AI builder and product owner writing on AI and Technologyhttps://janwillemaltink.com/en-ushttps://janwillemaltink.com/writings/making-claude-code-continuously-improve-itself/https://janwillemaltink.com/writings/making-claude-code-continuously-improve-itself/A continuous improvement system for Claude Code that extracts insights from coding sessions and turns them into concrete harness improvements.Wed, 25 Feb 2026 00:00:00 GMThttps://janwillemaltink.com/writings/ai-as-a-compiler-for-my-notes/https://janwillemaltink.com/writings/ai-as-a-compiler-for-my-notes/After years of experimenting with note-taking systems, AI has finally enabled me to create one that actually works for me.Sun, 08 Feb 2026 00:00:00 GMThttps://janwillemaltink.com/writings/n8n-sticker-generator-with-gemini-and-supabase/https://janwillemaltink.com/writings/n8n-sticker-generator-with-gemini-and-supabase/An n8n workflow that generates stickers with Gemini, removes backgrounds using Python task runners, and stores results in Supabase.Sun, 25 Jan 2026 00:00:00 GMThttps://janwillemaltink.com/writings/my-agent-rules/https://janwillemaltink.com/writings/my-agent-rules/Guidelines for structuring and operating codebases to maximize safe autonomy, high-quality output, and efficient use of agent context.Wed, 21 Jan 2026 00:00:00 GMThttps://janwillemaltink.com/writings/simple-loops-beat-multi-agent-architecture/https://janwillemaltink.com/writings/simple-loops-beat-multi-agent-architecture/Why a bash while loop often outperforms elaborate multi-agent systems. Research shows single agents with good validation beat coordination overhead.Thu, 08 Jan 2026 00:00:00 GMThttps://janwillemaltink.com/writings/good-validation-enables-good-agents/https://janwillemaltink.com/writings/good-validation-enables-good-agents/A walkthrough on building a unified validation system with gates and transformers that works across Claude Code, Cursor, and git hooks.Mon, 05 Jan 2026 00:00:00 GMThttps://janwillemaltink.com/writings/my-principles-for-successful-coding-agents/https://janwillemaltink.com/writings/my-principles-for-successful-coding-agents/A mental model for working with coding agents: Rules, Scopes, Validators, and Gates (RSVG). How to use deterministic checkpoints to guide probabilistic agents.Fri, 26 Dec 2025 00:00:00 GMThttps://janwillemaltink.com/writings/new-macbook-guide/https://janwillemaltink.com/writings/new-macbook-guide/My personal migration guide for setting up a fresh MacBook with 1Password, Homebrew, Zsh, and a developer-focused toolchain.Fri, 19 Dec 2025 00:00:00 GMThttps://janwillemaltink.com/writings/we-are-learning-to-speak-machine/https://janwillemaltink.com/writings/we-are-learning-to-speak-machine/The story of AI has been about machines understanding us. But something else is happening: we are changing how we write.Fri, 12 Dec 2025 00:00:00 GMThttps://janwillemaltink.com/writings/what-is-lora/https://janwillemaltink.com/writings/what-is-lora/LoRA lets you fine-tune large language models by training only small low-rank matrices instead of billions of parameters. Here is how it works.Wed, 10 Jan 2024 00:00:00 GMThttps://janwillemaltink.com/writings/what-is-byte-pair-encoding/https://janwillemaltink.com/writings/what-is-byte-pair-encoding/LLMs do not read words — they read tokens. Here is how tokenization works and how the BPE algorithm builds a vocabulary that balances efficiency with coverage.Wed, 20 Dec 2023 00:00:00 GMT