AI Daily Dev — May 29, 2026

Anthropic Ships Claude Opus 4.8 — and Dynamic Workflows Up to 1,000 Subagents

SWE-Bench Pro jumps from 64.3% to 69.2%, ahead of GPT-5.5 and Gemini 3.1 Pro on Anthropic's runs; agentic computer use 82.8% → 83.4%, knowledge-work Elo 1753 → 1890.
Anthropic says Opus 4.8 is ~4× less likely than 4.7 to let its own code flaws slip past unremarked, and is more willing to flag uncertainty mid-task.
Same $5/$25 per million token pricing as 4.7; fast mode is ~2.5× quicker, plus a new claude.ai 'effort' control.
Claude Code v2.1.154 ships Dynamic Workflows: Claude writes a JS orchestration script, runs up to 16 concurrent and 1,000 total subagents per run.
HN front page within hours; an 'Is Opus 4.8 broken?' thread on file-reading reliability is already trending alongside it.

models anthropic.com

Same model that powers the Grok Build CLI: agentic coding focus, 256K context, text + image input, always-on reasoning.
Priced at $1/$2 per million tokens in/out and served at 100+ tokens/second — undercuts Opus 4.8 and GPT-5.5 by ~5×.
70.8% on SWE-Bench Verified on xAI's internal harness — respectable for v1, still 15–18 points behind Claude Opus and GPT-5.5.
Built-in MCP support, function calling, and structured outputs; available via xAI API and OpenRouter from today.

models x.ai

Cursor: 'more efficient than Opus 4.7' on CursorBench and 'more persistent on harder tasks,' rolled out same day.
Windsurf added Opus 4.8 at unchanged pricing plus a new Fast Mode priced at $25/M output tokens.
GitHub Copilot made Opus 4.8 generally available across Pro, Pro+, Business and Enterprise on launch day.
Day-one model parity across every major coding IDE is now the norm — and the differentiator is harness quality, not model access.

tools x.com

Starting this fall in the US and Brazil, ChatGPT will surface live Associated Press results on election night.
OpenAI is offering Codex Security and its Trusted Access for Cyber program free to registered US voting-system manufacturers.
SynthID watermarks now embedded in images generated by ChatGPT, Codex and the OpenAI API — survives screenshots and resaves.
Tools restricted from political impersonation, voter suppression and deceptive campaign use; Democracy Works partnership covers registration info.

industry openai.com

Bloomberg report ahead of WWDC June 8: Siri overhaul is the centerpiece of iOS 27, iPadOS 27 and macOS 27.
New dedicated Siri app, plus a system-wide 'Search or Ask' bar in the Dynamic Island that lets users swap in ChatGPT or Gemini.
Web-grounded answers with bullet points and large images — Apple's first serious shot at ChatGPT/Claude/Gemini parity.
Ships alongside a revamped Image Playground, systemwide AI grammar checker and AI-generated wallpapers.

industry bloomberg.com

Sixth European office in a year — joining London, Dublin, Paris, Zurich and Munich.
Named Italian customers include Generali, Unipol, Pirelli, Enel, Angelini Pharma, Bracco, Bending Spoons and Satispay.
Office spans sales, technical pre/post-sales and policy; Anthropic plans to triple its international workforce.
Lands four days after Anthropic co-founder Chris Olah presented at the Vatican alongside Pope Leo's Magnifica Humanitas encyclical.

industry anthropic.com