Matthias Leyendecker

❯ terminal velocity_

Welcome to the perfect world of micro-blogging without the ad-driven algorithm. Short form notes, mostly with photos.

– AI will kill us all

What amazes me is not that yet again an AI researcher leaves one of the big labs to warn on national television that “AI will kill us all”, or to disappear into the mountains to write poetry. It’s the completely uncritical, unreflective platforming that the media gives these stories.

Perfect career pivot: spend barely three years in the industry, leave with a doom-and-gloom post on X, watch it go viral, take some softball questions from Anderson Cooper, and profit. To be fair, he walked away from his Anthropic equity (by his own words). We can only assume that his OpenAI stake keeps climbing with every new hype cycle.

And then, by pure coincidence of course, just a few days later Amodei publishes “We Must Pace the Frontier”. Within hours Altman, Musk and everyone else agrees. There’s a word for competitors (public and soon-to-be public) agreeing among themselves to restrict output.

Once again the news outlets repeat the tweets word for word, without pausing for one minute on the “why”, as if there were no deeper intent behind them.

– Too far removed from product

I had an amusing experience in a job interview not too long ago. It must have been stage 2 or 3, where my product background was evaluated, and eventually I got a no (official reason: the company was second-guessing their own role profile in general and wanted to reboot the hiring process), but as feedback I received a “you seemed slightly too far removed from product in your previous role”.

Reason being that I coached my product and engineering teams to the point where they ran almost completely autonomously without too much intervention from my side apart from strategic goalposts. So I had the capacity to focus on a lot of other topics surrounding the business, from funding to compliance risk – quite natural as a co-founder.

And here is silly me thinking that my team(s) are the actual product that I need to shape and improve, and that having them run like a well-oiled machine as an end goal just proves I am missing the mark.

I must say, I got a lot of fun feedback in the past decade, but this one will definitely enter the hall of fame.

– Building Pi with Pi

Really insightful piece by Armin on how AI is quietly eroding open source. Not just through the sheer volume of slop flooding GitHub, but by letting us skip the human collaboration that makes projects strong in the first place. Open source was never about the code – it was about the community around it.

Building Pi with Pi →