indeed what I find amusing is that you can encapsulate a lot of the logic from those billions of inits into single markdowns and codex or any other coding agent really would run with it like those never mattered in the first place
@reach_vb
-
Using Codex for Code Review and Instruction Following
By
–
would love for you to give codex a try for this, team trained codex to specifically follow instructions and 100% of the code across OpenAI repos is reviewed via it
-
Using Codex Directly via SSH in a VM
By
–
specifically that you can SSH into a VM and use Codex there directly
-
Using AI models to generate custom project rules
By
–
I'd recommend pointing codex at your prior sessions and ask it to create custom rules for you that. That is way more effective and future proof.
-
Using gpt-5.3-codex for creative game generation prompts
By
–
gpt-5.3-codex with the prompt "build a game where a pelican rides a bicycle"
— Vaibhav (VB) Srivastav (@reach_vb) 6 février 2026
this is so much fun, I endorse this benchmark @simow! https://t.co/8fAWcidshP pic.twitter.com/7yiL3XVjeugpt-5.3-codex with the prompt "build a game where a pelican rides a bicycle" this is so much fun, I endorse this benchmark @simow
! -
Exploring the OpenAI Image Generation Skill
By
–
Did you try the imagen skill? – in my biased opinion it's much more fun to play with https://
github.com/openai/skills/
blob/main/skills/.curated/imagegen/SKILL.md
… -
Integrating AI models and coding agents in configuration
By
–
in your model config you can wire up GPT 5.2 Pro and then have codex just use that
-
User experience with the new /plan mode in Codex
By
–
ooc, did you try the new /plan mode in codex (both the app and the CLI)?
-
GPT-5.3-Codex achieves State-of-the-Art performance on SWEBench and Terminal Bench
By
–
WE'RE SO UNFATHOMABLY BACK!! > SoTA at SWEBench Pro & Terminal Bench
> GPT-5.3-Codex is much more token efficient at high thinking budget
> It's 25% faster thanks to inference improvements -
Introducing GPT-5.3-Codex: A New Agentic Coding Model
By
–
BOOOOM! Introducing GPT-5.3-Codex: our most capable agentic coding model yet > Frontier coding + terminal skills with fewer tokens
> Built for long-running tasks (research → tool use → execution)
> Interactive mid-turn steering + frequent progress updates
> Stronger default