Ask HN: What Are You Working On? (June 2025)

Fantastic. I wish I'd started on writing something like this years ago (although I'd wanted to teach explicitly rather than having a collection of how-tos).

> when I finally decide to compile into a book, everyone now uses AI

This is part of what discourages me from starting now, sadly. That, and having more concepts for actual Python projects than I know what to do with.

I'm trying to create the best A/B test sample size & duration calculator: https://calculator.osc.garden/

It's free (https://github.com/welpo/ab-test-calculator), and it has no dependencies (vanilla JS + HTML + CSS).

Right now it only supports binary outcomes. Even with the current limitations, I feel it's way above many/most online calculators/planners.

I'm still working on a German health insurance calculator. It evolved into a very elaborate recommendation tool.

Health insurance is one of the earliest, most important decisions immigrants make, and they often choose wrong. It can delay visa applications, cause coverage issues, or create expensive problems down the road.

Now they click a few buttons and get very specific recommendations explained in plain English. If they're confused, they can involve an independent insurance expert for free. The guy replies within an hour or two, and is cool with Whatsapp. The way I gather feedback from users, he's strongly incentivised to stay honest.

There is no AI involved, just good old-fashioned business logic. It means that the advice is sound, well-tested and verified by multiple competing experts.

It's such a far cry from either trusting whatever reddit or your employer tells you, or the slow back and forth of getting a quote from a (possibly dishonest) broker.

The second version[0] has been live for about a month, and the results are phenomenal. This third version vastly improves the quality of the advice, adding information about gap insurance for visa applicants, and making actual recommendations instead of listing all options.

It's a really fun project, even if the topic is boring. It's a great research, UX, copywriting, coding and business project. It's the product of a few months of hard work, and so far it seems to pay for itself.

[0] https://allaboutberlin.com/guides/german-health-insurance

I've been working on a 3D voxel-based game engine for like 10 years in my spare time. The most recent big job has been to port the world gen and editor to the GPU, which has had some pretty cute knock-on effects. The most interesting is you can hot-reload the world gen shaders and out pop your changes on the screen, like a voxel version of shadertoy. https://github.com/scallyw4g/bonsai

I also wrote a metaprogramming language which generates a lot of the editor UI for the engine. It's a bespoke C parser that supports a small subset of C++, which is exposed to the user through a 'scripting-like' language you embed directly in your source files. I wrote it as a replacement for C++ templates and in my completely unbiased opinion it is WAY better.

https://github.com/scallyw4g/poof

I'm writing tests, fixing bugs, and adding features to improve the quality of a piece of financial software that transfers certain financial data on a special private network. It's way less fancy than it sounds, but I'm enjoying improving the tests and adding important security and legal compliance features. Knowing that others will depend on my hard work to keep their business financial records straight is a great reward, and I am taking my responsibility seriously.

I'm also working on learning about building software with LLMs, specifically I am building a small personal project that will allow me to experiment with them using measurable hypotheses and theories, rather than just tweaking a prompt a bunch and guessing when it is working the best. I know others have done this, but I am building it from the ground up because I'm using it as a learning experience.

I plan to take my experimentation platform and build a small "personal agent" software package to run on my own computer, again building from scratch for my own learning process, that will do small things for me like researching something and writing a report. I don't expect anything too useful to come out of it, since I am using 1.7B/4B models on a MacBook Air M2 (later I might use my 3080 but that won't be much improvement), but it will be interesting to build the architectural stuff even if the agents are effectively just useless cycle-wasters.