub Copilot's January release introduces public preview support for Anthropic's Claude agent SDK, enabling developers to delegate tasks directly within VS Code. The update also enhances agent session management with multi-agent orchestration, improved reasoning capabilities, and new security features like terminal command sandboxing.
releasefeatureapiintegrationsecurityplatform
NVIDIA releases Nemotron ColEmbed V2 multimodal models; 8B variant ranks #1 on ViDoRe V3 benchmark//NVIDIA has released the Nemotron ColEmbed V2 family, a set of late-interaction multimodal embedding models in 3B, 4B, and 8B sizes designed for visual document retrieval. The flagship 8B model achieves 63.42 NDCG@10 on the ViDoRe V3 benchmark, ranking first in its class for enterprise document retrieval tasks.
Vercel integrates Parallel web search across AI Gateway and AI SDK//Parallel's LLM-optimized web search tool is now available through Vercel's AI Gateway and AI SDK, enabling developers to add real-time web search to any model without provider-specific implementations. The integration supports both AI SDK and Vercel Marketplace with centralized billing and pricing at $5 per 1,000 requests.
Vercel build logs now support clickable links for faster debugging//URLs in Vercel build logs are now interactive, allowing developers to navigate directly to internal and external resources without manual copying and pasting. External links open in new tabs, streamlining the debugging workflow and reducing friction when investigating build issues.
featurereleaseplatform
AI2 publishes Nature paper on retrieval-augmented model for scientific literature synthesis with verifiable citations//AI2 and University of Washington researchers published a paper in Nature describing OpenScholar, an open-source retrieval-augmented language model designed to synthesize scientific literature with verifiable citations. The system searches a corpus of 45 million open-access papers and includes tools like ScholarQABench for evaluating citation quality.
Cloudflare releases AI Crawl Control reference documentation with GraphQL API and bot detection//Cloudflare has published comprehensive reference documentation for AI Crawl Control, including a GraphQL API reference, bot detection identifiers for major AI crawlers, and Worker templates for monetizing crawler access. Developers can now integrate crawler detection and access control directly into their applications.
featureapidocumentationplatform
Zed adds pluggable edit prediction with support for six providers including Ollama and Mercury Coder//Zed now supports multiple edit prediction providers through a new pluggable architecture, allowing developers to choose between Zeta, Mercury Coder, Sweep, Ollama, Codestral, and GitHub Copilot's Next Edit Suggestions. The refactored system simplifies adding new providers and enables local code prediction via Ollama with open-weight models.
Cloudflare Workflows gains visual diagram tool for debugging and code understanding//Cloudflare Workflows now automatically generates visual diagrams from your code, letting developers understand step execution, loops, and conditional logic at a glance. The feature is available in beta for all JavaScript and TypeScript Workflows in the Cloudflare dashboard.
GitHub Copilot in Visual Studio gains syntax-highlighted completions and partial acceptance//GitHub Copilot in Visual Studio received a January update featuring colorized code completions with syntax highlighting, the ability to partially accept suggestions by clicking, and productivity improvements like faster scrolling and enhanced Markdown preview controls. The update streamlines the coding experience with better visual feedback and granular control over code suggestions.
GitHub Copilot expands with Claude and Codex agents in public preview//Claude by Anthropic and OpenAI Codex are now available as coding agents for GitHub Copilot Pro+ and Enterprise subscribers. Users can create and assign agent sessions directly from GitHub.com, GitHub Mobile, and VS Code without additional subscriptions, with each session consuming one premium request.
featurereleaseintegrationplatform
Discord ships Rust audio backend for 80% of traffic, adds stream zoom and Linux timestamps//Discord's latest patch notes highlight significant performance improvements across desktop, mobile, and backend infrastructure. The update introduces user-facing features like mouse wheel zoom on screenshares, a new @time command for timezone-aware timestamps, and a "Bypass Slowmode" permission for server admins.
Magic Patterns launches unified Design Systems for consistent AI-generated UI//Magic Patterns has released Design Systems, consolidating its Presets and Libraries into a single source of truth for managing brand styles, typography, and components. The update includes new import capabilities from Figma, Storybook, and screenshots, allowing teams to bring existing design systems into the platform and automatically apply them across generated designs.
