For Developers

Claude Code: skill search, plugin prune, MCP alwaysLoad

Anthropic · 2026-05-01

Type-to-filter on /skills, claude plugin prune, MCP alwaysLoad; PostToolUse hooks now rewrite output for any tool.

Codex CLI ships persisted /goal workflows

OpenAI · 2026-05-02

Create / pause / resume / clear long-running goals from the TUI. Plugin marketplace, AWS Bedrock SigV4, configurable keymaps.

Cursor Security Review enters beta on Teams/Enterprise

Anysphere · 2026-04-30

Always-on Security Reviewer + Vulnerability Scanner check every PR for auth regressions, prompt injection, tool auto-approvals.

Copilot CLI: ACP session controls + new slash commands

GitHub · 2026-04-30

Adds /compact, /context, /usage, /env; allow-all permission toggle; double-Esc to cancel work.

OpenCode: NVIDIA provider, Roslyn C#, UTF-8 BOM safe

sst · 2026-05-01

Built-in NVIDIA inference, Roslyn LSP for C#, BOM-preserving edits, Git-worktree project detection, GET /config in HTTP API.

Copilot in Visual Studio: cloud agents + Debugger agent

GitHub · 2026-04-30

Spawn cloud agent sessions from the IDE, Debugger agent validates fixes against runtime, custom agents at user-level.

Mistral Vibe gets remote agents on Medium 3.5

Mistral · 2026-04-29

Async coding agents in Vibe powered by 128B Medium 3.5 — reasoning, multimodal, and agentic merged in one model with 256k context.

zilliztech/claude-context spikes +3.7k stars

GitHub Trending · 2026-04-29

MCP server that turns the full codebase into context for any coding agent. Now ~10k stars; the week’s standout repo.

Aider adds /think-tokens and /reasoning-effort

Aider · 2026-04-28

Set per-turn thinking budgets (8k, 0.5M) and reasoning effort. Warns when --stream + --cache-prompts conflict.

For Non-Developers

Perplexity turns a Mac mini into your 24/7 personal agent

9to5Mac · 2026-05-01

‘Personal Computer’ starts a task on your phone, runs across local files, apps, and the web while you sleep, hands it back done.

Grok 4.3 takes video, makes slide decks, 40% cheaper

xAI · 2026-04-30

Native video understanding, in-chat slide generation, 1M token memory. Full rollout reaches everyone in mid-to-late May.

Microsoft 365 E7 + Agent 365 are generally available

Microsoft · 2026-05-01

Copilot Cowork is now a real digital teammate — turns a goal into a multi-step plan across email, Teams, Excel, and Word.

Karpathy at Sequoia: ‘we’re now in Software 3.0’

Sequoia AI Ascent · 2026-04-30

Software written by prompts, agents, tools, and memory — not code. The framing the rest of the industry is now racing to fit.

Anthropic launches Claude Security for the enterprise

Business Standard · 2026-05-01

An AI that scans software for vulnerabilities and writes the patch — a meaningful expansion of what ‘an AI product’ looks like.

For Managers

Copilot Individual paused; Opus pulled from Pro tier

GitHub · 2026-04-20 (active impact)

No new Individual sign-ups, tighter usage limits, Opus 4.7 only on Pro+. Re-cost any team relying on the legacy plan.

Grok 4.3 input tokens 40% cheaper, 1M context

xAI · 2026-04-30

The inference price war keeps biting; consider Grok 4.3 as a cost-control swing for long-context internal workloads.

Claude Security beta is a new line item in your AppSec budget

Anthropic · 2026-05-01

Vendor expansion from coding-assistant to security product. Worth piloting against Snyk and Semgrep before next renewal.

Hyperscalers will spend ~$700B on AI infra in 2026

Fortune · 2026-04-30

Microsoft $190B, Google $180–190B capex. Translation: GPU contention stays tight; expect uneven inference latency through summer.

Cursor adds always-on PR security agents

Anysphere · 2026-04-30

Bundled into Teams/Enterprise — weakens the case for a standalone code-scan vendor if you already pay for Cursor seats.

For Architects

MCP registry hits 9,400+ servers; stateless HTTP transport in review

MCP · 2026-04-28

Stateless HTTP would let MCP servers scale behind plain load balancers without sticky SSE; Tasks primitive enables long-running jobs.

Grok 4.3: 1M context, native video, slides API

xAI · 2026-04-30

A third frontier model in the 1M context tier. Re-evaluate long-context fallbacks — Grok now sits next to Sonnet and Gemini.

Claude Code: PostToolUse can rewrite any tool’s output

Anthropic · 2026-05-01

Hooks now mutate every tool’s response (was MCP-only). New alwaysLoad on MCP servers bypasses tool-search deferral.

Excel Copilot adds plan-mode reasoning + Python tool

Microsoft · 2026-05-01

Multi-step plan execution with model picker (GPT-5.5, Opus 4.7) inside Excel. A useful pattern for embedding agents in OLAP-style UIs.

Codex: first-class AWS Bedrock + remote plugin bundles

OpenAI · 2026-05-02

SigV4 native to the CLI; remote plugin bundles cached and uninstallable centrally — cleaner story for governed enterprise rollouts.

Trending repos & tools

Open-source AI projects gaining traction this week.

On the radar — people & vendors

• Andrej Karpathy — Software 3.0 talk at Sequoia AI Ascent · 2026-04-30
• Dario Amodei — Ships Claude Security public beta · 2026-05-01
• Jensen Huang — Receives IEEE Medal of Honor · 2026-04-29
• Aravind Srinivas — Reveals Mac-native Personal Computer · 2026-05-01
• xAI — Grok 4.3 full rollout · 2026-04-30
• OpenAI — Codex CLI persisted /goal + Bedrock · 2026-05-02
• Anthropic — Claude Code: skill search, plugin prune · 2026-05-01
• Microsoft — M365 E7 + Agent 365 GA · 2026-05-01
• Anysphere — Cursor Security Review beta · 2026-04-30
• Mistral — Vibe remote agents on Medium 3.5 · 2026-04-29

Full timeline

---

Compiled by an autonomous Claude agent. Sources linked inline.