Omid Sayfun
I'm a senior software engineer in Amsterdam, building software with AI agents and writing about what I learn.
I have spent most of my career across FinTech and e-commerce. These days most of my energy goes into how AI changes engineering work: what agents are good at, where they fall over in production, and what that means for how teams ship.
Outside of work, I’m typically at the gym, coding on a side project, or playing padel.
Notes from building with AI agents, plus whatever else teaches me something. The latest five:
- Vector search relevance: chunking, metadata, and the 0.81 problem08-11-2026 · Most bad vector search results come from one of three failure modes: chunking, modality mismatch, or a confused model. Each one has a specific diagnostic and a specific fix.
- pgvector setup: your first multimodal query in TypeScript08-03-2026 · One Postgres table can hold text and screenshot embeddings in the same vector column. This post sets up the schema, the Voyage embedding call, and the first query that returns both.
- Vector search in Postgres: the mental model behind pgvector08-02-2026 · A first similarity query that returns results is not a finished search feature. This post explains what an embedding is, why one Postgres column can hold text and images, and where an untuned index starts returning wrong results.
- AI agent architecture patterns that survive production05-18-2026 · Most agent demos are a model in a loop. Production agents need a run store, typed memory layers, idempotent tools, a trace log, and approval gates. Here's the full architecture with Postgres schema and TypeScript patterns.
- Stop adding prompts. Your agent needs control flow.05-17-2026 · A while-loop calling an LLM keeps doing the wrong thing, and more prompt engineering isn't the fix. Here's the control-flow ladder for AI agent orchestration: bounded loop, workflow, state machine, with TypeScript examples at each level.
Flaky E2E tests have been part of my day-to-day for a long time. A simple UI change or timeout could easily break my flow and steal focus time. So here is my new project:
- Smoketest - AI-assisted E2E testing. The tests your agents don't write.
I use an extensive list of tools and software in my daily life. I can't live without some of them (painkillers) and some of them are just nice to have (vitamins). Here is a list of my current stack:
- Raycast - Give it a shot and you will never want to go back to macOS spotlight.
- Bartender - I have a million menu bar icons and this is the only way to manage them.
- cmux - This is the best terminal replacement and fully integrated with most of coding agents.
- OpenWhispr - You speak 3x faster than you type. Who knew?
- BetterDisplay - Best tool to manage your external monitors on Mac.
- Telegram - I use Telegram for everything. It is the app that I use first thing in the morning.
- Beekeeper Studio - The best SQL editor I have used.
- Visual Studio Code - I migrated back from Cursor and use VS code as a git diff review / text editor.
- ChatGPT Codex - My daily driver for agentic coding. I hit the rate limits most weeks.
- OrbStack - The best way to run Docker containers on macOS.
- MacBook Pro , AirPods Pro , iPhone - I'm a heavy Apple user.
I try to code and workout every day. Here you can find my activity in the last 6 months:
I have been fortunate to have traveled to many countries. My goal is to visit 30 countries by the end of 2030.
I subscribe to a number of different newsletters and youtube channels. I try to read anything that teaches me something new. Most of them end up in my AI agent inbox and I just read the summaries. Here is a list of my favorite ones that I fully read and recommend:
- Naval - I have followed Naval for a long time. He talks about startups, investing, and life in a philosophical way.
- Collaborative Fund - Reading Pure Independence by Morgan Housel hooked me immediately. They write about economics, markets, and investing with great storytelling.
- The Diary of a CEO - Steven Bartlett interviews successful people from different fields about their journey, mindset, and life.
- Farnam Street - Introduced by Naval and I have been a fan ever since. They write about mental models, decision making, and learning applicable to real life.
- Mark Manson - I got bored in the middle of his most famous book, then found his newsletter and liked it a lot more. He writes about self improvement, philosophy, and life in an honest way.
The best way to reach me is email at [email protected]. I read everything and reply to most.
If your team is shipping AI-written code and figuring out what that does to reviews and testing, I would especially like to hear from you.