Miloš Švaňa
ai, python, software engineering, decision-making and more
- Posts
- 28 from 19 Jan 2025 to 15 Oct 2026
- Posting
- monthly (9 in the last year)
- Kind
- personal blog
- Feed
- https://svana.name/feed/
2026 9
- My slow return to religion
- My Arduino RGB controller is now even better
- I rewrote the Embeddings Playground backend in Rust
- Reading Club: Build a Large Language Model (from Scratch)
- I built an RGB controller with Arduino
- Do LLMs hallucinate more in Czech than in English?
- Benchmarking LLMs for hallucinations
- Reading Club: Against the Machine
- Hallucinations in LLMs: What are they and why we have them
2025 19
- Moving local GPU workflows to the cloud with Rsync
- Implementing a fully local AI coding agent is hard
- Introducing FileChat: A local read-only AI coding assistant
- How I Solved PyTorch’s Cross-Platform Nightmare
- Reading Club: Lessons from AI Safety for Businesses
- I trained a GPT-1-like model from scratch
- Reading club: Concrete Problems in AI Safety
- Why we shoud take AI safety seriously
- Hope, Hype, and Hell: Mapping the AI Safety Spectrum
- Criteria in multiple criteria decision analysis
- Even Swaps: making decisions via trade-offs
- Better decisions with Multiple Criteria Decision Analysis
- On the value of commercial coding courses
- Keras, PyTorch, and an integrated Intel GPU
- “How will I use this in real life?”
- Introducing Embeddings playground
- Why use mathematics to make decisions
- Would I do a PhD again?
- The art of testing LLM-powered apps