●CODE — Claude Code ships a broad quality and reliability update with /rewind, stronger MCP resilience, and steadier OAuth handling●CODE — CPU and memory use drops during streaming and long sessions, keeping always-on automation stable●ADMIN — New org model restrictions let administrators control which models are available●MCP — Structured output, remote MCP, and session resume all get more reliable●MODEL — Claude Fable 5 is generally available, with a 1M-token context window, always-on adaptive thinking, and 128K output●LINEUP — Opus 4.8, Sonnet 4.6, and Haiku 4.5 lead the lineup; pick the right one per task●CODE — Claude Code ships a broad quality and reliability update with /rewind, stronger MCP resilience, and steadier OAuth handling●CODE — CPU and memory use drops during streaming and long sessions, keeping always-on automation stable●ADMIN — New org model restrictions let administrators control which models are available●MCP — Structured output, remote MCP, and session resume all get more reliable●MODEL — Claude Fable 5 is generally available, with a 1M-token context window, always-on adaptive thinking, and 128K output●LINEUP — Opus 4.8, Sonnet 4.6, and Haiku 4.5 lead the lineup; pick the right one per task

TAG

self-repair

1 articles

← Back to all tags

Claude API¹ retry design¹ error handling¹ production¹

⬡ API & SDK/2026-06-27Advanced

Designing the Give-Up Condition in Self-Repair Loops: Four Error Classes, Four Retry Budgets

LLM self-repair loops break on the fantasy that 'if you keep fixing, it eventually passes.' Classify errors into four classes, give each its own retry budget. Working TypeScript and real cost numbers included.