{"slug": "ai-agents-weekly-claude-opus-4-8-claude-code-dynamic-workflows-chrome-devtools-1", "title": "🤖 AI Agents Weekly: Claude Opus 4.8, Claude Code Dynamic Workflows, Chrome DevTools for Agents 1.0, DeepSWE, Agent Harness Scaling Laws, and More", "summary": "Anthropic released Claude Opus 4.8, an incremental upgrade with improved agentic judgment and honesty, achieving 84% on Online-Mind2Web and roughly 4x fewer code flaw overlooks. Harvard's Zitnik Lab introduced AutoScientists, a decentralized multi-agent system that self-organizes around promising research directions, reaching a 74.4% mean leaderboard percentile on biomedical ML. Other developments include Chrome DevTools for Agents 1.0, DeepSWE raising the coding-agent bar, and scaling laws for agent harnesses.", "body_md": "# 🤖 AI Agents Weekly: Claude Opus 4.8, Claude Code Dynamic Workflows, Chrome DevTools for Agents 1.0, DeepSWE, Agent Harness Scaling Laws, and More\n\n### Claude Opus 4.8, Claude Code Dynamic Workflows, Chrome DevTools for Agents 1.0, DeepSWE, Agent Harness Scaling Laws, and More\n\nIn today’s issue:\n\nAutoScientists self-organize agent teams\n\nAnthropic ships Claude Opus 4.8\n\nClaude Code adds dynamic workflows\n\nChrome DevTools for agents hits 1.0\n\nDeepSWE raises the coding-agent bar\n\nxAI opens grok-build-0.1 in beta\n\nMicrosoft open-sources Webwright for agents\n\nScaling laws for agent harnesses land\n\nHarness sensitivity proves non-monotone\n\nSIA co-updates harness and weights\n\nCUA-Gym scales computer-use RL data\n\nPolar trains agents on real harnesses\n\nAnthropic details how it contains Claude\n\nXiaomi slashes MiMo-V2.5 API prices\n\nLanguage models learn to sleep\n\na16z maps the AI application layer\n\nAnd all the top AI dev news, papers, and tools.\n\n**Top Stories**\n\n**AutoScientists Self-Organize for Long-Running Science**\n\nHarvard’s Zitnik Lab introduced AutoScientists, a decentralized multi-agent system for long-running computational science where agents self-organize around promising research directions instead of following a fixed plan.\n\n**Self-organizing teams:** Agents form around promising directions and vet proposals before allocating resources, so compute goes only to ideas that survive review.**Learning from failure:** The system documents failures as well as successes, building a record that steers future exploration.**Validated broadly:** Reaches a 74.4% mean leaderboard percentile on biomedical ML, 1.9x faster convergence on language model training, and gains on protein fitness.\n\n**Claude Opus 4.8 Sharpens Agentic Judgment**\n\nAnthropic released Claude Opus 4.8, an incremental upgrade over Opus 4.7 tuned for sharper judgment, more honesty about its own progress, and longer independent runs.\n\n**Agentic gains:** Posts 84% on Online-Mind2Web for computer-use and browser-agent tasks, and the team reports it is roughly 4x less likely than its predecessor to overlook code flaws.**Self-correction and honesty:** Early testers cite improved reliability, better self-correction, and more accurate reporting of how far it has actually gotten on a task.**New controls:** Ships alongside dynamic workflows, an effort control to dial response intensity, and a Systems API update that lets you change mid-task instructions without breaking the prompt cache.**Why it matters:** The honesty and judgment gains target the exact failure modes that break long-horizon agents, where a model that overstates progress derails an entire run.\n\nAvailable today via the `claude-opus-4-8`\n\nAPI identifier at the same price as before ($5/$25 per million tokens), with a 3x cheaper Fast mode.", "url": "https://wpnews.pro/news/ai-agents-weekly-claude-opus-4-8-claude-code-dynamic-workflows-chrome-devtools-1", "canonical_source": "https://nlp.elvissaravia.com/p/ai-agents-weekly-claude-opus-48-claude", "published_at": "2026-05-30 15:02:03+00:00", "updated_at": "2026-06-15 21:56:38.721626+00:00", "lang": "en", "topics": ["ai-agents", "large-language-models", "ai-research", "ai-products", "ai-tools"], "entities": ["Anthropic", "Claude Opus 4.8", "Harvard", "Zitnik Lab", "AutoScientists", "Chrome DevTools for Agents", "DeepSWE", "xAI"], "alternates": {"html": "https://wpnews.pro/news/ai-agents-weekly-claude-opus-4-8-claude-code-dynamic-workflows-chrome-devtools-1", "markdown": "https://wpnews.pro/news/ai-agents-weekly-claude-opus-4-8-claude-code-dynamic-workflows-chrome-devtools-1.md", "text": "https://wpnews.pro/news/ai-agents-weekly-claude-opus-4-8-claude-code-dynamic-workflows-chrome-devtools-1.txt", "jsonld": "https://wpnews.pro/news/ai-agents-weekly-claude-opus-4-8-claude-code-dynamic-workflows-chrome-devtools-1.jsonld"}}