Fetching from the wire…
Top 5 · 2026-04-03 · source-backed
Cursor 3 launched on April 2. Not an update. A full rebuild. The editor is now secondary to what they're calling an agent orchestration workspace. You can run unlimited parallel agents, locally or in the cloud, launch them from your phone or Slack or Linear, and manage them in a fleet view. There's a Design Mode for annotating UI elements in the browser. There's /worktree for isolated agent execution in git worktrees. There's /best-of-n for running the same prompt against multiple models side by side.
The same day, GitHub shipped custom .agent.md files for Copilot in Visual Studio. You define an agent in markdown, drop it in your repo, and it gets full workspace awareness, tool access, and MCP connections. A new find_symbol tool gives agents language-aware navigation across C++, C#, TypeScript, and anything with an LSP extension. They also shipped the Copilot SDK in public preview across five languages, exposing the same agent runtime that powers their cloud agent.
And Claude Code pushed two releases in a single day. v2.1.90 added /powerup (an in-terminal learning system), auto mode boundary enforcement, and .husky directory protection. v2.1.91 followed hours later with TaskCreated hooks, worktree HTTP hooks, deep links via claude-cli:// protocol, and YAML glob rules for skills.
I've been using Claude Code daily for months. The shift I'm seeing isn't incremental. The file tree is becoming an implementation detail. The agent control plane is becoming the primary interface. Cursor explicitly framed their 3.0 as a response to Claude Code's reported 54% market share, positioning themselves as a fleet manager rather than a file editor. GitHub is making every repo an agent workspace. Claude Code is building the hooks and lifecycle events that let you wire agents into CI/CD pipelines.
If you're still thinking about AI coding tools as "autocomplete with extra steps," you're working with a mental model from 2024. The pattern is clear: design your workflows around agent coordination. Task queues, verification loops, merge strategies. The editor is where agents happen to write code. It's not where you spend your time anymore.
Each link below shares sources, entities, or timing with this story.
Anysphere shipped Cursor 3 (codenamed "Glass") on April 2, and it's not really an IDE anymore. It's an agent workspace. The redesign is fundamental. Developers enter natural language task descriptions, select an LLM, and Cursor generates code plus a demo video. Local and cloud...
Anthropic invented a file convention. It's now shipping GA inside a competitor's product. Nobody wrote a spec, nobody held a standards meeting, it just happened. On July 29, GitHub made agent skills and MCP server support generally available in Copilot code review for all Pro,...
A spec is a press release until someone who didn't write it implements it. GitHub made Agent Plugins 1.0 generally available on August 12 across VS Code, Copilot CLI, the Copilot SDK, and the Copilot app on all plans. The spec, published August 6, was co-authored by AWS, Anysp...
Two things in one release, and only one of them is in the security section. CLAUDE_CODE_SUBAGENT_MODEL used to override the model for every subagent. In 2.1.251 it sets the default instead. An agent definition's model: frontmatter now wins, and so does an explicit per-spawn mo...
GitHub announced that every Copilot plan moves to usage-based billing via "AI Credits" on June 1, 2026. The flat-rate model is dead. Every plan gets a monthly credit allotment calculated by actual token consumption (input, output, cached) at listed API rates. Code completions...
"owner/*" entries in marketplace settings let you allow or block every repo under a GitHub org in one rule instead of enumerating them. Same release warns when a workflow agent requests a policy-restricted subagent model, and adds a /teleport hint in cloud sessions for continu...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.