Bad error handling has cost billions of dollars, crashed planes, killed people, tanked stock markets, wrecked vehicles, and delayed flights. As we reflect on Halloween, it’s fitting to consider these horror stories of software gone wrong.
Movies are handmade, and just like any other art form, sometimes the seams that hold movies together become visible to the audience. For mov...
I’m always frustrated by heavy processes. They usually feel like unneeded overhead to get the job done. Let’s explore that overhead world, as some is actually desirable.
Exploring trends in entropy across afrobeats, pop, and hip-hop.
A few weeks ago, I participated in a startup competition in Italy. Over a single weekend, my team had to develop a startup idea, validate it, and pitch it against other teams. In this article, I’ll share our experience and the key lessons that led us to victor…
Addendum: make computers suffer. ◼...
Website of Alex Zakhlestin
understand + work backwards from the root goal • don’t rely too much on permission or encouragement • make success inevitable • find your angle • think real hard • reflect on your thinking
Part 6 of my software journey, my depth phase.
Chris Dixon's blog.
Using technology to answer important questions like: Who would win in a fight, X or Y.
April 09, 2025 · 18 mins · 3350 words
Explore Gemma 3 models now offering state-of-the-art AI performance on consumer GPUs with new int4 quantized versions optimized with Quantization Aware Training (QAT).
Turns Codebase into Easy Tutorial with AI. Contribute to The-Pocket/Tutorial-Codebase-Knowledge development by creating an account on GitHub.
I guess you have all heard about the growing problem of AI companies trying to aggressively collect whatever data they can get their hands on to train their models. This has caused an explosive surge in web crawlers relentlessly hitting servers big and small. …
Batching speeds up training and inference, but for LLMs we can't just use matrices for it -- we need higher-order tensors.
It’s election season here in Iceland! The election is on Saturday, 30th of November, so next Saturday from when this is written. Every time elections are upcoming, somebody inevitably asks me how the voting system here works, probably expecting a simple answer…
OpenAI has recently introduced its new o3 and o4-mini AI models that are integrated into ChatGPT. Reports indicate these models not only hallucinate more often than previous versions but also demonstrate an uncanny ability to determine photo locations. This du…
Spoiler: you probably shouldn’t. A personal reflection on playful math, shifting passions, and nurturing curiosity.
2025-04-18
MCP vs ACP: AI Protocols for Context and Agent Orchestration
A review of Ansible and its philosophy's merits and shortcomings
My smart bike company went bankrupt, so I added a physical button to turn the light on/off
2025-04-19
While working on LLM-driven NPCs, I observed significant improvements in several areas by adding a simple component: in-memory free-text search
Rounding to 0% is bad because the user may think there's been no progress. Rounding to 100% is bad because the user may think things are done when they aren't.
Imagine walking into a shoe store and asking for size 10 sneakers. You may end up with shoes that fit, but “size 10” isn’t descriptive enough to be sure — the US, UK, EU, Australia, and Japan all have different shoe sizing systems. To ensure you get shoes that…
Contribute to PiLiDAR/PiLiDAR development by creating an account on GitHub.
The latest Google update will make your phone more secure if you don’t touch it…