An iPhone app that shows you hospitality options at each upcoming exit.

The "Add Stop" feature on both Google & Apple Maps never did it for me. The thing is, choosing, say, a food option during your drive has I think two distinct properties: It's prohibitive to go back, and you usually have an extremely narrow set of familiar brands to choose from that put you firmly in exploit territory. So you probably have a preference ranking in your head already, and now you need to weight their exit availability against your desired stopping time.

Seeing this app on HN today was one of those delightful moments when you realize that someone else had the exact same micro-frustration (*) you had, decided to do something about it, and did it exactly the right way. His write up about how he built it is also interesting and thought me about the OSRM routing engine.

I've always missed a "EUR stablecoin" category on coingecko, and when I came across usdc.cool a while ago, I thought it would be nice to have something similar for Euro stablecoins.

Codex pretty much one-shotted this, as in, the first version was basically functional. Surprisingly, it completely figured out how to get data directly from blockchain nodes of chains as diverse as XRP, Cosmos and Solana, including identifying public RPCs that work.

The hallway in our flat has one of these setups where the lights can be turned on via any of multiple rocker switches. So I wanted to understand how that works, to be able to upgrade that light with a smart relay.

I also love explorables. One exciting thing about the rise of vibe coding is that the barrier to creating this kind of interactive teaching tool is going towards nil.

So here is a quick vibe coded applet showing how a traveler-wire setup works. This was originally a one-shot ChatGPT canvas project, which I then asked Claude to polish a bit.

This was on HN a couple weeks ago. The idea is to let an LLM classify your bank statements into categories, but tally has an interesting approach. It's not just a prompt, and its not an app calling the LLM internally; instead, the your agent is supposed to call the app.

tally basically wants the agent to build a set of rules to match against your transactions, so it keeps telling the agent "call this subcommand to get some sample transactions that match no rules yet", the agents will generate rules that seems sensible, then invoke tally again to validate their work; tally itself merely implements the rule engine.

The agent still needs to be able to build a view of your categories primarily, and then your existing rules as well I suppose - because in some cases the right solution might be to update an existing one; I could see this becoming an issue as your ruleset grows. tally gives it the backpressure it needs.

I quite like the idea of externalizing the control loop to the user's existing agent of choice. It makes the entire thing feel instantly incredibly powerful compared to something that uses an LLM internally but gives you a fixed interface to work with. For example, you are able to tell Claude Code to make adjustments like "can we split this category in two" and it would work flawlessly. It also makes tools composible!

There is also a command (tally workflow) prints instructions to explain to agents how the process works. This reminds me of the beans prime subcommand. A pattern in the making!

Except for a small bug in HA, I'm quite happy with my Duette Top-Down Bottom-Up blinds. They have two rails, which allows you to cover, for example, only the bottom segment of the window, and letting light in from the top.

However, in Home Assistant they will be exposed two separate entities, which isn't that nice:

So today I vibe-coded a custom HA tile that combines two entities in a single slider:

This was done with Codex CLI; it struggled more than I would have expected to get the styling correct; I had it deploy directly to my HomeAssistant instance, then checked the dashboard manually. I would estimate overall time investment was 20 minutes. I have not looked at the code, but at 1.2k lines that is kind of more than I would expect. Then again, it does have a configuration modal too.