Roman

🔧 Engineering first, hype later. I’m building Sapphire ICDS, a high-performance platform written in modern C++ (C++20/23), with native modularity and deep AI integration as a core part of the system — not an add-on or a plugin. This isn’t a “framework” or a templating toy. ICDS is fully engineered software, built around: strict, documented interfaces signed native modules secure execution and low-level control AI treated as part of the architecture, not a feature The project is developed on Linux, focusing on performance, security, and long-term maintainability instead of buzzword-driven development. I share real engineering progress, architecture choices, and practical integration of AI into production-grade systems. If you care about software quality beyond trends, follow along.

@shikhr_ And price/codex limit use x2 over gpt 5.5🤣🤣🤣🙈

Imagine if GPT 5.6 absolutely mogs Opus 4.8

@ziwenxu_ Not bad

Stop everything you're doing right now Anthropic just dropped Claude 4.8 Time to cook! 🔥

@code_coded Image tool is good👍 With 5.4. Model 5.5 is rubish as I told and I avoid including it's integration into own product

@SapphireTeck That's quite against the public consensus though it seem. What about using imagegen within in? I love how Codex can iterate over it's own image generation prompt and review and re-review the image until it's finely tuned to the requirement. That's gotta count for something?

Anyone else feeling that Codex 5.5 has become frustratingly, near unusably slow for even simple tasks? Last 2 days it's been just horrendous. Having to jump back onto backup Claude account...maybe I cancelled Claude too earnestly =/

@TimJayas Hopefully it will be more like 5.4, not like 5.5 rubish

Goodbye Codex GPT-5.2 and GPT-5.3 GPT 5.6 Goblin incoming 👀

@thsottiaux Bad move. I hate 5.5. As only 5.4 will be taken away I cancel my pro subscription immediately. gpt 5.5 is terrible

To simplify our Codex compute fleet management, we will be sunsetting GPT-5.2 and GPT-5.3-Codex in Codex on June 2nd when logged in with your ChatGPT account. For free plans, GPT-5.5 will be the default frontier model to build and work with going forward. These models will remain available on our API.

@ziwenxu_ What a hell happened it melts that fast?

OpenAI Codex is burning credits at a terrifying pace. I've been using it lightly with 20x and it's already at 25% Imagine what's happening to people on the $20 or $100 plan Our extra usage vanishes in 3 days lol

@ziwenxu_ I run separate system user for codex and backup outside of that used so that ai has no access to it. and no backups is wild, man :D

I'm cooked. Codex just nuked my entire repo and everything inside it is gone 2 weeks of work he built for me that I never pushed to GitHub is now dust. Lessons 101: Always tell your AI to push to GitHub for every small win they complete

@drummatick 49h in a row???? What goal it was?👀

I am proud of you codex, you may rest now

@Conor_D_Dart @OpenAI Ah, I gnored from start gpt 5.5. Skipped it and still using 5.4. Cheaper, better, stable, reliable

I feel like GPT 5.5 isn't the same anymore. Putting 5.4 just fixed the issues I had, so I am telling you there is an issue with 5.5 whether @openai knows or doesn't. But us - the customers are smart and we notice the little changes. Things changed 3 resets ago. 5.5 on launch was king of the hill.

@maisha_anj63027 Grok delete, claude not in use, so count deleted, gemini also not in use.

Can you delete any one of them??

@md_kasif_uddin Gpt 5.4

Which AI model gives the most accurate responses?

@notjazii Does that mean you are that weak in programming?

me trying to code without claude, codex, cursor

@thsottiaux If lower cost/%usage - definitely

Should we bring batch compute to codex? Aka /slow mode

@ziwenxu_ Yep. Maybe Tibo has a holiday today

@SapphireTeck Fingers crossed 🤞

OpenAI just silently fixed the Codex usage bug everyone was complaining about Now we just need a reset

@LexnLin Who told you, that reset is coming at all? Bug being fixed != reset coming

Codex reset incoming

@ziwenxu_ I also have huge hopes on that, because their bug realy drained quota fast

@ziwenxu_ I have 39K c++ lines of code to port, so max tomorrow yet enough, then empty🙈

@ziwenxu_ I hope dj will hit reset button soon through

@SapphireTeck ok, you are saving up for the weekend lol

@ziwenxu_ I have 37% left yet😁

@SapphireTeck :( then i have to wait for 3 more days