Manuel Odendahl (@programwithai) 's Twitter Profile
Manuel Odendahl

@programwithai

Real-world programming with AI. No filler, just killer! Unique advice and creative techniques that you won't find anywhere else, straight from the trenches.

ID: 1723855763297247233

calendar_today13-11-2023 00:10:33

744 Tweet

369 Followers

228 Following

Richard Artoul (@richardartoul) 's Twitter Profile Photo

This is a great summary of the history of Go’s garbage collector, and a small glimpse into why Go became so prevalent in the distributed systems space so quickly

Pierre Vannier (@pierre_vannier) 's Twitter Profile Photo

Mes amis c’est le moment de faire pêter les likes, les partages, les étoiles et les avis pour faire monter “IA pas que la Data” au firmament des podcasts 🫶❤️

Mes amis c’est le moment de faire pêter les likes, les partages, les étoiles et les avis pour faire monter “IA pas que la Data” au firmament des podcasts 🫶❤️
Manuel Odendahl (@programwithai) 's Twitter Profile Photo

IMO, Language has never been a direct reflection of one person’s thoughts and ideas. When you read an article, the words are shaped by the culture of that language (English vs French vs …), the culture the author was brought up in, the way they were taught to write an article,

Manuel Odendahl (@programwithai) 's Twitter Profile Photo

A lot of developers struggle with the flexibility and more holistic product view that llms require to really shine (think prototyping, building internal tooling, etc…), and are instead experts at tasks that llms/agents kind of suck at: being very good at building and verifying

Manuel Odendahl (@programwithai) 's Twitter Profile Photo

I am politically quite on the outer left side (even for European standards), and yet I fully embrace AI for my work as a developer: what gives? The anti-AI crowd is IMO stuck in the narrative pushed by the top: that AI can replace work, and that workers who don’t use it will be

Manuel Odendahl (@programwithai) 's Twitter Profile Photo

Since I’m on a little rant roll, I think anti-AI for coding / writing documents (I can’t really vouch for anything else) often pushes LLMs as “stochastic parrots” in a Derogatory fashion, as well as pointing the fact that people can easily project consciousness onto a model.

Manuel Odendahl (@programwithai) 's Twitter Profile Photo

Ok the ac is down, I had a long day, I think I’m gonna punt this one because it’s already painful when I’m reasonably rested 🤣

Manuel Odendahl (@programwithai) 's Twitter Profile Photo

if you ever make a website that can be easily navigated and does not look like the 90ies, you immediately and irrevocably lose all ham radio privileges. i didn't make the rules.

Manuel Odendahl (@programwithai) 's Twitter Profile Photo

Had ManusAI build me a whole set of open modelica simulations and python analysis scripts. A bit rocky at the start (will learn how to prompt it better / provide better docs), but nifty! The interesting part is that while there is a decent chance this is nonsense, figuring

Had <a href="/ManusAI_HQ/">ManusAI</a> build me a whole set of open modelica simulations and python analysis scripts. A bit rocky at the start (will learn how to prompt it better / provide better docs), but nifty!

The interesting part is that while there is a decent chance this is nonsense, figuring
Manuel Odendahl (@programwithai) 's Twitter Profile Photo

The fact that vibecoded tools accumulate tech debt so quickly is partly because of the models and agents, for sure. But part of it is also the accelerated pace of development that make a product accumulate 20 features within a few days, the same way a larger codebase will accrue