On August 5, 2026, Meta released Muse Code, a multi-agent coding harness with event-log replay, in beta, together with an updated Muse Spark 1.2 coding model. Muse Spark 1.2 offer…
Vercel added Meta's Muse Spark 1.2 to AI Gateway with bring-your-own-key support and no markup. The 1.2 update focuses on code generation, complex debugging, codebase understandin…
On August 5, 2026 Visual Studio Code 1.132 shipped Agent Host, an architecture that runs AI agents in a separate process, fully isolated from the extension host, so a hung Copilot…
Cloudflare now captures OpenTelemetry traces automatically when running a Worker locally through Wrangler or the Vite plugin, and exposes a SQL-queryable Local Explorer API. AI co…
Cloudflare added structured OpenTelemetry tracing to local development in wrangler dev and the Cloudflare Vite plugin. Running a Worker locally now emits production-equivalent tra…
Cloudflare Artifacts repositories can now run a CI/CD pipeline automatically on every push. Developers define a CI Workflow with the new CI SDK that builds, lints, type-checks, an…
Cloudflare introduced @cloudflare/ci, a CI/CD system built on Cloudflare Workflows that replaces YAML pipelines with TypeScript. Because pipelines are Workflows, they can self-hea…
On August 3, 2026, Alibaba released Qwen3.8-Max, a mixture-of-experts model with 2.4 trillion total parameters (about 95B active per query) and a 1M-token context window, natively…
GitHub shipped Copilot CLI v1.0.78, adding a live tool-call timeline that shows how long each tool call takes (ticking live for calls of five seconds or more, on by default), an e…
GitHub added a per-task reasoning-level control for the Copilot cloud agent. Users pick a reasoning level alongside the model when starting a task, and the agent uses it for that…
Cloudflare added new Workers runtime tracing APIs, startActiveSpan() and span.end(), that let developers create custom spans beyond a single callback to instrument long-running op…
Cloudflare added Worker startup profiling to Wrangler: wrangler check startup now reports raw and compressed bundle sizes plus a CPU activity summary and generates a .cpuprofile f…
On July 30, 2026, GitHub's July Visual Studio Copilot update introduced a new Agent (Preview) in the agent picker, built on the same GitHub Copilot SDK that powers the Copilot CLI…
On July 30, 2026, GitHub summarized Copilot's July VS Code releases (v1.127-v1.131), centered on the Agents window: reviewing code beside chat with diff counts, Git worktree suppo…
GitHub launched stacked pull requests in public preview on July 30, 2026, letting developers break large changes into an ordered series of small, independently reviewable pull req…
GitHub retired GitHub Models as of July 30, 2026, shutting down its playground, model catalog, inference API, and bring-your-own-key (BYOK) access for all customers. GitHub direct…
On July 30, 2026, OpenAI lowered GPT-5.6 API prices, cutting Luna by 80% (to $0.20/$1.20 per million input/output tokens) and Terra by 20% (to $2/$12), while flagship Sol held at…
On July 29, 2026, Cursor released a native iPad app for all paid plans, with a rebuilt large-screen layout featuring pinned sidebar chats, split-screen review, and full file diffs…
Cloudflare now requires the Workers Paid plan (from $5/month) to run three resource-intensive Workers AI models: Kimi K2.6, Kimi K2.7-Code, and GLM-5.2. Cloudflare frames the chan…
On July 28, 2026, Cloudflare began requiring the Workers Paid plan to run three resource-intensive models on Workers AI: @cf/moonshotai/kimi-k2.6, @cf/moonshotai/kimi-k2.7-code, a…
On July 28, 2026, Cursor introduced Cursor Start, a monthly plan priced at 649 rupees for developers in India, bundling access to Grok 4.5, the Composer model, always-on cloud age…
On July 27, 2026, Cloudflare added createTestHarness() to Wrangler, an API for running integration tests across Workers. It supports multi-Worker routing, request mocking, and Pla…
On July 24, 2026, Anthropic released Claude Opus 5, its new flagship model, positioning it as near Fable 5's frontier intelligence at half the price. Opus 5 pairs a 1M-token conte…
On July 24, 2026, Anthropic released Claude Opus 5, a frontier model priced at $5/$25 per million input/output tokens (unchanged from Opus 4.8 and roughly half the cost of Fable 5…
On July 24, 2026, Replit updated its Agent: after connecting Slack, Agent can search messages, files, channels, and people, read conversations, send messages, and manage Slack can…
On July 23, 2026, Vercel added Ant Group's Ling 3.0 Flash to AI Gateway, a Mixture-of-Experts model with 124B total parameters (about 5.1B active per token), a 256K context window…
In late July 2026, GitHub introduced custom agents for the Copilot CLI: developers define an agent in a Markdown file stored in a repository's .github/agents directory (or ~/.copi…
Released July 22, 2026, Claude Code v2.1.218 fixes a bug where message-normalization cost grew quadratically with the number of turns, causing multi-second stalls and slow resumes…
Cursor Router, the request-level classifier behind Auto mode, added Auto Intelligence and Auto Balance presets that now learn continuously from production traffic and route to new…
