Petr Glaser

@borekb Jo, je to prostě vykradený model, aniž by správně uvedli licenci... parchanti.

Zajímavé, ale asi logické. Takže Composer 2 by měl být fakt dobrý 😄

@JaroslavKlim @borekb Codex je miláček. Zdaleka nejvyšší kvalita kódu, nejspolehlivější. <3

@borekb Codex je hračka, takový kámoš junior

Můj aktuální názor: - Kdo nepoužívá Codex, výrazně si škodí. - Kdo používá pouze Codex, výrazně si škodí.

@elegantnikuryr @borekb A na jaké use-cases? Na FE je GPT 5.3 Codex i GPT 5.4 (neexistuje codex verze) mnohem horší, ale i tak z něj můžou lézt moc pěkné výsledky.

@borekb No pokud zalezi na vysledku tak za me ne. Tandem CC a Gemini v Antigravity je uplne jinde nez Codex. O kontextovem okne ani nemluve. Tim nerikam ze Codex je blbost, jen me v agentickem fungovani nedava smysl ho pouzivat, ze to muze byt lepší model, asi jo.

@aichinge @borekb To je trošku hloupé srovnání. Codex není model, když už tak 5.3 Codex, který ale byl nahrazen 5.4 většinově (byť se občas vracím k 5.3 Codex). Ale když tu otázku vezmu takhle upravenou, tak... je mnohem spolehlivější. Výsledky z něj lezou mnohem lepší, víc production-ready.

@borekb Na co je podl tebe lepsi / horsi Codex oproti Opus 4.6 (1M)?

@leerob Are you scamming people and not providing correct licenses that your model is based on Kimi K2.5?

Should I do a live stream tomorrow to talk about Composer 2 and new Cursor updates? Lmk what questions I can answer or demos to show.

@theo Do you plan to add Pi (or even Oh My Pi)? Or should I just rely on @kcosr's AppServer plugin for that?

T3 Code now supports Claude. If you have the Claude Code CLI installed and signed in, you can use it with T3 Code. Hopefully the lawyers won't make us remove this 🙃

@kcosr Love it. I'll use it for my alternative to Codex Desktop. I'm already building on top of App Server.

Pi in Codex Desktop. Incomplete but functional. Maybe fun for someone to hack on. github.com/kcosr/codapter/

@sawyerhood What do you plan to do with dev-browser? How do you plan to beat agent-browser and playwright cli interactive? :)

does codex have plugins yet? i'm working on a revamped version of dev-browser and want to make sure that i support them (if they exist)

@dillon_mulroy I have this SO. FUCKING. MUCH.

5.4's idea of cleaning up stuff

@ScriptedAlchemy Did you also try 5.3 Codex Spark with your setup?

OpenAI cooked with GPT 5.4 mini. Spent the whole day having 5.4 high drive the mini model for all writes in codebases.

@nstlopez Looking forward to that!

I'll be speaking over here next month :)

@wesbos I just... wanna stay in @rspack_dev ecosystem. It just feels better. Maybe Vite will change with move to Rolldown and Vite8, but I am not sure yet... testing it right now.

Vite+ is SO GOOD I’ve moved 3 sites over and scaffolded out a few more. So sold

@thekitze Better than T3 Code? Better than Schaltwerk?

the Codex app is a trillion and one time better than the codex cli and any other cli as a matter of fact. fuck tui, gimme ui all day every day

@aidenybai Sounds like cmux. :D

what if ghostty had vertical tabs? i'm too lazy to learn tmux and i want an interactive UI to manage my agents/terminals

@ScriptedAlchemy I've been using that on my AI Course when I ran it. Even the lowest package had more than Claude Max. Not sure about current situation. Or if @pierceboggan adds GPT 5.4 Mini as free model, then GHC subscription would be nuts.

@BleedingDev It’s sold out all the time! But yeah I been looking. GLM seems to be the best. How much token use you get out of GLM? You tried their subs?

God I love gpt, but I can’t keep spending $80 a day buying more credits. I’m gonna try 5.4-mini and see how that does with 5.4 high as the driver. Work instance is basically 429 error 24 hours a day 🥲 I’m going to have to look at GLM to offload. Only thing keeping me going is @cursor_ai credits they gave me 💕

@pierceboggan @WonderingDavid Yes please, replace it, make it 0x.

@WonderingDavid For 0x?

Pretty please @pierceboggan can this replace 5 mini in copilot 🙏

@kcosr Maybe @reach_vb can take a look at it and push it towards OpenAI engineers?

@kcosr You are not doing it wrong. It keeps killing them, waiting for them, thinking what happened to them. And 5.4 is worse at it then 5.3 Codex.

Guys, am I doing Codex subagents wrong? It keeps stopping and getting confused about files subagents created.

@thekitze Only that one thing pushes me more towards joining Tinkerers Club. :D

the best agent orchestrator i've tried yet has this badge and is available only to a certain group of cracked ppl 🤓