Posts about Technology

When I was a kid, I had access to a workshop full of scraps and could build all sorts of gadgets.

When I got a computer, I had a BASIC interpreter and the manual. I’d occasionally buy or borrow a reference book from the school library, and I could study, experiment, and crank out all sorts of code using every page of those books. I always appreciated that I didn’t need to consume raw resources to explore the computer and create things. The internet just brought more access.

For nearly 40 years, I’ve tried to stick mostly to freely-available and open-source tools. I want to know that I can keep creating things without great expenses of raw materials or subscriptions. Free tools are likely to be the most accessible and have a community of users.

I don’t count electricity and internet access as programming expenses, since they’re useful for other everyday tasks. (I do sometimes worry about how I’d work on an airplane, though, with limited battery and maybe no internet access.) Will I consider subscription access to AI Models a utility?

I’d be uncomfortable having the AI agent do all this work, but I take the time to review, understand, and decide to keep or toss the agent’s work. If I don’t completely understand the work it did, I ask it for explanations.

I can still study and do work manually, and I often do, so I don’t quite feel like I’m held hostage to that subscription. The agent still needs to use the same tools I would use to do the work by hand. Without the agent, I may just need to focus my attention a bit stronger.

Years ago, I played a Timeline game with some coworkers, and I immediately wanted to play this game to sequence other events in more specific (or even personal) domains.

Years passed with only a TODO note, so this year, I figured this might be a good candidate to see what Gemini CLI can mostly do on its own in "one shot".

I wrote an initial spec, and I told the agent to read it and implement the whole thing. Of course, it ran off in the wrong direction for the first pass, so I deleted the resulting code, expanded the spec with some more detail, and kicked it off again. I did this about 7 more times, and the results kind of worked, but randomly included and ignored some of my directions. As usual, the guesses it took were mostly welcome, since I just needed to see something, but didn’t know what yet.

Spending tokens to generate the whole thing and throw it away spent lots of tokens and time. Is it really "one shot" if you need to do it multiple times?

A handful of those candidate applications would have been a fine starting point, so I kept one, and went into iterative mode. From that point, I asked the agent for small focused changes like I had for other apps. It’s really good at gathering the context from the existing project and implementing those fixes and enhancements. I (we?) worked much faster and consistently in small bites.

The agent was able to locate and build some datasets for me, but I also scraped and transformed a dataset from the Computer History Museum. That was one of my main motivations to get this project going.

My implementation of Timeline is fun to explore very specific domains and to learn. It’s like flashcards.

The Gemini AI will make some pretty good guesses about how a 3rd-party API may work. It is good at searching the internet, but when APIs have changed across versions, the old and new docs and examples it’ll find can confuse it. In a dynamic language and environment you’ll not spot these errors until runtime.

To combat the ambiguity and to give the AI agent more power to solve its own problems, ask it to add some tests around the code that uses the API. (In my case, the API is the XTDB client API.) Once it has a way to execute the code through tests, it’ll quickly start figuring out where it’s made mistakes and start running its own experiments to observe errors, search for fixes, and applying those fixes around the codebase. I exhibit the same pattern when I’m doing it by hand.

The tests also give you, the human, an easier entry point to evaluate the code the AI generated. If the tests look gnarly, you know to suggest refactorings to improve the architecture and make it easier to test. When the AI has the tests passing, and the test code is easy enough to read, then you can have a closer look at the application code to refine and keep that maintainable too.

I saw a tire was low on the display of my 2017 Chevy Bolt EV, so I topped up the tire with the tire inflator jumpstarter I keep in the car. The next day, I noticed it was still low, so I hit it again and brought it up a little higher. When I checked in the car again, I saw it’s still low, but the back tire is high. The TPMS sensors aren’t in the right place.

They apparently are the type of sensors that I can receive with my RTL-SDR as they drive by the house.

To get the car to register each sensor in the right place, I needed to tell it to relearn the IDs. I bought a little tool for $10 to trigger the sensors, but I apparently could have also done it by deflating the tires in turn to also trigger the sensor on demand.