Ask HN: What are you working on? (September 2026) (self)

I have ALOT of free time since quitting my job.

0) money-money-money: A sub 15Kb server-less opinionated household budgeting tool [0]. Why? Its made to be extremely compressible, so you can edit your household budget and share back and forth with your partner via short-links or QR codes. The idea is you edit your budget, send the link to other party, they can view or edit and send back to you without a server in the loop. [0]

1) slices: An alternative AI-first ADHD-friendly architecture experiment where everything you code is a "slice" with a shared event pool. A slice is a fully self contained "feature". Slices can only communicate with other slices through a shared event pool. Inspired by small-talk, vertical-slice architecture, and ADHD. I used slices to build a self-modifying IDE too. [1] [2]

2) text compression experiments: I'm experimenting with approaches to compress text in pursuit of the Hutter prize. The idea is to take the entirety of the English Wikipedia, and compress the text as small as possible whilst still being able to reproduce it. [3]

3) tiniest maze solving neural network: A write-up work in progress on a maze solver I built with heavy LLM help, managed to get a 14 byte neural network solving 96% of unseen mazes. [4] [5]

[0] - https://con-dog.github.io/money-money-money

[1] - https://github.com/con-dog/slices-demo

[2] - https://con-dog.github.io/slices-demo/

[3] - http://prize.hutter1.net

[4] - https://github.com/con-dog/tiny-neural-network

[5] - https://minimio.ai/

I've been working on a voxel game engine called Bonsai for ~10 years.

Probably the most interesting thing about it at the moment is the editor. The world, and most things in it, are represented as collections of SDFs. More accurately, they're density fields, but, potato-tomato.

Bonsai has undergone a large rewrite over the last couple years that's nearing completion. A world edit is defined as a bunch of SDF parameters which get projected/rasterized into the voxel grid by a shader on the GPU. One neat thing about SDFs is they've been thoroughly researched and documented by a guy named Inigo Quilez, and they have a lot of nice mathematical properties. For example, you can do a smooth union of arbitrary SDFs to get nice rounded contours where shapes join.

I've written every system from scratch, all the way from the memory allocators and font rasterizer to the collision detector and simulation loop. I even wrote a metaprogramming language as a replacement for C++ templates, which is a whole other story. IIRC the only external dependency is the C runtime library for starting the process and my very occasional use of variadic functions arguments.

For a long time, an explicit non-goal of the project was to ship a game. It sounded insane to me to write an entire 3D engine and then ship a game. As it turns out, I've gotten it to the point where I can actually make a game. I've got a start on the game systems in a closed-source repo, and hope to have a steam page for it by the end of the year.

I'll do some shameless self-promotion and leave some links here for anyone interested in looking at the engine, language code, or some pretty pictures.

https://github.com/scallyw4g/bonsai

https://github.com/scallyw4g/poof

Two years ago, I learned that less than 5% of the Latin Renaissance had been translated to English and less than 40% had been scanned. To me, this seemed urgent to address — if we wanted the Renaissance in AI training data…

Now, we’ve made SourceLibrary into a hub that provides over 20,000 translations of classical literary traditions from Sumerian to Chinese to Armenian to Sanskrit… and of course Latin and Greek. Recently, we realized we have more words than English Wikipedia!

You can freely connect to the public MCP for research; in Claude, it retrieves images into the chat.

As a philanthropic project, we are almost completely free and open source. However, we are hoping to fund our stewardship work with AI training budgets (is that reasonable?). We are based at the non-profit “The Embassy of the Free Mind” in Amsterdam, a rare book library devoted to free thought and mysticism. https://embassyofthefreemind.com

This has been the most fun project to work on!! I’d suggest trying the free research agent with a rabbithole topic of your choice: https://sourcelibrary.org/librarian/

I'm buildingan open-source way to fill U.S. immigration forms without fighting clunky PDFs

Github: https://github.com/athos2113/fillvisa-os

Demo: https://fillvisa.com/demo/

US Immigration still relies on outdated XFA PDFs - you can't fill them on your browser. Most immigrants end up printing the form and fill it manually.

So, I converted the PDF forms into smart web forms. They are replica of the official USCIS forms and follow conditonal logic. In the output, you get the official USCIS pdf form filled.

I'm building InterviewMe ( https://interviewme.now/ ). It lets employers interview you via AI and is intended as a better way for people to get deep answers about your work history and experience than they could by looking at a simple online resume or LinkedIn. You upload your resume, answer some questions about your background, optionally upload any other documents you have (cover letters, awards, articles, etc) and it ingests all of that and creates a conversational agent people can use to ask questions about your professional experience. Check out my profile to get an idea of how it works: https://interviewme.now/@max and then get your own: It's free!

