Martin

hahah that was in jest at mario - he's definitely the brain child of it and it's so good. we used a very similar idea in our Think stuff we shipped today that was inspired by him/pi as far as my workflow goes - if i know exactly what i want and need to build i use what i guess is called "socratic method" for building up the agents initial context - basically asking it questions i know the answers to (e.g. how does x api work in y library, how is it implemented. what are the existing patterns for doing z in this codebase) so it build up its understanding. if the topics are kinda deep dives/longer thread i rewind with /tree picking out the import things it discovered. i'll continue that process for what i want to build e.g. "how would you implement a feature that x,y,z in this code, how would manage this constraint, and that edge case" at this point once it's speculated/"planned" an implementation I fire up plannator (plannotator.ai) and iterate back and forth using /plannotator-last (e.g. review and annotate the agents last message which coincidentally is kind of like a plan at this point) once i'm happy with it's understanding and approach i /tree back with the clear summary of how to implement and let it rip (sometimes i'll use @mattpocockuk's tdd skill here) once it's done its initial implementation i /tree back to the the clear summary of how to implement and then use /plannotator-review and iterate using that until i'm happy with it

@lucasmeijer Really good talk, thanks for sharing.

My talk "A love letter to pi", about what is working and not working for me in having AI help me write better software faster. (if you only care about pi zap to 13m00s)

recommended viewing. one more time, on it's own. this is probably yhe most practical talk on using coding agents i've watched to date. watch it. by @lucasmeijer it's also a great demo of pi and captures exactly why i built it. youtu.be/fdbXNWkpPMY?si…

Hey @thorstenball and @sqs now that @AmpCode has deep mode, is there still a reason to consult the oracle separately while using deep mode?

@badlogicgames You didn't take the @AmpCode offer 😉

shit. guess friendship was ended by thorsten :(

@sqs @badlogicgames Tell us more about...

@badlogicgames I made one on ampcode.com but it's not released yet and is buggy. Least helpful X reply ever.

is there something like google docs, but for markdown? i need a cloud based collaborative markdown editor please.

So let's start from this post to tell you about my journey with the LLMs and the new Redis data structure I'm implementing right now. Vector Sets were hand-coded, this time I decided to use Claude/Codex as a helpers, and guess what? The work I had to do was huge. Thread:

This is my first post on the Cursor blog: an interactive survey on the state of the art for n-gram indexes.

@almonk Do you plan to release it? Would love to play with it.

spatial ui for terminals is such a fun idea – I have a similar prototype that's Ghostty-based, and very intuitive.

Why can I not have one browser window with different profiles? I mean, only one window and profiles are organised like tab groups?

Didn’t expect @thorstenball to show up in a @diabrowser video (x.com/charliedeets/s…). Keep reading, man.

@elithrar I’m not sure. I’m running several agent and Node apps on my M2/16GB Air (hooked up to a 5k LG), and it gets so sluggish that I can’t use it for much else.

recently bought a refurb 13” Air (M4/16GB/512GB) at $849. still a good deal, but would have seriously considered the Neo. 8GB RAM is plenty for a text editor, agent, LSPs and Chrome. modern OS’ will use what you give them, which is often why it looks like you need more.

The delight of novel experiences will be shadowed by nostalgia for what we leave behind. It’s normal. Pour one out for the Old Ways, then move on. Figure out the new world. Do things. The New Way is learned through doing. There will be highs and there will be lows. Great movements are built on emotion, not always positive ones. There will be those who lean into the negative, and the world has no shortage of demagogues who will gladly trade you more cortisol for influence and power over you. You decide which path you want for yourself. A big part of that is choosing your community. For me, I want to be among those who are leaning into the delight and joy and wonder of it all. Someday, you'll tell your grandkids: I was there when the world changed. And I set my hand to the stone to build the one you’re living in now.

@mitchellh I would also like some new kind of concepts something like sidequest: x.com/taehkimmm/stat… or some interface like the 9term or Acme (Plan9) where you could go back and edit history + forking...

I’ve been forcing myself to use the GUI agent apps the past week just to learn. One week isn’t enough experience yet, but I think they all miss the mark (so far) and I’m deeply fighting the urge to do one with my own taste. I’m not gonna do it. I’m not gonna do it.

@cramforce @robipop22 @vercel @rauchg Personal and Work (don't want to have multiple GitHub Accounts)

@robipop22 @vercel @rauchg Explain your use case plz

day 100 of asking for @vercel @rauchg @cramforce to please implement multi github account linking on multiple teams

@theo I think it's time that we need something like PromptHub... Instead of Pull Request you can only submit Prompt Request.

Maria is speaking straight facts here. Please don’t spam us with giant PRs adding new features. Just tell us what you want and we’ll triage accordingly. Appreciate the hype around T3 Code but please chill a LITTLE bit guys

@sappomanno @1Password 🤔 we already have one for our team.

@mazebuhu @1Password Ask ICT for a license and get a family sub for free 🐽

So what are you all using for storing your secrets now a days? @1Password got to expensive for the whole family.

@mitchellh @AmpCode Wow thanks didn't know that!

@mazebuhu @AmpCode You can use the command palette to toggle reasoning effort

Ahhhh, Codex 5.3 (xhigh) with a vague prompt just solved a bug that I and others have been struggling to fix for over 6 months. Other reasoning levels with Codex failed, Opus 4.6 failed. Cost $4.14 and 45 minutes. Full trace plus includes original issue: ampcode.com/threads/T-019c… I know this prompt is relatively bad. Honestly, our stable release is in a week, and I was throwing some Hail Marys at the frontier models to see if I could get a clean, understandable fix for some of these bugs. By using `gh`, it grabs much better context from the issue, so its not terrible. The best thing that Codex did was eventually start reading GTK4 source code. That's where I ended up (see my GH issue), and I knew the answer was somewhere in there, but I didn't have the time or motivation to do it myself. The other models never went there, and lower reasoning efforts with 5.3 didn't go there either. Only xhigh went there. I think that was a critical difference. The final fix was decent. It was small, all in a single file, and very understandable. It had one bug I identified (you can see in the trace), and then I manually cleaned up some style. But, it did a great job. Definitely an "it's so over" moment. But at the same time, it feels amazing because now our next stable release will have this fix and I was able to spend the time working on other fixes as it went.

Resharing this ML Street Talk episode with @jeremyphoward again, because I think you should watch it. So many grounded thoughts cutting through the hype and breathlessness. Also discusses juniors in the age of agentic coding. youtu.be/dHBEQ-Ryo24?si…