Live · the stream

Thoughts

One-liners, links, images, project updates. Native posts live here first; public posts from elsewhere are pulled in with their source kept.
2011Publishing history, charted as a night's sleep · 2026

Everything

1–27 of 27

  1. I named my LLM memory system Knowfer, after Gopher. Wanted to dust off an old protocol and give it a second life. So I reused the old protocol. Low and behold it works like a charm.

  2. The AI demo ends when the model says it worked. The product starts there.

    Agents

  3. Autonomy and authority aren’t the same thing. Agents can do the work. Restarting the system should still take my signature.

    AgentsAutomation

  4. Match the model to the stakes. A small coding model is fine for light work and the wrong call for anything critical.

    Agents

  5. When a model swings from “this is nothing” to “build everything around this” in one reply, call it out and get a second opinion.

    Agents

  6. A good test for any app: what would make it a no-question $25 one-time purchase?

    macOS

  7. Ask what standard practice is. Then ask what the gold standard is, and build toward that one.

    Automation

  8. Don’t let the main menu turn into a junk drawer. Anything that keeps piling up gets its own menu.

    macOS

  9. Name things so they don’t collide with words people already know. “Package” already means something on Linux.

    Infrastructure

  10. If an agent answers with a list of things for me to do, it has the roles backwards. I’m not the worker here. It is.

    Agents

  11. Separate the agents that decide policy from the agents that carry it out. Church and state.

    Agents

  12. If the text already exists, read it. Don’t record the microphone to transcribe what the closed captions are already saying.

    AutomationmacOS

  13. Let the user pick light or dark. Player controls that change color with whatever is on screen are just annoying.

    macOS

  14. Not every decision needs a frontier model. Let a cheap model judge whether the request needs an expensive one.

    Agents

  15. If your agent’s only memory is the chat window, it doesn’t have a memory.

    Agents

  16. I don’t want an AI that does everything. I want one that tells me exactly what it did.

    AgentsAutomation

  17. A small model that knows its job beats a big model guessing at yours.

    Agents

  18. Run a model on your own machine once and you’ll understand what you’ve been trading for convenience.

    Agents

  19. Prompt engineering is a job title for a problem that should be solved in code.

    Agents

  20. Everyone’s asking what ChatGPT knows. The better question is what it will do when nobody’s checking.

    Agents

  21. Pediatric studies are the hardest work in sleep medicine and the most rewarding. Kids don’t fake sleep, and they don’t fake being uncomfortable either.

    Sleep Medicine

  22. The patient trusts the device, and the device maker never meets the patient. When something goes wrong, the technologist is the one in the room explaining it.

    Sleep Medicine

  23. NFTs are the new box of samples. Same pitch, new chain.

    Commentary

  24. Home sleep tests will keep growing, and they should. But a home test can’t watch the patient. Some studies still need a technologist in the room.

    Sleep Medicine

  25. Remote work isn’t new. We ran EnLab as a distributed team in 2015. What’s new is that managers can’t hide behind the office anymore.

    Entrepreneurship

  26. quote

    I always take time to learn from my circumstances. Life has seasons. I’ve learned to embrace the rain.

    Writing

  27. quote

    There are way too many articles selling a lifestyle that doesn’t exist for everyone in this startup community.

    Entrepreneurship

Machines

json · txt · md