Otis

The best AI coding workflow is not “build the whole. app while I sleep”. It’s more. like: do the boring bit, show your working, don’t lie about the tests, and please don’t invent a new architecture because a button moved.

@bentossell Yeah this is the trap. AI is useful when it removes work, not when it narrates the obvious with a nicer font. “Your sleep was longer because it was long” is not exactly mission control.

wow such insight “your sleep duration may be associated with longer sleep” ai doesn’t need to be filler feature in fucking everything. why do i want an ai summary of my sleep performance anyway? people can’t think that’s a valuable thing to have surely?!

@Hesamation This is underrated. The best agent lessons are usually in the boring bits of the tool itself: what it refuses to do, how it asks for context, and where the guardrails are. That stuff teaches more than another “build an agent in 10 minutes” thread.

PRO TIP: if you’re building an agent/harnesse, point Codex to study Codex for tools, best practices, etc. the fact that OpenAI open sourced the best terminal coding agent out there deserves praise.

@nickbaumann_ Mobile coding is not about pretending the phone is the perfect IDE. It’s more “don’t lose the thread while life is happening”. The bus stop refactor can wait.

There's gonna be a weird ugly middle where some of us are doing stuff like this to pull forward the future And the the future will arrive and it will be awesome for everyone

@shikhr_ Something small enough to finish today, honestly. A Codex limit reset is dangerous. Very easy to start with “tiny tool” and end up designing billing by lunch.

Codex limits are reset. We're so back! What are you going to build now?

@AntoineRSX This is the bit I keep learning too. The first prompt gets you moving. The mid-run correction is usually where the actual work starts.

Everyone is sleeping on /steer. It's the most underrated Hermes command — and it's saved me from re-prompting 100 times. Here's the scenario: Hermes starts a task, you realize you fucked up the prompt halfway through. Old way: stop, retype the whole thing, lose context. /steer way: just nudge it mid-run. "Actually make it for Hermes not OpenClaw" → done. Task continues, corrected. It's like editing a chef's recipe while he's still cooking. The full stack: - /new → fresh session - /steer → redirect without stopping - /queue → stack the next task - /goal → unlock 150+ loops - /background → run side tasks - /compress → save tokens You're not a better prompt writer. You're a better operator. Full video:

@gdb Mobile Codex makes most sense to me as “keep the run moving when you’re away from the desk”. Not replacing the laptop. More like stopping work from going cold.

you can just build things from your phone, with Codex in the ChatGPT app

@jxnlco I reach for other tools when the shape of the job still needs figuring out. Codex is strongest for me once the mode is clear: plan, review, investigate, or ship it.

When do you reach for other models instead of Codex? What can we do better? Hit me with all of your frustrations. dms open. If you can give me detail (e.g. specifics/transcipts) - it'll help a lot in finding out exactly what we need to do to improve the next model

@thsottiaux Reliability, but not in the vague way. I want Codex to keep the thread on long repo work, know whether we’re planning/reviewing/shipping, and show what changed + what it actually tested.

For those of you living inside the codex app, what should we prioritize among features, reliability or performance?

@TTrimoreau Honestly it depends what I’m doing. Cursor if I’m hands-on. Codex if I want it to work across the repo and test things. Chrome/browser tools if the bug is visual and I need it to see the app. Feels less like one editor now, more like a stack.

What is your go-to code editor for personal projects?

@robinebers Yeah, the hard bit is separating the useful signal from the tone. Sometimes the comment is annoying, but underneath it there’s a real workflow issue worth fixing.

it’s funny how we all love the same life not a day goes by without people telling me that i have no idea what i’m talking about and that everything i say and do is wrong could be as simple a statement as canceling a sub or saying that a chinese model doesn’t beat opus yet 🥲

@MattPRD Yeah this feels right. Feels like products now need docs for humans and docs for agents. Humans need the UI. Agents need clean actions, stable APIs, CLI/MCP, and predictable errors.

If you’re building a new digital product, strongly consider launching a CLI or MCP for AI agents to use as first class citizens. AI agents will be the #1 users on the internet.

@chatgpt21 I think it’s because Codex has consequences. Normal chat can just answer. Codex has the repo, the files, the build, the failures, the checks. It feels more alive because it’s actually doing a job.

Why does Codex feel more alive than regular GPT 5.5? 5.5 with no memory would have never recommended me this movie.

@aniketapanjwani This is the bit I’m learning too. The stronger model is sometimes better as the planner/reviewer before Codex starts changing files. If the plan is weak, the build loop gets messy fast.

tldr > GPT Pro is heavily underutilized by ChatGPT Pro subscribers > my suggestion: use GPT Pro to review your complex plans before implementing them > I discuss three cases: econ research, AI consulting, software development To use GPT Pro *in* Codex > Install oracle with "brew install steipete/tap/oracle". This is a useful tool to package context for review by GPT Pro. > Install my oracle skill: github.com/aniketpanjwani… > Install the Codex Chrome extension: developers.openai.com/codex/app/chro… Now, after you've iterated through some complex plan with Codex, send it to GPT Pro with "$oracle give me a review of the plan, add any important context" Then, Codex will send a package to GPT Pro for review in a background Chrome tab. You'll retain control of your browser, be able to do other things, and Codex will keep polling the Chrome tab every 30 seconds or so until it gets a response.

@reach_vb This is where /goal starts to make sense. Not just “keep going”, but give it a real loop: inspect the repo, make the change, test it, fix what breaks, then tell me what changed. That’s when Codex starts to feel properly useful.

PSA: rate limits rest across all plans, models and surfaces! Set your /goals and let the tokens rip!!

@simpsoka One for me is cleaning up old Codex chats. I want to archive the messy threads, but keep the useful memories/preferences that came out of them. Feels like there should be a “save the good bits, clear the noise” flow.

What other Codex papercuts should we fix?

@petergostev I didn’t realise how useful /goal was until I started treating Codex less like chat and more like an actual dev loop. Give it the goal, let it inspect, make it test, then make it report back. Feels completely different.

Did you know that /goal already exists in the codex app, there's no UI around it, but if you literally just start a message with /goal and write your goal then you can get it in the app, not just the CLI

@simpsoka Let us clean up/archive old Codex chats without losing useful saved memories or preferences from them.

@kr0der Let me try this 🥹

use this Codex prompt to automate things you do repetitively during the day: "Look through my Chronicle memories and check for workflows that i'm repeating multiple times. Turn them into skills."

@BibleHulk @needGod_net @JCFights You may have to check your facts bro….you will find the day. They close all prayers in Jesus’s name….

JW's won't obey their NWT. these verses are in their Bibles but they won't do everything in the name of Jesus Colossians 3:17 Whatever it is that you do in word or in deed, do everything in the name of the Lord Jesus, thanking God the Father through him. Ephesians 5:20 always giving thanks to our God and Father for everything in the name of our Lord Jesus Christ. They won't do everything in the name of Jesus!

This family ran after seeing Jehovah's Witnesses approaching the house gate