inguz

Nice work!

Nice! I look forward to digging in to this.

My own slightly-similar thing (uses an explicit sequence, and signed objects in refs/): https://github.com/generalbusiness-ai/gitseq

"software is made between commits" (their DeltaDb blog at https://zed.dev/blog/introducing-deltadb)... yes!

I tend to think the important bit is traceability of why the software changed in the way it did, but not yet convinced that Delta will help with that.

Not mentioned in the article: macros, which surely are the traditional way to work with this in C and C++.

Reminded me of when I had a Teletype ASR33 on AirPrint (https://github.com/hughpyle/ASR33/tree/master/cups). With LLMs that would have taken a half hour. It would have maybe been less fun? but there's a lot of time faffing around on PPD that I'll never get back.

This is great. Would prefer an eight-year-old play/explore/build in this sort of world than just about anything else!

I quite clearly remember this magazine issue from the time.

A giant, god-like hand reaching out from the clouds. Tagline "THE JOLLY GIANT DELIVERS THE GOODS".

https://www.computinghistory.org.uk/det/1706/Personal-Comput...

Personally I think MUDs are a great platform for the agent/human coordination things that everyone's trying to build.

They walk among us https://x.com/hughpyle/status/2050255675255669222

In particular: tell your agent to connect to it using MCP. That's the cleanest way for agents to "live" in the world; I'm using it from Codex and Claude Code, but haven't yet tried with OpenClaw/Hermes/similar.

Retro-plug: here's a Teletype ASR33 simulator with sound https://www.youtube.com/watch?v=jd5oomwEBb0

Tinkerpop/Gremlin is not really SQL-like at all. My experience with it has not been fun: it's difficult to reason about, has obscenely bad documentation, poor tooling, idiosyncratic library support, and simple data access patterns are hard.

A little more digging turned up a recent paper that uses Histogram of Oriented Gradients (HOG) along with neural-network training: http://www.jsoftware.us/vol13/355-SE3002.pdf

After some more investigation... their paper is interesting. But: it's $25 to download their model file, and the trainer code and training dataset aren't published. Oh well.

Nice!

There's a text-printer 'Cedilla' (https://www.irif.fr/~jch/software/cedilla/) that uses this composition across large parts of Unicode - to quite good effect it seems.

The Teletype doesn't do backspace (!) but can do carriage-return... so e.g. to print "Björk" , the text is: BJORK(CR)(space)(space)(quote). I have a Teensy microcontroller that already handles some ANSI-escape characters, and implements the 'backspace' ^[[D this way (https://github.com/hughpyle/ASR33/blob/master/firmware/ansi_...). Next on the backlog is "bold", using a similar strategy. Fun times :)

Very very cool. I love their approach of looking for similarity metrics that are less sensitive to local misalignment, and this is something I'me very interested in exploring in the future.

I'm not completely convinced by their distinction between "tone-based" and "structure-based", though. When they talk about structure, it seems to primarily mean 'edges'. Instead maybe a more useful distinction would be between "local optimization" (at the level of a single printed character), and optimizing "globally" or over a wider area of the image, and so allowing an iterative optimizer to find a really compact representation.

Very interesting - thanks for the link. (This is exactly what I hoped HN would show me!)

Several interesting things there. The use of Structured Similarity Index (SSIM) is probably better/faster than the Histogram of Oriented Gradients (HOG) that my project uses. And then, at runtime, I'm still doing a brute-force match (which is really slow) -- clearly there are big performance gains to be made by training a classifier.

Their results are really good!

Thanks :)