Cloudflare Queues launches on Workers Free plan with 10,000 daily operations//Cloudflare Queues, the company's message delivery service, is now available to all free-tier Workers accounts. Free users get 10,000 operations per day and access to up to 10,000 queues, with all existing features included except message retention is capped at 24 hours instead of 14 days.
featurepricingrelease
Hugging Face launches community-driven model evaluations with decentralized benchmarking//Hugging Face is introducing a decentralized evaluation system that allows the community to submit and verify model benchmark results directly on the Hub. The system links benchmark datasets, model repositories, and evaluation results through git-based pull requests, creating transparent and reproducible eval scores across the community.
featureplatformapiopen-source
Granola launches MCP integration, bringing meeting context to Claude, ChatGPT, and Cursor//Granola has released a Model Context Protocol (MCP) integration that allows AI tools like Claude, ChatGPT, and Cursor to access meeting notes directly. The integration eliminates context-switching by enabling developers and teams to reference meeting discussions, action items, and decisions without leaving their current workflow.
featureapiintegrationplatform
H Company releases Holo2-235B-A22B, reaching 78.5% accuracy on UI localization benchmarks//H Company has released the Holo2-235B-A22B Preview model, their largest UI localization model to date, achieving state-of-the-art performance on Screenspot-Pro and OSWorld G benchmarks. The model introduces "agentic localization," an iterative refinement approach that delivers 10-20% relative accuracy gains by allowing the model to progressively improve its predictions across multiple steps.
Cloudflare launches R2 Local Uploads in beta; up to 75% faster cross-region uploads//Cloudflare has released Local Uploads for R2 object storage, an open beta feature that writes upload data to storage near the client first, then replicates asynchronously to the bucket's primary location. The feature reduces upload latency by up to 75% for cross-region uploads while maintaining strong consistency and immediate data accessibility.
Vercel Workflow 4.1 Beta introduces event-sourced architecture with self-healing capabilities//Workflow 4.1 Beta fundamentally changes how the system tracks state by moving to an event-sourced architecture where state changes are stored as an immutable log of events rather than updating records in place. The update brings automatic recovery from queue failures, complete audit trails for debugging, and improved throughput supporting thousands of steps per second.
Vercel makes Turbo build machines default for new Pro projects; promises 30-70% faster builds//Turbo build machines—featuring 30vCPUs and 60GB of memory—are now automatically enabled for all new Pro projects and accounts upgrading from Hobby. The change delivers significant performance improvements across different build durations, with teams reporting up to 70% faster builds for longer compilation times.
Vercel adds zero-configuration support for Koa framework//Vercel now supports Koa applications with zero-configuration deployment, allowing developers to run expressive HTTP middleware frameworks on the platform. Koa apps automatically benefit from Fluid compute with Active CPU pricing, scaling dynamically based on traffic.
Vercel's AI Gateway integrates with ByteDance's TRAE, enabling unified model access and one-click deployments//Vercel AI Gateway is now available within ByteDance's TRAE coding agent, allowing developers to access hundreds of AI models through a single API key and deploy directly to Vercel production. The integration supports both SOLO and IDE modes with automatic failover, zero markup on tokens, and unified observability across AI usage and deployments.
Vercel Toolbar adds "Copy for Agents" to share visual context with AI coding assistants//Vercel Toolbar now includes functionality to copy complete visual context from deployment comments, providing coding agents with structured technical details including page URLs, viewport dimensions, component trees, and selector information. This eliminates manual context entry when asking AI agents to make changes to deployed applications.
Deno launches Sandbox for running untrusted code with network isolation and secret protection//Deno Sandbox is a new feature that lets developers safely execute untrusted or LLM-generated code in lightweight Linux microVMs with configurable network access and secret management. The service boots sandboxes in under a second and allows direct deployment to Deno Deploy without rebuilding.
releasefeatureplatformsecuritysdk
Deno Deploy reaches general availability with zero-config deployment and integrated databases//Deno Deploy is now generally available, offering streamlined JavaScript and TypeScript deployment across any framework without vendor-specific configuration. The platform includes automatic framework detection, continuous deployment from GitHub, integrated Postgres and Deno KV databases, and a new Deno Sandbox service for secure microVM execution.
Supabase launches official Claude connector with 32 database management tools//Supabase is now an official Claude connector, allowing developers to manage their entire database infrastructure through natural language conversations with Claude. The MCP server integration provides 32 tools covering SQL queries, schema modifications, edge function deployment, branching, and project operations.
GitHub Mobile now supports commenting on unchanged lines in pull request diffs//GitHub Mobile's Files Changed view has been expanded to allow developers to comment on any line within a changed file, not just those directly surrounding modifications. This update is available immediately on iOS and Android, bringing feature parity with the web experience.
