Rémy

Especially given they could be charging said cars and thus circumvent the need for batteries in the infrastructure altogether (make the kWh much cheaper when the panels are working)

You could also charge the car at work for free and bring back electricity at home as a perk

That's a very interesting idea. LLMs are utter shit at new stuff.

Very disappointed to see that releases are chronological

"Pharaoh with two eagles... Fucking idiots"

must not be a very good president

As a telecommunications engineer, I get why they wrote what they wrote but it's so incredibly inaccurate...

Also Switzerland doesn't have trains that go above 30 km/h so that kind of explains the situation

Honestly I think the core issue is that nobody gives a fuck

How would you even know? Each line is a blend isn't it?

I think however that issues are mostly due to agents, which are super fucking slow on top of making mistakes.

AI as a replacement for StackOverflow and/or as a teacher is a good use-case

Although more often than not I'll just stop the AI and code the thing myself because the LLM outputs crap

Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity We conduct a randomized controlled trial to understand how early-2025 AI tools affect the productivity of experienced open-source developers working on their own repositories. Surprisingly, we find th...

First serious study I find on the productivity gains for developers using AI.

Turns out the gains are significantly negative. 🫠

These results are absolutely not surprising given my experience. And it's with seniors, with juniors it ought to be explosive.

What is X?

Not necessarily 👀

I mean. It entirely fell over. Everything stopped "non core" working overnight, and now it's just a bot wasteland.

Documentation Archaeology: How to Extract Knowledge from Abandoned Codebases with AI Developers are no strangers to being thrown in a wild project where their ability to understand technical debt will be key to their survival and mental health. LLMs are amazing tools for this use-case

Did you try piping the whole codebase into a LLM and asking it to do the refactor? It's potentially very efficient at that.

I just explained that in an article (not exactly the same topic but the technique works basically the same)

Hahaha. No it's my Slack notification settings (which are entirely muted)

Let's see if people from work can figure this one out

It's not exactly tight for OpenAI. They lost all B2B market shares while 3/4 of the market is split between Google, Anthropic and DeepSeek. B2C holds with brand name and UX. GPT-5 better be making coffee because Gemini does everything else 😇

I'd wager that porn only provides the ideas. But the attention economy is the root cause.

Nothing as fulfilling as the most vanilla sex in the right context 🧐

Yet like all the good healthy habits it's not appealing. We are way too disconnected from our bodies. 🫠

(he says, on social media)

Not sure if there is only porn to blame though.

Some kinky people genuinely are kinky just like you could be gay.

But how many people are kinky for the dopamine hit (hi ADHD people) and regret afterwards?

Younger generations are running out of dopamine with every corporation sucking it dry

Ça doit être un driver spécifique, Fedora est supposée être assez à jour donc à mon avis ça se joue à un tout petit truc

It's been on a steady decline for many years. Every time I contributed to the platform my question got rewritten and my answer got censored. Personally I now hate this platform and given there is a convenient alternative it's all for the best IMHO

Documentation Archaeology: How to Extract Knowledge from Abandoned Codebases with AI Developers are no strangers to being thrown in a wild project where their ability to understand technical debt will be key to their survival and mental health. LLMs are amazing tools for this use-case

Had to dig in legacy projects yesterday and everyone got amazed by the documentation I was able to produce. A pretty useful case for LLMs I must say!

Please, I already can't solve captchas anymore

Let's mention however that actually new technologies perform like utter shit in LLMs, so until there is enough open source projects using said tech to give a good corpus in GitHub, you also won't be getting any help

a close up of an older man 's face with the words `` ironic '' written on his face . ALT: a close up of an older man 's face with the words `` ironic '' written on his face .

StackOverflow is so dead that even if AI doesn't bring any productivity, eventually there won't be relevant content for newer technologies left on StackOverflow and you will have to use AI if you need help

It's pretty neat that the `> ''` condition in SQL matches a true-ish string in Python, even if it is nullable

Although what I really want is 10 million tokens context window, that would be fucking great (my largest code bases are around 2 million tokens, if I can pipe it all there is no need for agents anymore)

Any KPI becomes a bad KPI when you start to follow it 😭

Although I'd offer the cyclomatic complexity as a good approximation (provided that it understands when the system gets gamed, especially in FP-leaning languages such as JS)

I'm starting to understand that LLMs are passing coding benchmarks not because they make smarter code but because they can consider more "branches" at the same time to fix their own mistakes, writing literally 5~10x too much code. I seriously doubt that those are useful improvements 🙈

I'll wait to see if it bites me in the ass, but so far Gemini has proven quite good at converting the whole codebase into a nice documentation, even if the project in itself is initially documented like shit/not documented

Inkjet printers are a mystery to me, they have so many drawbacks in comparison to laser and yet people keep on buying them? (admitedly nobody has needed a color printer in more than 10 years)