← All topics

Agents & Tooling

Agents, MCP, and wiring AI into the systems you already run.

Showing 1–20 of 144 posts

  1. Skills, then agents, then context

    Google is turning Gems into skills. If a big AI job overwhelms your team, break it into simple steps, write each as a skill, and put an agent in charge of running them.

    4 min read Interactive

  2. Did you or did you not?

    Decision models are getting cheaper, so more people will be handed one. They pick from allowed answers but still give probabilities and can change between runs, so send the unsure cases to a person.

    3 min read Interactive

  3. Steam saved a 60 fps cap for 112 games. The Performance panel still showed 144

    An AI agent changed a setting on my Legion Go, and Steam's config file showed it saved. A screenshot of the settings panel showed it hadn't taken effect. Anything new or recently adjusted needs oversight.

    2 min read

  4. Give a shared AI assistant a weekly checkup, and keep what it adds terse

    Once a week, ask the assistant your team shares what would improve its instructions. Keep any rules it adds succinct, and give the bot two human owners.

    2 min read

  5. Plan big, do small

    A big model should plan the work and find the tools, libraries and free databases that already exist. Smaller models, or programs on your own computer, do what I like to call the donkey work.

    5 min read

  6. School family knowledge base, part 2

    Part 2 of a build log about a school family knowledge base: what information each of the eight sources holds, and the format it arrives in.

    3 min read

  7. School family knowledge base, part 1

    Schools send us information in eight places. That's a lot of channels for parents to coordinate and understand. Part 1 of a build log: the main issue, and how the tasks have to be broken down.

    2 min read

  8. Logo.dev Returns a Company's Logo From Its Web Address, and Its MCP Server Lets Claude Fetch It

    Logo.dev returns a company's logo, colours and brand details from its web address. Here is what it does, how to connect it to Claude, and what to use it for.

    3 min read

  9. When Gemini Runs Claude, What Are You Buying?

    Google’s workplace agent can use models from more than one provider. I want to know what happens to my instructions, files and work history when I switch.

    3 min read

  10. An Agent Needs a Stopping Point

    A failed connection or a missing file should have a defined outcome. My inbox cleanup and Anthropic’s latest report offer a practical way to specify it.

    2 min read

  11. The 3,848-Message Sieve

    What It Takes to Let an Agent Clean Your Inbox

    3 min read

  12. The Infinite Sandwich and the Made Bed

    Why Systems Thinking Is Your Real AI Superpower

    5 min read

  13. Gems Are Becoming Skills, but I Can Find My Gems, Not My Skills

    I am using Gemini more and more for editorial and creative work. I like it for developing ideas and roughing out illustrations, while I tend to use other tools for purely operational jobs. That has meant spending quite a bit of time playing with Gems.

    2 min read

  14. OpenClaw versus Paperclip

    I tried OpenClaw to coordinate agents from a Rabbit R1 and it turned into engineering for engineering's sake. I went back to Paperclip, and it is getting a lot better.

    3 min read

  15. The difference between skills and context

    A skill is the reusable instructions for doing a job. Context is the material that goes into this particular job. I use my one-panel cartoon Gem as the example.

    1 min read

  16. The Importance of Being Jev

    TypeSafe's Jev and Cloudflare's Clef return a yes or a no, a one or a zero. They are for questions with a fixed set of answers, such as whether an email is urgent.

    2 min read

  17. Stupid Gemini Tricks: One Panel Comics

    A quick and dirty Gemini Gem demo on creating one-panel comics.

    1 min read

  18. Meta's Muse asks for a lot of trust

    Meta's Muse agent connects to your email, calendar, payments and health services, trains on your conversations by default, and has already shared a user's home address with a Marketplace buyer.

    2 min read

  19. They're the same video

    OpenAI's new 'Get things done with voice mode' video is oddly similar to this SNL parody.

    1 min read

  20. Dumb Little Robots, All the Way Down

    A team gave one agent the whole job of pulling in data, analysing it and reporting on it. Its context window filled up and it lost track of its purpose.

    1 min read