minikomi

Does the openai logo look like that because .. OpenAI? Then others followed suit?

Thank you! Something about the site allowed for wonderful serendipity for finding music. I had a blast.

Very fond memories of using Audiogalaxy, and also soulseek.

Soulseek especially had a community where you found someone who was into the same kind of music as you (obscure breakcore! japanese garage punk!) and could browse their collections, and chat to them also! What a wonderful way to make music friends and get good recommendations.

A clojure / fennel dsl for generating pure data patches, looking to make a small drum machine in love2d and being able to live update the internal patches would be fun.

I think it might be fun if the words told a story

    <somebody> just <did something>. 
    
    And it changes everything.

I have one, it's great

Crosspoint just released a new version

https://github.com/crosspoint-reader/crosspoint-reader/relea...

With:

- custom fonts

- better syncing

- quick-press refresh

Etc etc

Now translate it into zig!

Hello from eww

Sometimes it's better

To keep it all in a clump

Than spread it about

It's like a heavier commone colde

Or that they hate the project more than using AI

Cheers mate I added one at https://poyo.co/feed.xml

Asolutely. It doesn't have to be an either-or. I use gptel and org mode when I was to be really hands on driving the development. It's a very different mode of interacting with models, and the way newer models are trained to play nice with harnesses makes them very obedient.

https://poyo.co/note/20260202T150723/

Thank you for sharing.

Would you be willing to share some artists? I'm curious

The descriptions generated from the prompts are almost always great, but the generated music is always terrible. The sound pallete seems so limited.

A generalized llm prompting library for clojure, and seeing what falls out from that. I wanted something which was fun to use in an interactive way, but not too abstracted.

- Introduction: https://poyo.co/note/20260318T184012/

- Tool loops: https://poyo.co/note/20260329T034500/

- Playing with receipt extraction: https://poyo.co/note/20260323T120532/

- Use with async flow: https://poyo.co/note/20260410T164710/

Yep. Semantically distinct and meaningful chunks wins every time over any kind of windowing or slice and dicing.

Unfortunately, many people are looking for a fire and forget solution over an existing rats nest of documentation debt..

I think it's number 9 in the list

Component library page in the docs gives 404

I don't let it execute emacs lisp itself, but elisp generated in org mode babel blocks which is instantly executable is a fine way to have gptel improve itself.

( See "meta tooling" in https://poyo.co/note/20260202T150723/ )

The more you can afford to build up your understanding of the problem space and define what inputs & outputs look like, the more flexible you can be with evals. Unfortunately, this is a lot of work and requires thinking and discussion with your team and those involved.

https://poyo.co/note/20260217T130137/

I wrote about general ideas I take towards simple single prompt features, but most of it is applicable to more involved agentic approaches too.

I agree, and this is why I tend to use gptel in emacs for planning - the document is the conversation context, and can be edited and annotated as you like.

Given the recent surge of popularity in home-grown coding agents ala pi-coding-agent, I thought it would be nice to share my setup for semi-hands-on coding.

It's not a replacement for a more sophisitcated coding agent harnesses, rather an alternative interaction mode, and tools to support it.

Although the content runs fairly emacs-centric, consider it rather a nudge toward building your own small, self-owned tools.

When writing Janet, i enjoy the judge[0] style of testing because it's so interactive. I added emacs helpers, one for running existing tests and one for updating the tests. It made for a nice repl-like experience, especially nice when writing grammars[1].

[0] https://github.com/ianthehenry/judge

[1] https://github.com/minikomi/advent-of-code/blob/d73e0b622b26...

i too could not resist

https://g.co/gemini/share/6aa102571d75

I use gptel[0] with my denote[1] notes, and a tool that can search/retrieve tags/grep/create notes (in a specific sub folder). It's been good enough as a memory for me.

0: https://github.com/karthink/gptel

1: https://protesilaos.com/emacs/denote

Group rides, coffee outside, bike packing are all amazing ways to make friends. Shared adventures however small make long lasting bonds.

An example of why a basic understanding is helpful:

A common sentiment on HN is that LLMs generate too many comments in code.

For good reason -- comment sparsity improves code quality, due to the way causal transformers and positional encoding work. The model has learned that real, in-distribution code carries meaning in structure, naming, and control flow, not dense commentary. Fewer comments keep next-token prediction closer to the statistical shape of the code it was trained on.

Comments aren’t a free scratchpad. They inject natural-language tokens into the context window, compete for attention, and bias generation toward explanation rather than implementation, increasing drift over longer spans.

The solution to comment spam isn’t post-processing. It’s keeping generation in-distribution. Less commentary forces intent into the code itself, producing outputs that better match how code is written in the wild, and forcing the model into more realistic context avenues.