Cloudflare R2 Local Uploads enters open beta; reduces upload latency by up to 75%//Cloudflare has launched R2 Local Uploads in open beta, allowing users to dramatically improve upload performance for globally distributed clients. Data is written to storage near the client and asynchronously replicated to the primary bucket, maintaining strong consistency while reducing Time to Last Byte by up to 75%.
Xcode 26.3 integrates Claude Agent SDK for autonomous coding tasks within Apple's IDE//Apple's Xcode 26.3 now natively supports the Claude Agent SDK, enabling developers to leverage agentic coding capabilities directly in the IDE. The integration allows Claude to autonomously handle complex, long-running tasks including visual verification with Previews, cross-project reasoning, and intelligent task execution without leaving Xcode.
Cloudflare Agents SDK v0.3.7 adds Workflows integration and synchronous state management//The latest Agents SDK release brings native integration with Cloudflare Workflows, enabling agents to handle real-time communication while workflows manage durable, long-running tasks. The update also introduces synchronous state updates with validation hooks, a new `scheduleEvery()` method for recurring tasks, and improvements to the callable system.
GitHub releases Dependabot Proxy as open source under MIT license//GitHub has open sourced the Dependabot Proxy, the HTTP authentication layer that handles connections between Dependabot and the GitHub API, private package registries, and Git servers. The proxy supports nine package managers including npm, Maven, Docker, and Cargo, enabling organizations to audit how their dependency updates are authenticated and contribute improvements to the codebase.
Dependabot adds OIDC authentication support for private registries//Dependabot now supports OpenID Connect (OIDC) authentication with private registries including AWS CodeArtifact, Azure DevOps Artifacts, and JFrog Artifactory. This eliminates the need for long-lived credentials stored as secrets, replacing them with short-lived dynamically generated tokens similar to GitHub Actions OIDC federation.
GitHub Copilot simplifies model selection with automatic discovery for Pro and Free users//GitHub has streamlined how users access Copilot models by making newly released models immediately available within Copilot features without requiring settings changes or policy management. The update applies to all user tiers—Pro+, Pro, and Free—while maintaining existing data-usage policies and user control over model selection.
featureplatformrelease
Deepgram adds Hebrew, Farsi, and Urdu support to Nova-3 speech-to-text model//Deepgram's Nova-3 speech-to-text model now supports three additional languages: Hebrew, Farsi (Persian), and Urdu. This expansion enables developers to build voice applications for these language communities with industry-leading accuracy.
Cloudflare Security Center adds threat actor alias lookup to simplify cross-vendor identification//Cloudflare Threat Events now include an "Also known as" field displaying common aliases and industry-standard names for tracked threat groups. This feature is available in both the dashboard and via API, helping security teams map actors to other vendors' naming conventions without manual cross-referencing.
Vercel adds Python 3.13 and 3.14 support; 3.14 becomes default in coming months//Vercel's Builds and Functions now support Python 3.13 and 3.14 alongside Python 3.12. The platform will switch to Python 3.14 as the default runtime in the coming months, though developers can pin to 3.12 in their project manifests to maintain compatibility.
Cloudflare Email Security improves monitoring dashboard with search, accessibility enhancements//Cloudflare has updated its Email Security monitoring page with improved visual layouts, enhanced accessibility features, and new search functionality across widgets. The update includes stacked bar charts for better data visualization, widget search for Policies and Submitters, and granular data breakdowns by month and threat type across all Email Security tiers.
Harvey scales legal knowledge coverage to 60+ jurisdictions with autonomous agent pipeline//Harvey has built "The Data Factory," an automated system using AI agents to discover, validate, and integrate legal data sources at scale. Since August 2025, the pipeline has expanded knowledge source coverage from 6 to 60+ jurisdictions and integrated over 400 legal data sources, enabling agents to handle complex queries across global legal databases without manual setup.
featureplatformapi
ElevenLabs ships Eleven v3 GA, WAV support, and Agents Platform enhancements//ElevenLabs released Eleven v3 out of alpha with improved stability, accuracy, and lower latency. The update includes WAV output format support for Text-to-Dialogue, expanded Agents Platform capabilities with branch renaming and guardrails, and multiple SDK updates across Python, JavaScript, React, and widget packages.
