News
Industry updates, tool releases, and notable developments in the AI coding tools space.
Industry updates, tool releases, and notable developments in the AI coding tools space.
Claude Code 2.1.152 shipped today, May 27. /simplify is back — as a wrapper for /code-review --fix, the new flag that applies review findings to the working tree. Six days after the verb was removed, it returns aliased to a different command.
OpenAI shipped Codex CLI 0.134.0 on May 26. Two headline items: search across local conversation history with previews, and read-only MCP tools that run concurrently when they advertise readOnlyHint. Plus profile-config migration enforcement and richer hook context.
GitHub shipped targeted model rules on May 26. Enterprise owners can allow specific Copilot models for specific organizations, with each model set Enabled (automatically on by default) or Optional. Public preview, Business and Enterprise plans.
Anthropic published the first results from Project Glasswing on May 22. Around fifty partners running Claude Mythos Preview found more than ten thousand high- or critical-severity vulnerabilities in the first month, including 2,000 at Cloudflare and 271 fixed in Firefox 150.
Claude Code 2.1.149 shipped on May 22 with a per-category breakdown of /usage. For the first time, the tool reports which sub-component of a session — skills, subagents, plugins, or per-MCP-server — is driving consumption against your limit.
Claude Code 2.1.147 shipped on May 21 and removed /simplify outright, replacing it with /code-review — a command that reports correctness bugs and posts inline GitHub PR comments instead of applying fixes. The old cleanup-and-fix behavior is gone.
From May 17, GPT-5.3-Codex is the base model for every Copilot Business and Enterprise organization, replacing GPT-4.1. It is also Copilot's first long-term-support model, guaranteed available for a year.
Google released Gemini 3.5 Flash and Antigravity 2.0 at I/O on May 19. The model is faster and lands across Google's products and GitHub Copilot — but at $1.50 per million input tokens, the Flash tier is now three times the price of its predecessor.
Researchers disclosed a cluster of vulnerabilities in AI coding agents in May 2026: prompt-injection-to-RCE in CrewAI, an unauthenticated Azure SRE Agent endpoint, a poisoned AGENTS.md chain against Codex, and over a thousand malicious skills.
OpenAI put its Codex coding agent inside the ChatGPT mobile app on May 14. The phone pairs to a Mac running Codex and can review diffs, approve commands, and redirect a running task — supervision without the desk.
xAI opened the Grok Build beta on May 14 for SuperGrok Heavy subscribers. The terminal agent runs up to eight workers in parallel, carries a 2M-token context window, and ships with an approve-first plan mode.
Anthropic's May 11 update adds a unified session list to Claude Code, letting you track, background, and switch between multiple concurrent coding sessions from the terminal.
Copilot Workspace, GitHub's task-oriented AI environment, is now generally available. Here's what's in the box, what isn't, and how it stacks up against Cursor's agent features.
Copilot Business and Enterprise users can now choose between Claude 3.5 Sonnet, Gemini 1.5 Pro, and GPT-4o within the same subscription. Individual plan access comes later this quarter.
Cline now supports the Model Context Protocol natively. Configure an MCP server, and Cline can call it as a tool — connecting the agent to your databases, APIs, and internal services.
The new Sonnet release lands with modest capability improvements and a meaningful pricing change for cached prompts. Here's what it means for your AI coding tool of choice.
Cursor's latest update adds first-class support for team-shared rules and prompt templates across a workspace. The feature targets the team-fragmentation problem that's grown with AI tooling adoption.
Claude 3.7 Haiku launches with significant speed improvements over 3.5 Haiku and pricing on par with the older model. The implications for autocomplete-style workflows are real.
JetBrains updated their AI assistant with Claude 3.5 Sonnet support and a free tier for individual developers. The IntelliJ ecosystem catches up to Cursor on AI parity.
Anthropic ships team features for Claude Code, including shared CLAUDE.md sync and team-level usage analytics. The CLI category goes upmarket.
Anthropic doubled the default prompt cache TTL from 5 minutes to 1 hour. For long coding sessions, this is a meaningful cost reduction.
Batch API offers half-price tokens for jobs completing within 24 hours. Useful for non-interactive AI coding workflows.
Claude now supports configurable thinking time per request. Useful for tools that want to balance reasoning quality with latency.
Claude 3.7 Opus launches as the new flagship for hard tasks. Strong on coding benchmarks; pricier than Sonnet.