Designing Loops for Production-Grade Work — Blog — Liquid AI

kept by eddie aug 19

Liquid AI shows that coding agents can autonomously build a production‑grade BPE tokenizer trainer only when given iterative loops, real‑scale data, and external verification.

Opus 4.5 is not the normal AI agent experience that I have had thus far
Hacker News commenters report that Opus 4.5 enables rapid development of small projects but often produces overly complex, unmaintainable code unsuitable for production.
also on Claude Opus 4.5, #ai-agents, #ai-coding, #software-engineering, #developer-tools
Claude Codes #3 - by Zvi Mowshowitz
Claude Code and Claude Cowork dominate AI coding tools, with Anthropic's safety approach enabling permissive agents, but users warned against over-optimizing setups.
also on Claude Opus 4.5, #ai-agents, #ai-coding, #developer-tools