Ask HN: What Are You Working On? (March 2026)

Deep link now ( https://Deeplinknow.com ) - deferred deep linking for developers / people who dont want their links blocked by adockers because Branch/Appsflyer et al are actually under-the-hood cross platform ad tracking services.

I do no tracking, no analytics, just help you cross the airgap between web and mobile app so you can send users to the right place (and track them however you deem necessary)

I'm rewriting a shipping app, that is just over two years old.

This is a "full rewrite," because I need to migrate away from my previous server, which was developed as a high-security, general-purpose application server, and is way overkill for this app.

Migration is likely to take a couple more years, but this is a big first step.

I've rewritten the server, to present a much smaller API. Unfortunately, I'm not yet ready to change the server SQL schema yet, so "behind the curtain" is still pretty hairy. Once the new API and client app are stable, I'll look at the SQL schema. The whole deal is to not interfere with the many users of the app.

I should note that I never would have tried this, without the help of an LLM. It has been invaluable. The development speed is pretty crazy.

Still a lot of work ahead, but the server is done, and I'm a good part of the way through the client communication SDK.

A few things.

I've been on/off working on a Forth compiler for the NES. It will be open source soon enough but I'm not happy with the code right now as it's extremely messy, repetitive, and buggy, but I think it's turning out ok. I am resisting the urge to use Claude to do all the work for me, since that's depressing.

I've also been working on a clone of the old podcasting website TalkShoe. It's nothing too complicated. It's mostly an excuse to learn a bit more about Asterisk and telephony stuff. I'm hoping to have something fully usable in about a month or two.

I forked the main MiSTer binary due to some disagreements I had with Sorg in how he's running things [1]. My fork was largely done by Codex and Claude, but the tl;dr of it is that it has automatic backup of your saves, tagging and versioning of your saves, and it abuses the hell out of SQLite to give better guarantees of write safety than the vanilla MiSTer binary gives you. I've been using it for a few weeks now and it seems to work fine, and it's neat to be able to tag and version saves.

I think that's mostly it. I'm always hacking on something so there might be a straggler there.

[1] https://github.com/Tombert/Main_MiSSus/blob/master/README.md

https://fitcal.app syncs Strava activities to your Google calendar. No fancy features, just does what it says on the tin. Really fun to build out with elixir + phoenix.

When training I like to have every day mapped out with how many miles to run, at what pace, etc as an event in my calendar. My actual workout gets uploaded into Garmin and Strava, but I always wanted it back in the calendar so I could see at a glance the consistency over time. It's been really fun to see other people use and get value out of something I built for myself.

Training a tiny LLM for fun using Rust/Candle - I constantly tweak stuff and keep track of results in a spreadsheet and work on generating a bigger corpus with LLMs. It's a project for fun, so I don't care about finding actual human generated text, I'd rather craft data in the format I want using LLMs - Probably not the best practice, but I can sleep properly despite doing that.

My favorite output so far is that I asked it what life was and in a random stroke of genius, it answered plainly: "It is.".

It's able to answer simple questions where the answer is in the question with up to 75% accuracy. Example success: 'The car was red. Q: What was red? ' |> 'the car' - Example failure: 'The stars twinkled at night. Q: What twinkled at night? ' |> 'the night'.

So nothing crazy, but I'm learning and having fun. My current corpus is ~17mb of stories, generated encyclopedia content, json examples, etc. JSON content is new from this weekend and the model is pretty bad at it so far, but I'm curious to see if I can get it somewhere interesting in the next few weeks.

https://github.com/antoineMoPa/rust-text-experiments