I’m building TraceDB[1] because I think observability needs an overhaul. We pay a premium to index millions of near-identical spans, sample away the rare ones that might explain an outage, and still have to piece together what failed, where it started, and how it spread. All the data is largely there, but no one can make sense of it! Observability was hallucinating before AI made it fashionable.

I’m seeing near-100% root-cause accuracy on synthetic test data and am looking to validate that on real production workloads using OpenTelemetry tracing. If you’re open to chatting or experimenting, drop me a line! I’d also love to hear ideas or how we can improve observability.

[1] https://tracedb.ai

I'm building https://matgoat.com/en/

A way for BJJ and martial arts academy owners to manage attendance, payments, class schedules, and student progress.

I train BJJ myself and saw how much time instructors lose to admin work, so I wanted something simpler and more focused than the generic gym software out there and have been having a lot of fun building it with feedback from real academies.

I’m building https://www.freefocusgames.com, a free collection of browser-based attention and working-memory exercises, including Dual N-Back, Stroop, Schulte tables, attention-control exercises, and memory games.

I started it because I wanted simple brain-training exercises that didn’t require an account, app install, or subscription.

It’s also open source: https://github.com/loethen/freefocusgames

About fifteen years ago a puzzle game called Trainyard came out for iOS and took my dev team by storm for a few weeks. It ceased getting updates some years ago and isn't on the app store anymore. So I came up with a variant on it and just put up https://www.geoffcanyon.com/lightyard.html I need more puzzles to include in the base pack, if you come up with something interesting, export it and send it to me!

I'm making a stair lift for my Roomba. It fits into a small box and you can retrofit it on any stairs. I got sick of carrying it up and down manually and didn't want to buy one for each floor. It also carries my laundry up and down.

Bunch of things: 1. Building a local ai news feed to learn what's going on latest in terms of technical discussion, research papers, projects - https://local-ai-signal.vercel.app/ 2. Build a second brain by injecting years of my notes using hermes, luna model and obisidian

I'm working on building open source control firmware that runs on the Masterbuilt Gravity series electronically controlled charcoal smokers.

https://gitlab.com/prbs23/freefall_800

The smoker hardware itself is pretty good, but like most IoT devices, the firmware and app sucked. So I reverse engineered the whole control PCB, and have been rebuilding new open source firmware from scratch. All the basic functionality is working now with a fully local control web app. Plus an Andoid app, and Home Assistant integration.

Currently working on interface refinements, and some more advanced control sequencing features.

I wrote up all the reverse engineering details in a blog post here: https://www.prbs23.com/blog/posts/reverse-engineering-gravit...

Sunday dinner. Roast pork loin, rice and sauteed veggies on the side; and a big salad. Simple, and I get leftovers for another night or two.

I'm working on making bcachefs useful when you mix both SSD and HDD in one filesystem.

At the moment, even if you set up SSD for foreground operations and HDD only for background, more often then not, your write latencies are those of the HDD. We can do better.

See https://paquari.com/posts/bcachefs-ssd-hdd-progress/ for a progress report. (Disclosure: the write-up is an AI summary of my notes, so don't expect any brilliant prose.)

I'm also working on resurrecting stabilizer (https://github.com/matthiasgoergens/stabilizer). Stabilizer is a way to make program performance less dependent on the linker and layout lottery. But it has sadly fallen victim to bitrot.

I'm doing some minor contributions to other parts of Linux like ZFS and ext4, mostly as a byproduct of experiments I'm running for my bcachefs work. Another somewhat ambitions project is to make swapfiles on ZFS and bcachefs work (or work better), at the moment the problem is that writing swap on them might need to allocate memory exactly when you are out of memory.

I'm working on a new independent LaTeX engine written in Rust and no C dependencies at all. You can see it working at https://telox.dev/app/scratchpad/

There's currently not a way to invert LaTeX engine output for bidirectional editing. So you can't easily drag objects around or interact with the page on LaTeX. That's what I'm setting out to change in the name of mathematical exposition.

I have more details on my plans here https://news.ycombinator.com/item?id=49689856

I've experimented with a few ways of making the editing realtime and I'm getting in the 10 millisecond ballpark on some edits, but I want to work backwards from the light speed path to see what it takes (besides a lot of time).

My quirky wishlist app https://thingstohave.app is finally done and heading to public launch this or next week.

I already partially covered the last few months of app development in this topic. But actually, its development took way longer, as only this iteration started in 2024. So it is the end of a long milestone— this launch date initially was to be a year ago, but I procrastinated and missed important for such apps holiday season

This year I was able to go way beyond last year’s backlog, and now I’m releasing a fully finished app instead of early access.