Cloudflare WAF adds detections for Django SQL injection and XWiki RCE vulnerabilities//Cloudflare has released new WAF rules to block two critical vulnerabilities: CVE-2025-64459 affecting Django ORM and CVE-2025-24893 enabling unauthenticated RCE in XWiki. Both detections are now set to block mode by default in the Cloudflare Managed Ruleset.
OpenAI launches Codex app for macOS with multi-agent orchestration and skill framework//OpenAI has released the Codex app for macOS, a dedicated interface for managing and running multiple coding agents in parallel on long-running tasks. The release includes expanded access to Codex through ChatGPT Free and Go plans, doubled rate limits across paid tiers, and a new skills framework that extends Codex beyond code generation to handle complex workflows and integrations.
OpenAI launches Codex app, a multi-agent development environment with native code review and automations//OpenAI has released the Codex app for macOS, a desktop application that serves as a command center for managing and supervising AI agents working on software development tasks. The app enables developers to run multiple agents in parallel, review diffs, leave inline feedback, and schedule automated workflows—all without juggling terminal windows.
Sim v0.5.79 adds enrichment tools, increases rate limits across plans//Sim Studio releases v0.5.79 with a new enrichment tool feature and increased rate limits and socket connections across all subscription tiers. The release also includes workflow optimization and bug fixes for MCP tool timeouts.
Sim launches v0.5.78 with Google Maps and DSPy integration, plus bug fixes//Sim releases v0.5.78 featuring new integrations with Google Maps and DSPy, enabling developers to build more sophisticated workflows. The release also includes several bug fixes addressing visibility controls, executor conditions, and billing detection.
Sim Studio releases v0.5.77 with SimilarWeb and Cal.com integrations//The latest update adds two new tool integrations (SimilarWeb and Cal.com) alongside improvements to block dragging and deployment version descriptions. The release also includes bug fixes for streaming, token limits, and various UI refinements.
Vercel Sandboxes now generally available for safely executing untrusted code//Vercel has launched Sandboxes as a general availability product—ephemeral compute environments for safely running untrusted code, AI-generated outputs, and third-party applications in isolated Firecracker microVMs. The offering protects production infrastructure by blocking sandbox code from accessing environment variables, databases, and cloud resources.
AssistLoop joins Vercel Agents Marketplace with native Next.js integration//AssistLoop, an AI-powered customer support solution, is now available directly in the Vercel Marketplace. Teams can deploy AI-driven support to Next.js applications in minutes with automatic environment variable injection and unified billing.
Sim Studio v0.5.76 fixes PostHog session recording and React hooks errors//Sim Studio releases v0.5.76 with improvements to PostHog analytics and bug fixes addressing session recording payload handling and OAuth component rendering. This incremental patch focuses on stability and infrastructure improvements.
Linear launches time-in-status tracking to identify workflow bottlenecks//Linear now tracks cumulative time spent in each issue status, enabling teams to spot bottlenecks and stuck work across their development process. The feature integrates across lists, boards, Insights, and Dashboards, while also adding bi-directional Jira Epic syncing capabilities.
Cloudflare launches Moltworker: serverless AI agent platform achieves 98.5% npm package compatibility//Cloudflare has introduced Moltworker, a middleware that enables running Moltbot (an open-source AI personal assistant) on Cloudflare Workers instead of dedicated hardware. The launch coincides with a milestone: Cloudflare Workers now supports 98.5% of the top 1,000 npm packages natively, thanks to expanded Node.js API compatibility.
Sim releases v0.5.75 with nested workflow execution snapshots and YouTube API enhancements//Sim v0.5.75 adds support for nested execution snapshots in child workflows and expands YouTube integration with new captions, trending, and video category tools. The release also introduces API timeout configuration, structured terminal output, and deployment version descriptions alongside multiple bug fixes and UX improvements.
OpenAI retires GPT-4o from ChatGPT, shifts users to GPT-5.2 with enhanced personalization//OpenAI is retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT on February 13, 2026, as usage has declined to just 0.1% of daily active users. The company has incorporated the feedback that drove GPT-4o's restoration into newer models, introducing customizable personality controls and improved creative capabilities in GPT-5.1 and 5.2.
PostHog's Logs feature exits beta with GA release, unified debugging platform pricing at $0.25/GiB//PostHog has moved its Logs product to general availability, completing its unified debugging platform that combines product analytics, session replay, error tracking, and now centralized log management. The service uses standard OpenTelemetry ingestion, integrates frontend and backend logs with automatic user/session linking, and includes AI-powered pattern summarization.
GitHub
Vercel
Stripe