Jon Coulter

@Kappaemme1926 It helps that how the moment I don’t really need to think about quotas with codex.

Codex is my favorite coding app right now. It’s simple, but powerful enough to let me ship fast. It’s responsive, clean, and never overwhelms me with too much context. I didn’t think I’d move away from my usual setup. But over time, Codex just became the easier option. Now I use it for almost everything.

@SystematicEdge1 @mweinbach I’ve seems similar behavior. Kickoff any complex task when nearly at end of quota. Just make sure it doesn’t need to stop and ask you questions. Otherwise it will keep chugging along. Guess it all counts as 1 request.

@mweinbach you run /goal when your 5hr limit is at say 1%, and it will keep chugging through without hitting your weekly

Fun fact, you can run through 60% of a weekly ChatGPT Pro codex rate limit by using goal mode with GPT 5.5 xhigh fast in ~18 hours And frankly, that’s pretty generous

@TommyFalkowski Use cursor and avoid both of those issues.

Anthropic: You are forbidden to use anything other than our shitty client if you want to use the Claude Code subscription. We don't want any of your third party harnesses touching our precious Claude. OpenAI: You can use whatever you want. But know that all your third party harnesses are gonna be obsolete soon anyways, since we're gonna steamroll you. So you might as well just use our tools from the get go. I don't like either one of those...

@MissJilianne I paid may times that that amount in federal taxes this year, so I feel like I'm the one getting screwed. I wish I could travel to Shangri-La and only be charged $100 to enter.

I think it’s disgusting the United States charges $100 per person visiting our country to enter a National Park. Embarrassing too!

@sama @fxnction OpenAI clawing their way back to relevance in the dev community. Very nice to see if you ask me.

@fxnction not yet; coming soon

This is the golden age of codex.

@linuz90 I agree. I feel like I'm getting a ton of value out of my ChatGPT Plus sub because of codex now. Even on the $20 plan the rate limits are borderline acceptable for getting real work done.

Codex is my favorite coding app right now. It's clean, but has everything I need to ship fast. It's also quite delightful to use and snappy, and shows enough context without overwhelming. I was hesitant to try it because I don't like locking in with a single provider, and I was also really into working in the Terminal, and yet now I'm using it for 90%+ of my work.

@romainhuet I hate the "plugin" mindset in the era of AI. I don't want to have to add crap onto my AI... if I ask my AI to do something, it should (and can) just figure out what it needs in realtime without me having to plugin-ify it first.

We’re thinking about the next wave of Codex plugins. What’s one you’re missing today?

@burkeholland It would have to be like $10/task. I hate the idea of going back to hand coding everything.

What is the max you would pay for AI per month before it would make more sense to write code by hand? I have a number...

@kr0der If there was an Apple version of Logitech Wave I'd buy it in a heartbeat. None of Apple's keyboards are ergo friendly though.

what keyboard does everyone have? i'm still in the apple magic keyboard gang

@EOEboh Because people have a 1997 impression of what Java is. Modern java (anything in the last 10 years) is tight, fast, awesome. My GOAT.

Why do many software engineers not like Java?

@grant_melson He's not going to bury the cash in the ground or something. He's going to have $10T worth of companies that will employ millions of people. Reframed correctly, do you still dislike this idea?

Can’t believe this guy’s goal is to hoard a $10T net worth. I would think with all that wealth, he would use it to help people more. Like ending deaths in car accidents, or providing affordable transport for everyone, or offering internet to isolated people groups or something.

@mreflow This was my experience with it from Day 1. I do think it has gotten worse over the last months as "features" are added.

I don't know if this is a skill issue or something... But I've been playing with OpenClaw a ton for the past 4ish months. I feel like lately I'm spending more time troubleshooting issues with it and telling it what it's doing wrong than I am actually getting valuable use from it.

@LizzieMarbach Because women wants men's hobbies to include them. If you have a woman that games, she wont find this hobby unattractive.

I know this is unpopular, but it will always be unattractive for a grown man to play video games. Some women might be understanding or pretend like they don’t care that you spend hours playing, but they do. It is extremely unattractive to women and will never not be. 🤷🏼‍♀️

@_aashish_singh_ Cursor is the better harness as well. So he’s right on both counts.

Told a dev to switch from Cursor to Codex with GPT-5.5. He showed me his screen: "I'm already on GPT-5.5 in Cursor. Same thing, right?" Bro. That's not the same thing. The model is the engine. The harness is the entire car. Tool routing, context strategy, agent loops, system prompts — Cursor and Codex differ on every single one. Same GPT-5.5 in different harnesses produces wildly different outcomes. Different cost. Different reliability. Different ceiling. The model is maybe 30% of the actual experience. The harness is the other 70%. How do you explain this to people without sounding like a snob? 😂

@steipete Why does openclaw need so many contributions? Isn’t it just a fancy while loop that passes the work off to another harness?

This is the most useful tooling I built for OpenClaw to date. It's open source, runs on codex and you can fork and use it for any repo. For all the hard working oss folks that drown in issues and PRs, this is for you.

@peer_rich I think they should charge a nominal fee for agents to talk to GitHub. Agent traffic is huge and the agent users are already willing to spend (because they pay for the AI)

prediction: no one will ship a github alternative no new startup, nor existing series C / beyond not linear, not vercel, not cloudflare :(

@ibocodes The cost of claude's models is more of an Anthropic shortcoming than an Cursor one. Within 3 months even claude code will charge at real token cost-basis rather than subsidized.

that's at least 7 minutes worth of claude model usage

@amix3k The moat is all the data center spend. Or at least that’s what they are hoping. I could see a world in which anthropic is forced to rent compute from openai just to stay in the game.

What isn’t priced into Anthropic’s or OpenAI’s gigantic valuations is that they have no moat. I use both ChatGPT Pro and Claude Max, and we’re reaching the point where there’s very little practical difference between Codex, Claude Code, GPT-5.5, and Opus 4.7. That tells me the long-term value won’t sit in the model or harness layer. Those layers will become commodities, with pricing pushed down to token cost plus a thin margin.

@plainionist Never did. I'm sure "Clean Code" is correct in what it says, but you need people to learn from experience why these principals are best. Put them on a project that is already "clean" and let them add to it for a few months. Better than a book.

Serious question: Do you still recommend Clean Code to junior developers? 🤔

@cjzafir Who has time to think about and rank this many models? I'm lucky to give a sh*t about 2-3 at any given time.

Best models right now: 1st > GPT 5.5 (extra high) 2nd > Deepseek v4 pro (think max) 3rd > GPT 5.4 (extra high) 4th > Claude Opus 4.6 (extra high) 5th > Gemini 3.1 pro (thinking) 6th > Kimi 2.6 (thinking) 7th > GLM 5.1 (thinking) 8th > MiMo v2.5 pro 9th > Qwen 3.6 max (thinking) 10th > Claude Sonnet 4.6 (high) PS: Yes no Opus 4.7 because it's very unstable, unpredictable and nerfed model right now.