marcink

marcink@stolat.town

I build software some of it even works (these days mostly <a href="https://stolat.town/tags/golang" class="mention hashtag" rel="tag">#golang</a> infra for <a href="https://stolat.town/tags/python" class="mention hashtag" rel="tag">#python</a> apps on top of <a href="https://stolat.town/tags/kubernetes" class="mention hashtag" rel="tag">#kubernetes</a>) he/him, polyam pl/en/fr, originally 🇵🇱, nowadays 🇫🇷 And I must not forget, we must not forget, that we are human beings. (Ren) <a href="https://stolat.town/tags/nobots" class="mention hashtag" rel="tag">#nobots</a> Fuck genAI.

Posts

Post #3767876
I bought this watch as my first smart watch as soon as this model came out in 2015. It has seen a lot of places since, done fairly serious kilometers. A couple of years ago I dropped it and it fell flat on its face, chipping off part of it and leaving a couple of cracks, I fixed it with a dab of uv resin. A year later I dropped it again, had to use more resin and start using screen protectors to protect me from the screen (a sharp edge here, a shard there). It still works just fine.
Post #1958594
The more I interact with people who use LLMs to work on codebases which I know and they don’t the more I empathize with doctors facing patients who use the Internet for the diagnosis and treatment ideas.
Post #1958593
The Vercel/context.ai security issue: I am more than a little peeved that people tend to point to AI only where it sounds useful. “The attacker must have used AI to breach us so badly and quickly” sure sure, could someone ask about the software delivery process at context.ai? To what degree it depends on LLMs, what is their review process, how do they manage risk?
Post #1958592
Living in France means a lot of “this phone conversation could have been an email”.
Post #1958591
Happy root exploit day to all who celebrate!
Post #1958590
The trouble with AI is that eventually you run out of other people’s data.