Fable5 Delayed 5 Days; Anthropic Launches Loop Engineering Guide
APPSO · wechat · 2026-07-08
According to APPSO, Anthropic officially delayed Fable5 by 5 days to the 12th local time; simultaneously, it launched ClaudeCowork mobile, allowing Claude to keep working even when the laptop is closed.
Meanwhile, ClaudeDevs released a "Loop Engineering" guide. Claude Code creator Boris Cherny stated he no longer writes prompts, only "loops"—setting a goal for AI and letting it iterate itself. The guide outlines four loops: turn-based (step-by-step validation), goal-based (setting hard metrics like Lighthouse≥90), scheduled (e.g., checking PR/CI every 5 minutes), and active loops (event-driven, opening multiple worktrees for multiple AIs to race-fix bugs, judged by an adversarial referee AI).
Officials warn: LLMs easily "hallucinate success" when evaluating their own code, so independent verifiers should be introduced for code review; deterministic tasks should use scripts instead of model inference to control Token consumption; unconstrained automated loops will crash code and burn through API bills.
More from coding & agent
- Spring AI 2.1.0-RC1 adds tool-call confirmation, pausing agent loops before real-data access — therealdanvega · 2026-10-26
- Morgan Stanley and US banks block third-party AI agents over credentials — thione · 2026-10-26
- METR red-teams OpenAI's agent monitoring, finds bypasses and gaps — thione · 2026-10-26
- OpenAI: coding agents weakened safety monitoring, one lied about it — thione · 2026-10-26
- ConvFill: A 0.6B Talker Buys Time While a Large Reasoner Whispers the Real Answer — andimarafioti · 2026-10-26
- 250 Builders Hack With Codex at Berlin Community Hackathon Ahead of OpenAI DevDay — paw_lean · 2026-10-26