BetaWeekly digests are a beta — we're trying something new. Feedback welcome.
Frontier AI Labs weekly digests
- Week of July 13, 2026Permission hardening and mode control across the agent toolchain
Anthropic's Claude Code shipped three major security-focused releases this week, closing a permission bypass in PowerShell and fixing worktree isolation for subagents, while OpenAI's ChatGPT desktop app unified its Chat, Work, and Codex modes. The Claude Platform also introduced mid-conversation system messages for three flagship models.
- Week of June 29, 2026Sonnet 5 and Claude Code reshape the frontier
Anthropic launched Claude Sonnet 5 with near-Opus performance at a fraction of the cost, while Claude Code — the new autonomous engineering assistant — delivered a landmark government security audit and shipped major stability improvements.
- Week of June 22, 2026Claude Tag brings collaborative AI agents to Slack as GPT-5.6 previews
Anthropic launched Claude Tag, a team-oriented AI agent for Slack, while OpenAI began previewing GPT-5.6 Sol and Codex Remote reached general availability across all plans.
- Week of June 15, 2026Health intelligence, agent safety, and enterprise controls
GPT-5.5 Instant brought frontier health evaluations to all free ChatGPT users with a 71% reduction in factuality issues, while Claude Code hardened agent safety with destructive command blocking and Anthropic launched enterprise-managed MCP auth. OpenAI also shipped credit usage analytics for Enterprise admins and granular spend controls.
- Week of June 8, 2026Fable 5 debuts, then vanishes — and tool transparency deepens
Anthropic launched Claude Fable 5 and Mythos 5 with 1M token context, then had access suspended by a US government export directive by week's end. Across both Anthropic and OpenAI, tooling gained sharper controls: Claude Code sub-agents can now spawn sub-agents, web search tools expose response-inclusion parameters, and ChatGPT added rate-limit reset banking.
- Week of June 1, 2026Codex expands beyond coding while Claude and OpenAI harden security and cost controls
OpenAI rolled out role-specific Codex plugins, interactive Sites, and enterprise admin controls for ChatGPT, while Anthropic stopped billing for Claude refusals without output and added a max_tokens parameter to the advisor tool. Both platforms shipped significant security hardening across Claude Code, ChatGPT Lockdown Mode, and the Claude API.
- Week of May 25, 2026Anthropic launches Opus 4.8 as Claude Code gains dynamic workflows
Anthropic launched Claude Opus 4.8 with a 1M token context window, while Claude Code introduced dynamic workflows for multi-agent orchestration. OpenAI shipped GPT-5.5 Instant improvements, Codex computer use on Windows, and new ChatGPT app templates for GitHub Enterprise.
- Week of May 18, 2026Codex ships goal mode, appshots, and remote computer use
Codex exited goal mode from experimental and launched appshots for macOS, while Claude Code fixed a series of sandbox permission escapes and added a Workflow tool for multi-agent orchestration. Anthropic also opened MCP tunnels as a research preview.
- Week of May 11, 2026Codex goes mobile as AI coding agents get remote control
OpenAI brought Codex to the ChatGPT mobile app, letting developers monitor and approve coding tasks from their phone, while Anthropic shipped agent view and goal-completion mode in Claude Code and expanded Opus 4.7 fast mode across the platform.