Codex
OpenAI's AI coding agent, available as an open-source CLI, desktop app and IDE extension.
Liked for solid code output,123 for cross-reviewing code alongside Claude Code,456 and for automating varied jobs from homelab management to porting legacy apps.78910 The most common complaint is buggy, unreliable tooling,1112131415 including session logs that can consume hundreds of gigabytes of disk.1617
Why people recommend it
Broad automation · 3
Commenters use it for far more than app code, from homelab management to collecting web pages and maintaining a llama.cpp build.
“I use Claude or codex to manage the entire thing, interchangeably.”
“Claude Code/Codex has made my entire macOS environment as malleable as Emacs. That’s huge!”
Show 1 more
“I have a perpetual codex thread open in my llama.cpp checkout that I periodically ask to take a look at currently pending llama.cpp PRs”
Output quality · 3
Several say it gives effective, often higher-quality results and is the agent they use most.
“I use Codex far more than I use Claude for coding and my results are generally higher quality.”
“And I use codex more.”
Show 1 more
“I find Codex to be quite effective but it doesn't feel familiar.”
Cross-model review · 3
Commenters pair it with Claude Code so each reviews the other's code.
“if I code with Claude (usually Sonnet 5.5-medium), Codex reviews it, and vice versa”
“I generally start a new Claude Code and a Codex and have them review the code”
Show 1 more
“a two part adversarial code review with claude code & codex plus automated reviews from coderabbit”
Porting old code · 3
Used to transpile or revive old PHP, Visual Basic 6 and Go codebases.
“had codex transpile it to rust”
“I used Codex to convert some of my old Visual Basic 6 apps and games and it was a joy to see them running again”
Show 1 more
“Luckily LLMs are quite good at these surgical changes. In fact I only used ~10% of my weekly limit on Codex”
Easier than alternatives · 3
Some find it less buggy or easier to manage than Claude Code's harness or OpenCode.
“I've mostly switched from Claude to Codex because I find the harness so buggy and annoying”
“compared to managing Claude and Codex, it's such a pain”
Show 1 more
“Codex will emit a lightweight ream of "thinking", then on completion it will erase this and only show the final, coherent answer.”
Open source · 2
The harness is open source and written in Rust, so users can read how it works.
“Codex is open source [1] and written in Rust.”
“Since Codex is open source, you can even take a look at how they do it.”
Caveats and dissent
Bugs · 3
Users report frequent bugs, TUI display glitches, a broken VS Code extension and regressions that block tasks.
“it is extremely annoying how many bugs Codex has. I basically can't trust it to execute on any goal”
“the true state of their toolings is closer to a burning dumpster fire. If you look at Codex Github repo for example, the amount of existing issues”
Show 1 more
Disk usage · 2
Session logs can grow to many gigabytes, with one user losing 330GB.
“even normal happy-path codex creates many GB of logs”
“this stupid bug that ate up 330GB of my Mac's disk space”
Sandbox problems · 3
Security blocks can lock up sessions, and one user saw it write to a folder granted read-only access.
“Codex was trying to use gdb to debug a capture crash and the session locked up with security blocks.”
“one day Codex+GLM went ahead and did things in my ~/.dotfiles folder which I had granted only READ access for”
Show 1 more
“all the cruft OpenAI loads by default before you even send your own messages”
Speed · 2
Some find it the slowest harness they use, and subagents made tasks much slower.
“openai models, and codex are clearly the slowest”
“When i used subagents with codex, it made tasks take literally 10x longer with a worse result”
Token usage · 3
Default context loading and heavy tasks consume many tokens, and one user found the $20 plan too limited.
“all the cruft OpenAI loads by default before you even send your own messages”
“Highly recommend. Uses quite a lot of tokens.”
Show 1 more
“I use $20 codex subscription, it's practically useless for anything other than luna.”
Account bans · 1
One user warns of permanent loss of access after ToS warnings.
“Losing access permanently to Codex made me feel like a second range citizen ever since.”
Often compared with
43 moved from Claude Code to Codex
3 moved from Codex to Claude CodeOn the home page's Debates
47 moved from Claude to Codex
2 moved from Codex to Claude
2 moved from OpenCode to Codex
3 moved from Codex to OpenCode
6 moved from Claude Opus to Codex
2 moved from Codex to Claude OpusOn the home page's Debates
10 moved from Cursor to Codex
2 moved from Codex to CursorPi80 comments mention both
8 moved from Codex to Pi
3 moved from Claude Fable to CodexOn the home page's Debates
2 moved from Gemini to Codex
Who recommends it in their own words
From the makers not counted in the ranking
“I'm more familiar with Codex than Claude as I worked on Codex when I was at OpenAI”
“Ongoing maintenance of my badly-named open-sourced tool called Codex”
“Disclaimer, I work on Codex since about a month ago.”