On July 22, 2026, Cursor made its Auto mode powered by Cursor Router, an intelligent router that classifies each request by query, context, task complexity, and domain and sends i…
On July 21, 2026, Google released Gemini 3.6 Flash at $1.50/$7.50 per 1M input/output tokens with a 1M-token context window, available day one in AI Studio, the Gemini API, and th…
On July 21, 2026, Vercel made Poolside's Laguna S 2.1 available on AI Gateway in free and paid tiers. Laguna S 2.1 is an open-weight Mixture-of-Experts model built for agentic cod…
DeepSeek is transitioning V4 from preview to official status and introducing time-based API pricing: calls during Beijing-time peak windows cost roughly double the off-peak rate,…
DeepSeek moved its V4 family out of preview into general availability on July 20, 2026, shipping V4-Pro (1.6T total / 49B active parameters) and V4-Flash (284B / 13B active), both…
Moonshot AI has temporarily suspended new subscriptions to Kimi K3, its 2.8-trillion-parameter, 1M-token frontier model, after about 48 hours of demand pushed serving infrastructu…
On July 17, 2026 Cursor improved its Slack integration so its agent posts a plan before it starts work, lets users redirect early, operates across multiple repository environments…
Claude Code version 2.1.212, released mid-July 2026, lets developers fork a conversation into a background session, adds session-wide limits of 200 subagent spawns and 200 WebSear…
On July 16, 2026 Databricks released the Lakebridge Agentic Converter in beta, an agent powered by Genie Code that converts legacy SQL dialects, including T-SQL, Snowflake, Redshi…
On July 27, 2026, Moonshot AI released the full open weights for Kimi K3 (about 1.4TB at MXFP4) under a Modified MIT license, following the model's July 16 API launch. K3 is a 2.8…
On July 15, 2026, Anaconda acquired Kilo Code, an open-source, model-agnostic agentic engineering platform used by more than 3 million developers across VS Code, JetBrains, web, a…
On July 15, 2026, OpenAI entered hardware with Codex Micro, a $230 mechanical keypad co-designed with Work Louder that gives Codex power users lighted Agent Keys showing agent sta…
On July 15, 2026, Thinking Machines Lab released Inkling, its first open-weights model, under Apache 2.0. Inkling is a multimodal mixture-of-experts model (975B total / ~41B activ…
On July 15, 2026, Vercel's Chat SDK added opt-in support for Discord Components V2, letting developers arrange text, images, files, and buttons as flexible components in any order…
On July 15, 2026, Vercel changed how the CLI links projects: it now resolves your team first before searching for projects, so linking is faster and more predictable across vercel…
GitHub expanded bring-your-own-key (BYOK) capabilities for GitHub Copilot in JetBrains IDEs, giving JetBrains developers broader control over which model providers and keys back t…
Vercel Blob now supports read-after-write consistency on private storage through a useCache: false option on get() and presignUrl(), returning the latest write instead of a possib…
Vercel is releasing the data behind its AI Gateway model leaderboards under a CC BY 4.0 license, accessible via a downloadable export and a leaderboard-export API endpoint, and le…
Vercel released a plugin that gives GitHub Copilot in VS Code and the Copilot CLI on-demand access to Vercel platform knowledge, including skills for Next.js, the AI SDK, and Verc…
On July 13, 2026, Sourcegraph added inner-loop agent rendering to Batch Changes, streaming the thinking and execution progress of coding agents like Claude Code and Codex directly…
On July 9, 2026 GitHub added OpenAI's newly released GPT-5.6 models — Sol, Terra, and Luna — to GitHub Copilot, letting developers select the flagship, balanced, or low-cost tier…
On July 9, 2026, GitHub shipped a Copilot feature that offers to generate a high-level overview when a user visits the home page of a repository they haven't contributed to. Copil…
OpenAI released GPT-5.6 in three tiers on July 9, 2026: Sol at $5/$30 per million tokens (91.9% on Terminal-Bench 2.1), Terra at $2.50/$15 for everyday use, and Luna at $1/$6 as t…
On July 9, 2026 Vercel added the ability to deploy apps built in Lovable directly to Vercel, giving projects generated in the AI app builder a native path to Vercel's hosting, pre…
Cloudflare's Workers AI toMarkdown and AI Search now convert GIF and BMP images into searchable Markdown descriptions, extending the existing JPEG, PNG, WebP, and SVG support with…
On July 8, 2026 Cloudflare launched Drop, which turns a dragged folder or zip of static assets into a live URL on Cloudflare's network in seconds — no Cloudflare account, Wrangler…
Cloudflare shipped Drop, letting anyone drag a folder or zip into the browser and get a live site on Cloudflare's global network in seconds — with no account, no Wrangler config,…
GitHub's July 8, 2026 Copilot changelog for VS Code (covering versions 1.123–1.127) makes agentic browser tools generally available — letting agents navigate pages, inspect conten…
SpaceXAI released Grok 4.5, its first model since going public and acquiring Cursor, pitched as a coding and agentic-work model rather than a consumer chatbot. Priced at $2 per mi…
SpaceXAI (formerly xAI) launched Grok 4.5, built on a ~1.5-trillion-parameter V9 foundation and supplemented with Cursor coding data, positioning it for coding, agentic tasks, and…