Building Valori, a Python-native vector database. It’s basically Lego blocks for embeddings: storage, indexing, quantization — all modular, all hackable. The goal? Let anyone plug in their own models and search pipelines without touching C++.

PYPI: https://pypi.org/project/valori/

How big of a threat is AI to your career? Upload your CV and it will be analyzed. You'll also get steps on how to reduce this risk.

https://www.isairisk.com/

Still working on this and some things will definitely change, but IMO the system prompt is already solid, so that the response isn't unnecessarily scary on one hand, but not too general on the other

A database populated with audio metadata (including a link back to YouTube or Spotify or whatever) that includes vector embeddings for the audio. That way I can grab clips of music I like from YouTube, generate vectors for them, then find similar things in the database.

It's off to a rocky start though, as I've initially populated it with YouTube-8M and AudioSet, neither of which are music-specific. The search results can be... Weird.

New (open source) PostgreSQL index type for analytics workloads, which is a read-only drop-in replacement for B-trees. Smol is multiplicatively faster than B-Trees and radically smaller.

https://github.com/asah/smol

Help, alpha testers, etc all welcome. Sorry RDS/Aurora users: smol is for embedded and self-hosted pg instances only for the foreseeable future.

Currently working on training language models steered towards certain "states of consciousness".

I have a model trained on publics datasets tied to brainwaves and/eye tracking and text comprehension (have this working well enough to experiment). Now I am training an adapter for various llm architectures to generate text steered to certain neural oscillation patterns (let's call them "states of consciousness" for brevity). I also have a 'rephraser' that rephrases text to elicit these certain states of consciousness. Overall experimenting with creating an suite of tools off my findings with how text relates to the eigenmodes of consciousness. My theory is once I do this I'll be able to do some...interesting things with "AI" agents. lmk if you want to talk about it if you're someone with knowledge in neuroscience/ML. My background is as a Software/ML Engineer so I could use additional thoughts. I do wish I could send a Github/docs which I will soon but this is currently a private project seeking investment for various research/public/private sector applications.

Unwound a couple of things from the stack!

Finished: the 100%-vibe-coded "GPT-5 reviews all my PRs on max reasoning" GitHub app (which is shockingly effective, https://github.com/Smaug123/robocop - probably nothing new for people who already use some product like this, but I like owning my own infrastructure as far as possible, and GPT-5 and perhaps Gemini are the only models smart enough to do this so I can't take this any further).

Currently: back on "write an immediate-mode TUI framework that uses a vdom as its fundamental abstraction" (https://github.com/Smaug123/WoofWare.Zoomies), in the hope that this is the first UI framework that I don't absolutely loathe.

Next: using the TUI framework, write a debugger to inspect the internal state of my deterministic .NET runtime (https://github.com/Smaug123/WoofWare.PawPrint) and to step forward and backward in time.

Next: get the deterministic .NET runtime to a point where a property-based testing framework can identify the deadlock in some very simple buggy multithreaded code. (The framework is not yet able to run Hello World - did you know that's an incredibly complicated program in .NET? - but it can solve a few Advent of Code problems right now, can perform some limited exception handling, limited virtual method dispatch, limited casting between types. Even getting to Hello World might take a year if I'm unlucky.)

Made a game: https://fourmula.awsum.info

I think it's too difficult in its current form.

A gamified bird sound ID app. Been working on it since May of 2024. While the core functionality of “catching” birds has been live for awhile, I’ve been working on a learning module built into the app. I’m experimenting with ways to teach new birders / nature curious people how to “tune” into different bird sounds. As a birder one of my favorite things to hone has been trying to differentiate songs and calls in the field and I want to capture that experience in the app.

Wings & Whistles: https://www.wingsandwhistles.app/

Love any and all feedback!

hosting a month-long residency for indie hackers in Da Nang, Vietnam!

we invited 10 of the best indie devs from around the world to live & build alongside us for a month at a beautiful villa. for free. (we have sponsors like OpenRouter, Cognition, n8n, and CodeRabbit!)

https://x.com/HackerResidency

we're 10 days into our first batch – would love feedback :)

I'm working on https://yap.town - an SRS based language learning app.

I would say it combines the best parts of Duolingo and Anki. Anki is great for memorizing words, but you don't see the words in the context of novel sentences. Duolingo is great for exposure to new sentences, but it's oriented around "lessons" and SRS is an afterthought. (Duolingo is also not designed for people serious about learning a language IMO, it's too easy and goes too slowly.)

Had to do quite a bit to get it to work well.

1. At first you would think that if you know all the words in a sentence, that should be enough to understand the sentence. But it doesn't work like that. For starters, words can have multiple meanings. The french word "bois" can mean "(you) drink" or "wood". You want to learn these separately. I trained an NLP model (a gemma3 finetune) that I use to understand the manner each word is used in each sentence: https://huggingface.co/collections/anchpop/lexide-nlp-models

2. Even then, what about a sentence like "you'd better not"? Even if you know the words "you" "had" "better" and "not", you still won't really get this. So I use the wiktionary "multiword terms" category for each language to get a huge list of terms like "'d better" , "you better believe it", etc, and teach these in addition to individual words. And then I only show sentences where you know all the individual words as well as all the terms.