Threads where it came up
- GPT-5.6 24 mentions
- Ask HN: What is your (AI) dev tech stack / workflow? (June 2026) 15 mentions
- GPT-6 Sol and Luna 10 mentions
- Ask HN: How do you manage skills files? 8 mentions
- Ask HN: Who wants to be hired? (October 2026) 7 mentions
- A week of using Codex more than Claude 7 mentions
Also written as Codex CLI, OpenAI Codex, Codex 5.5, OpenAI Tokenizer. Data: codex.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 55
“Codex will emit a lightweight ream of "thinking", then on completion it will erase this and only show the final, coherent answer.”
“I use Claude Code, Codex, opencode and a couple other but I do not like the DX/UX.”
“My Codex and Claude Code bots (and Copilot at work) run on it pretty much around the clock”
“it runs Claude code, codex, opencode, pi, whatever model is appropriate”
“I use Claude, Codex and Cursor/SpaceX agents”
“I used Claude and Codex to label the symbols”
“I used a prompt similar to what I use with Codex and Claude Code”
“Been sitting on 5.6 / Astra and Codex since.”
“I built it with Claude Code and Codex.”
“I used codex in computer mode to collect the relevant pages to my project.”
“Majority of the people at my company switched to Codex.”
“Codex models have no limitations on reverse engineering”
“through Claude Code and Codex, 766 runs in total”
“of the set {Claude Code, Codex, Pi, OMP} I must say I find CC to be the worst.”
“Also should work in as pair of Astra + Luna in CodeX.”
“I was given _free_ Codex and realized... it still wasn't worth it.”
“I was able to transcribe the textbook and make straightforward corrections with Codex + GPT 6.1 Sol.”
“I find Codex to be quite effective but it doesn't feel familiar.”
“I used Zed Agent with Codex as it has native integration with Zed's ACP so it's pretty convenient.”
“Challenge source (mostly courtesy of Codex...)”
“it kinda stopped running tasks. I would ask it to do something and the it would just kind of do an analysis, then give me a recommendation”
“I will give it a try (aka ask Codex).”
“I love interacting with Codex models more”
“a common UI across the Claude / Codex / Oh My Pi that I use”
“created in 3 weeks using a combination of Claude Code and Codex”
“running Claude Code or Codex on their bare metal”
“we used Codex to collaboratively build an alloy model”
“Luckily LLMs are quite good at these surgical changes. In fact I only used ~10% of my weekly limit on Codex”
“codex told me it didn't need me..it could call claude by itself”
“it cannot match the simplicity of a "marketer" opening Codex/Claude desktop app and writing a prompt”
“When should I use this over Codex?”
“supplement my codex usage”
“when I run Codex at work, Codex does not have access to my Azure AI Foundry API credentials”
“ask claude code or codex or whatever harness to explain things you don't understand”
“Codex was trying to use gdb to debug a capture crash and the session locked up with security blocks.”
“Personal accounts I run the $17 Anthropic plan + $100 Codex Pro + DeepSeek.”
“And I use codex more.”
“I use codex/claude daily”
“I made a ZX Spectrum development environment with Codex”
“Both Codex[0] and Claude Code[1] output OTEL traces that you can ingest and process.”
“even normal happy-path codex creates many GB of logs”
“if I code with Claude (usually Sonnet 5.5-medium), Codex reviews it, and vice versa”
“this stupid bug that ate up 330GB of my Mac's disk space”
“I generally start a new Claude Code and a Codex and have them review the code”
“Claude Code/Codex has made my entire macOS environment as malleable as Emacs. That’s huge!”
“I use Claude or codex to manage the entire thing, interchangeably.”
“the Codex TUI has constant display issues for me”
“I don't know how and why they ruined this CLI with needless "features".”
“it is extremely annoying how many bugs Codex has. I basically can't trust it to execute on any goal”
“had codex transpile it to rust”
“the true state of their toolings is closer to a burning dumpster fire. If you look at Codex Github repo for example, the amount of existing issues”
“At this point though, codex is completely unusable for me”
“a two part adversarial code review with claude code & codex plus automated reviews from coderabbit”
“I used Codex to convert some of my old Visual Basic 6 apps and games and it was a joy to see them running again”
“I use Codex far more than I use Claude for coding and my results are generally higher quality.”