Benjamin Latger

CTO & builder · Shipping products & sharing the process

⭐️ $0/m
🛠️ better-cycle
🛠️ avolia
🛠️ focco
🪦 vellify
🪦 getnps
🪦 memoriae

[Benjamin Latger @benlatger]

82 years. 52 weeks per year. ~4200 boxes.

Your whole life on one screen.

It hits different when you start counting the ones already behind you.

https://t.co/Y6g1WcT133

[Benjamin Latger @benlatger]

@thsottiaux thanks a lot ! is it possible to see the reset's expiration inside codex please ?

[Benjamin Latger @benlatger]

@justbyte_ You should ask « and why? »

[Benjamin Latger @benlatger]

@petergyang Add new language support for your app using something like next-intl (I did this in a little 4-hour loop this week).

[Benjamin Latger @benlatger]

@thsottiaux Could you add an option to get insights into our prompts and flows to spotlight areas for improvement? It would be good to have a side agent that can review our flows and give advice.

[Benjamin Latger @benlatger]

@thsottiaux I'm used to logging myself into multiple Linear accounts, and it's a little difficult to handle. You have to switch multiple times.

[Benjamin Latger @benlatger]

@Simk933 J’ai aussi testé et c’est effectivement très fort ! Surtout le fait de pouvoir connecter par MCP à un agent externe :)

Hâte de voir la suite

[Marcos Rico Peng @Marcos12345rico]

today we're launching @Palmier_io, a video editor Claude can edit.

use AI to edit, organize, and generate footage directly in the timeline.

finally, a video editor built for AI.

open-source. mac native. available now.

[Benjamin Latger @benlatger]

@T_Zahil I've been thinking about this more and more, honestly!

[Benjamin Latger @benlatger]

@cinamarina Good job! Huge milestone 💪

[Benjamin Latger @benlatger]

@T_Zahil Good idea, and it's clearly something you build over time. But it takes a while, and honestly staying consistent long-term is no easy feat

[Benjamin Latger @benlatger]

@T_Zahil Looks like the link isn't accessible on my end Thomas !

[Benjamin Latger @benlatger]

Must read

thanks @addyosmani for your articles

[Addy Osmani @addyosmani]

https://t.co/BkYXbsx92x

[Benjamin Latger @benlatger]

@T_Zahil Haha, don't worry you're not stupid ! 😆

I think a lot of us (builders) work that way

[Benjamin Latger @benlatger]

Thanks a lot for the reply. I get the value of a separate file to track progress, but what I'm really struggling with is how to actually constrain the LLM to follow very specific directions over long sessions.

Even with a progress file, I often end up after hours of work realizing the output has slowly drifted, it took shortcuts, ignored some architectural constraints, or the quality/style became inconsistent. Then I have to spend a lot of time fixing it.

How do you prevent that? What concrete techniques do you use to keep the model tightly aligned with what you want (specific rules, output style, quality bar, etc.) across many iterations?

Maybe I'm missing the point here.

[Benjamin Latger @benlatger]

@daicandev Building a retrospective tool for teams working with Linear. Coming soon!

[Benjamin Latger @benlatger]

@Taniyatweets_ Honestly, I don't think I could ever go back to Windows for work

[Benjamin Latger @benlatger]

@cinamarina Thanks you

[Benjamin Latger @benlatger]

Honestly, learning to code 100% by hand, no. But understanding how everything works and what agents actually write when you prompt them, yes, absolutely. I've been developing every day for almost 10 years now, and when you have an agent that can churn out 1000 lines of code in a few minutes, you better be able to read whether it's good or not. Otherwise it's really easy to end up with a completely unmaintainable codebase

[Benjamin Latger @benlatger]

@MatthewBerman Makes sense.

[Benjamin Latger @benlatger]

@theo I recently read that OpenAI is likely considering a significant reduction in token costs. Maybe this is part of that move.