This is why you need to have Local model with your own tools around it

ሰላም

A practical tip for anyone using Codex or AI coding agents on a large codebase:

You must design the codebase in a way that guides the agent.

In CRUder, I made Codex much more effective by setting hard architecture rules and restructuring the project for AI-assisted development:

✦ Split huge files into smaller focused files
• Kept files around 450 to 500 lines max
✦ Added folder-level AGENT.md guide files
• Defined what each folder does
• Explained what each file cover
- This Reduced unnecessary context loading
✦ Built the system in a modular, function-based structure
• Made the codebase easier to inspect
• Made edits safer and more predictable
• Reduced regressions during feature implementation

After this: Codex became more reliable, used context more efficiently.

AI coding agents are powerful, but they need proper engineering guidance.

Good architecture is not only for human developers. It is also how we enable AI collaborators to work at production scale.

This is a significant solution to reduce regressions caused by unrelated fixes.

While building CRUDER initially, the model just kept stacking solutions onto the initial basic files it created, so I lost some progress due to regressions and a disproportionate consumption of context resulting from oversized documents.

So I restructured the project's overall architecture the way I described above, and after that, I have seen no regressions at all, and the token consumption shrank by a good amount.

☆50 m people are fighting cancer,
☆173k peoole didn't wake up today,
☆ 80 m people can't walk,
☆300 m people can't see colors,
☆ 55 m people can't remember memories,
☆ 430 m can't hear.
So what i am trying to say is...

Even if you got problems to deal with ( i am not downgrading them!), your ordinary is someone's dream. So try to be grateful for what you have rn ☺️😗

https://www.youtube.com/watch?v=WNZehEJJPAg&list=RDWNZehEJJPAg&start_radio=1

ተጋበዙልኝ ከልቤ ነው የምወደው በያንዳድዱ ግጥሞቹ ላይ ያሉትን መልእክቶች አዳምጡት ወርቅ እንኳን ከዚ ያንሳል

በየቀኑ ቢነበብ የማይጠገብ
ነብሰን የሚያለመልም
ስርአትን የሚያስተምር
ከፈጣሪ ጋር የሚያቀራርበው
የዳዊት መዝሙር።

I believe that as a nation, our research and development capabilities, especially state-funded ones, should be explored and begin focusing on this area from the start, as this will give us an edge in the upcoming 20 to 50 years of technology.

It's been a while since GPT 5.5 and Cloude Opus 4.7 and 4.8 were out,

but for the first time in the past two months, out of the 11, I stepped back and prefer GPT 5.4 and Sonet 4.6 for many of my tasks.

I felt comfortable there; they do well, cost better, and they have this odd symphony when you use them together. It's empowering, honestly.

Disclaimer: i am not a fan of coffe it came with the image.

Agents' Last Exam is building the largest-scale, broadest-coverage agent evaluation benchmark to date, measuring performance on long-horizon, economically valuable tasks with verifiable outcomes.

https://agents-last-exam.org/

እስካሁን ያልተኛ እየሰራ ያለ ሰው ካለ ኮመንት ላይ ልያቹ