Silicon Valley models beat everyone by a mile... until they don't. Very interesting news have arrived from China this week that might impact the future of AI and AI companies in the West. Kimi K3 weights, scheduled to release tomorrow, may disrupt the Western AI market - the market players are acting as if it had.
For any AI company out there, using their own AI platform more than any of their clients has always been mandatory. Anthropic understood it, and took the bet with Claude Code. In March 2026 the source code leaked, and it was mocked. Three months later we know, that this was the right choice.
This year, I played 'Advent of Code Golf': a Code Golf over Advent of Code: calendar of coding puzzles by Eric Wastl. Miraculously, I have managed to fit all my solutions in 2024 symbols (a perfect number for the 2025 calendar!). But as soon as I have written the part 1, I became nagged by my results. 'All solutions' weren't really all solutions. That was worth improving...
Anthropic, the company behind the most powerful (at the moment) AI software development agent Claude Code, recently had to retire the take-home job interview assignment. The reason? They 'were about to release a model where the best strategy on take-home would be delegating to Claude Code'. According to Anthropic benchmarks, their Opus 4.5 model could beat in 2 hours what a human can make in 2 hours; and significantly beat most humans results if given 12 hours to work. Is it an incredible feat, or is it just a hype?
Every year, I play Advent of Code, an advent calendar of coding puzzles by Eric Wastl. This year I wanted an extra challenge. I decided to spend hours writing horrible, unreadable, bodged-together code full of code smells, bad practices, and hacks. I decided to do Code Golf.
I have beefs with many conventional critiques of python. People complain about GIL, people say python is slow. Along these lines lies the 'match-case' pattern matching. I'll be clear: match-case is bad, don't use it.
You are a developer, and your team lead sucks at their job. If only you were a team lead, you would have done way better. Rejoice! For the time has come for you to prove yourself!...
There are two reasons why python is 'bad'. First, it is not statically typed like C. Second, it is not as fast as C...
We live in a world where “I need 2 weeks to do it” means “I can do it in 1 hour two weeks later”, and we accept it. ...
We can't productively do creative work for more than 4 hours a day. In 2014, I graduated from university and got a job at a biotech company, where I was doing data science and developing data pipelines for large biological data sequencing, specifically). There I learned a lot about the way I personally work...
Smartphones have stolen from us the time and the capacity to become BORED. And that’s a bad thing...
I started journaling in January 2022, and it is easily the best habit I have ever made. And the most surprising one. You should try it too! 😃...
The more I date an author, the more I realize how similar are founding a startup and writing a piece of fiction: 1. You have a wild idea of how to change the world. 2. Despite it being wild, many people had tried something similar before, unsuccessfully...
Raytracing in python - now with animation!
I read majestic Fabien Sanglard’s article about the “business card raytracer” a few years ago. These 1337 bytes of code would fit on a business card...
How to tell if your potential employer or business partner is great before signing a contract
The idea of living in a computer simulation is nice, but we are obviously just a lab experiment of 11D aliens 👽 Evidence...
No spoilers here ;-)
