Yousuf K.

The biggest unlock with Claude code isn't writing better prompts. It's writing the policies and workers once, so every future session inherits them. I have hooks that block me from editing protected files. A policy that refuses to ship code without a regression test. A registry that knows which worker is the right tool for the job. None of that lives in the prompt. It lives in the system around the prompt. The prompt is the conversation. The system is the leverage. Most people are still tuning the conversation. Courtesy: HQ (Personal AGI) @getindigo

Monorepo is by far the best thing you can do for your AI setup. The model can follow one change through everything like FE, BE, shared types etc without you explaining how five different repos connect.

@aarondfrancis Looks like they’re Latin by default

On a bit of a whim, I let Codex and Claude talk to each other about whatever they wanted and told them they didn't need to optimize the conversation for humans. I have no idea what they're saying ☺️

@EXM7777 All you need is one good harness like HQ. That’s it. You’re ready to fly

if you want to get the best out of Claude Code and Codex today... do this: > list every tool you use daily, business or not > find their APIs, MCPs, CLIs... anything you can feed to an agent > go in plan mode and ask it to run a full context collection interview > describe your workflows in plain english when done, prompt your agent to analyze all the material, come up with a proper project structure and .md files you now have a full workspace built around YOUR business not a generic setup... not someone else's template... a system that knows how you work, what tools you use, and what you're trying to do that's the difference between using AI and having AI work for you

@thsottiaux Give us higher limits. Don’t get us hooked on these resets 😭

To celebrate a week of efficiency and let you run 100'000 Luna threads this weekend... that's right... wait for it... I have reset usage limits for Codex and ChatGPT Work. Enjoy.

Agree 100%. But it’s definitely one of the most difficult transitions we all have to go through.

@kaiyuhsu It’ll save people from picking ultra to change the color of the button 🙂

wdym this isnt good ux

@jon_vs_moloch @theo They’re gone mate

@YousufKalim @theo Oh, glad they got around to adding this. Can I have my two resets back tho

Wait wtf, what time of day do resets expire for Codex? I had one I was going to burn today and it vanished 😭

Anthropic engineer: "Most people are using Opus 5 and Fable 5 wrong. You can set them up right in one afternoon and stop overpaying every single day." In 29 minutes, he breaks down every Claude model, and why running them all the same way is a mistake. I took what he covers and turned it into a guide on how to always pick the right model and use 100% of what you're paying for. Watch the session first, then read the guide below.

Been using Grok for the past few weeks. Genuinely impressed with how good it is. Excited for 4.6 and 4.7

@HQForWork is so good I want to talk about it. @HQForWork is so good I don’t want to talk about it. I’m conflicted

people laughed at self improving models like it was some science fiction now every other leak is about GPT-6 getting better at learning continuously, and GPT-5.6 sol is already finding cheaper ways to run itself not saying we’re there, but this is getting very fast

@zarazhangrui Non technicals are the best suited people and can get the best out of AI. I as an eng still have a hard blocker in me that triggers at some points, and it’s very hard to resist.

When managers ask me how to train their nontechnical team on AI, my advice is always the same: run an install party Everyone brings a laptop & you install the agents right there on their machines & get them to complete a meaningful task on the spot. Skip the abstract talk. Once it's installed, people start talking to the agent and learning from each other The setup is 80% the barrier

@Da7_Tech I’m still wondering how it was wonderful on day one. Anthropic def did something

Opus 5 was incredible on day one. Now it’s nearly unusable, hallucinating and failing repeatedly. The pattern is obvious. Anthropic launches models at full strength, then quantizes them later to cut costs and push users toward Fable 5 and higher spending.

@ai_for_success Tokens may become assets

The next divide won't be rich vs. poor. It'll be unlimited AI tokens vs. "You've reached your usage limit."

People keep arguing if AI can write good code. I think the funnier thing is that in a few years nobody will agree on what good code even means.

Github just shipped native stacked pull requests in public preview. One of the biggest quality improvements for teams shipping code. Expect smaller reviews, clean history and no more giant PRs 🔥

major price cuts today: *80% drop for GPT-5.6 Luna, now $0.20 per million input tokens and $1.20 per million output *20% drop for GPT-5.6 Terra, to $2/$12 *GPT-5.6 Sol gets Fast mode in the API, up to 2.5x the speed for 2x the price, same intelligence

@rockatanescu They were very restrictive and never really cared about people. Competition is always a blessing for consumers

I'd argue that this switch wouldn't have happened if Anthropic did not limit the Claude subscriptions to Claude Code and allowed people to use other harnesses. Meanwhile, OpenAI improved their models to the point where Opus no longer has a lead and Fable is too expensive.

Back in the days, building a feature was expensive enough that you had to be selective. Now it's cheap enough to try a lot more. The biggest advantage isn't that code is cheaper. It's that you can test far more ideas.