AI Development Lead. Notes on LLM systems, agents and building software with AI.https://piotrludwig.space/https://piotrludwig.space/posts/software-houses-the-sdlc-factories-of-the-future/https://piotrludwig.space/posts/software-houses-the-sdlc-factories-of-the-future/Could software houses become SDLC factories, designing and overseeing AI-driven processes that turn a client’s idea into a working product?Sun, 18 Oct 2026 04:56:30 GMThttps://piotrludwig.space/posts/question-the-critics/https://piotrludwig.space/posts/question-the-critics/Before the critics’ certainty becomes our doubt, ask whether they understand what they’re judging and can see possibilities alongside limitations.Mon, 12 Oct 2026 18:39:56 GMThttps://piotrludwig.space/posts/fastball/https://piotrludwig.space/posts/fastball/I wanted to play table tennis with friends, not spend the evening managing scores. One day was enough to build and test a prototype that could help.Mon, 21 Sep 2026 10:00:00 GMThttps://piotrludwig.space/posts/ill-build-my-own-in-the-age-of-ai/https://piotrludwig.space/posts/ill-build-my-own-in-the-age-of-ai/AI makes building your own solution faster. Discovering edge cases and living with the maintenance costs still take work.Tue, 01 Sep 2026 10:00:00 GMThttps://piotrludwig.space/posts/when-code-gets-faster-requirements-become-the-bottleneck/https://piotrludwig.space/posts/when-code-gets-faster-requirements-become-the-bottleneck/Faster implementation shifts the bottleneck to understanding problems and capturing requirements that people and AI agents can act on.Sun, 23 Aug 2026 10:00:00 GMThttps://piotrludwig.space/posts/nural-study/https://piotrludwig.space/posts/nural-study/I’m building Nural.Study to connect saved materials, notes, flashcards, study sessions and reviews into one learning system. The project entered beta this year.Sat, 22 Aug 2026 10:00:00 GMThttps://piotrludwig.space/posts/ai-agent-evaluation-the-unwanted-foundation-of-development/https://piotrludwig.space/posts/ai-agent-evaluation-the-unwanted-foundation-of-development/Systematic evaluation makes AI agent development a measurable process, revealing consistency, edge cases, and opportunities for improvement.Thu, 09 Oct 2025 10:00:00 GMThttps://piotrludwig.space/posts/hallucinations-in-isolation/https://piotrludwig.space/posts/hallucinations-in-isolation/Language models hallucinate for the same reason a mind in sensory deprivation does: they have lost their point of reference. Notes from twenty meditation retreats, applied to LLMs.Tue, 17 Jun 2025 11:49:00 GMThttps://piotrludwig.space/posts/improving-product-comparison-with-llms/https://piotrludwig.space/posts/improving-product-comparison-with-llms/Two indexing strategies for a product discovery assistant, tested side by side: 500-character chunks versus granular per-parameter indexing. What actually improved the answers, and what turned out to be needless complexity.Tue, 25 Mar 2025 09:00:00 GMT