thereeldeel

Will Codex App support new context window, rather than compaction, for "unrelated" sub-tasks during long horizon tasks?

The core issue with the traditional MCP/Tool Calling approach is that adding too many tools or having to chain multiple calls leads to models behaving worse and inefficiently using tokens. Cloudflare's new concept proposes converting MCP tools into a TypeScript API and letting the LLM write code to call that API, instead of listing all the tools for the model to call directly

Could you elaborate on the 'headaches of various vendor reviews and declarations'? (I thought paperwork etc was table stakes for a SOC 2 certificate)

What do you think about statements such that “AI has IQ of 155”, “It has same IQ as Elon Musk”? This seems detached LLM’s actual performance results!