Twitxy Saved

Robert Youssef

Verified account This account is verified. Learn more

@rryssf

Building @godofprompt & @njectai
Agentic Memory and Context Engineering
Global nject.ai Joined June 2023
104 Following 41K Followers 26K Posts

Latest posts 5

Robert Youssef Verified account @rryssf
the most useful line in theo's rant is the workaround

he lets 6.2 astra do the long run, then switches to opus 5.5 just to get a readable summary of what it did

you don't have to pick one model. route by job: one for the heavy work, a cheaper one that actually explains things for the handoff
Theo - t3.gg Verified account @theo 2h
View quoted post on X
Robert Youssef Verified account @rryssf
buried in the Sol 6.2 announcement and it matters more than the benchmarks

cached input is $0.40/M. fresh input is $8/M. that's 20x cheaper

and openai straight up tells you how to get it: keep your prompt prefix unchanged and append new instructions to the end

so if your agent rewrites the system prompt every turn, puts a timestamp at the top, or reshuffles tools, you're paying 20x for the same tokens

> static stuff first (system prompt, tools, docs)
> dynamic stuff last
> never touch the prefix mid-session

here is most of context engineering in one rule
OpenAI Verified account @OpenAI 3h
Some of you have been asking very politely. Some of you have just been asking.

GPT-6.2 Astra and Sol are now available.

Super efficient and built to perform where it counts—the real world.

openai.com/index/introduc…
View quoted post on X