I’m proud of the result — a lot of overengineered stuff that is not viable in any commercial product, especially in what appears to be a simple CRUD app. For example: image loading from blur, unadvertised but very advanced DnD everywhere, natural language currency inputs with sorting, platform-agnostic quick add, and many more.

Its stack is also very fun: it has been running on the Cloudflare Workers ecosystem since 2023, and is powered by a custom Inertia.js adapter for Hono (that predates the official adapter by one month, lol), with Vue SSR. That all enables me to write old-school MVC with Vue as a template engine for views. It also proved to be very easy to work with and cheap to run.

With a tsunami of vibcoded beige wishlists, my weird app stands apart, for sure, but I’m unsure about its appeal to the average customer.

I've been slowly working on https://github.com/CanineHQ/canine for about 2 years now. Basically a self hosted, FOSS that turns a Kubernetes cluster into something as easy to use as Heroku. Its grown to about 2000 active developers using it host small things at https://canine.sh

It was a pain point from a start up I cofounded before where we needed Kubernetes flexibility but developers hated using it. I figured there was no reason why it had to be hard to use and thus Canine was born.

Lately been playing around with setting up remote dev containers on your own compute as well -- basically like github codespaces, except self hosted.

Built in Rails + Stimulus

Recently worked on https://sleepsignal.app

It's an app to get sleepy without meds. Simply close your eyes and imagine the micro-scenarios you hear. Like if you hear "moonlight on a white flower", imagine that scenario until you hear the next one. Session duration is also not too long (10 minutes default, you may change the session duration too).

It's like a digital melatonin pill - but without chemical side effects. Try it!

I'm needlessly reinventing the wheel in replacing almost every part of my X11 environment like a menu, bar, wm+hkd (except terminal since I don't want to mess with carrying terminfo around) to be portable and not at all interdependent.

AI is making it possible to (expeditiously) have everything I want or need on Linux or BSD written in POSIX shell or C is doing wonders for me. The initial effort and time pays off in knowing it will essentially function as intended forever, and no developer can make changes that disrupt my workflow or preferences for my environment.

My wife and I continue to work on a private, paid, ad-free and personally customizable search engine: Uruky [1].

To signup you don't provide any information (a randomly-generated account number is assigned to you), and you can run a proof-of-work captcha to get 2h for free. You can choose among many different search providers (for defaults and per query), including Uruky Site Search, powering our own index.

Last month we reached 300 monthly active accounts and released an image search gallery mode, plus a simple calculator and conversion widget!

The main differences between Uruky and Kagi, DuckDuckGo, SearXNG, etc. are visible in the footer (right side), but one huge difference is that with Uruky, after being a paying customer for 12 months, you get copy of the source code (licensed as BUSL, into AGPLv3 in 2 years — a suggestion made here on HN)!

Our main challenge continues to be discoverability and outreach because we want to do it ethically (no Big Tech and no GenAI/LLMs). Ideas are welcome! We’ve been sponsoring open source projects, open source maintainers, and indie, small-web, and privacy-related websites and applications/groups/orgs. This month we're sponsoring NOYB [2]!

Feature-wise, for September the most visible things that shipped already were Tags and Scopes (top feature requests for a while). We’re also still (slowly and sustainably) increasing our own index, focused on indie/small web.

Thank you for reading this!

[NO-AI]: There is no generative AI product or service being offered, here.

[1]: https://uruky.com

[2]: https://noyb.eu

Jane Street released a new challenge so I'm considering picking it up. It hasn't fully captured my brain, but I can feel it in there from time to time and depending how it lands it might be my next project [0]. I'm not really a hardware person but I'm close enough that it's maybe within reach

[0] https://blog.janestreet.com/protocol-emulator-asic-competiti...

Me and my wife (then my girlfriend) have written "realistic" space sci-fi in Finnish for over 20 years. Now with AI we are finally able to translate our stories to English. The quality starts to be there.

There are a lot of small languages in the world, and translation used to be extremely expensive. And there is never too much good space sci-fi.

(We don't write space battles or wars. Our style is slower and bit more philosophical. Light speed is the limit, and travel takes centuries.)

At the beginning of this year I was fractional CTO at a small b2b startup and was surprised by how what I considered to be the basic requirements of a customer support platform was split across multiple tools and all priced per-user. This essentially forced a small team to limit who in the org had visibility into what was being communicated to customers.

I set out to solve that with https://stayupfront.com and now that the build is in a good place i’m finally having to bite the bullet and learn marketing. I’m also realising that all the advice to start with marketing first was probably good advice!

I’m working on two projects that make public data easier to explore through maps.

US Energy Cost Explorer (https://energy-maps.com) is a free tool for exploring electricity and natural gas costs across the US. You can compare states, follow prices over time, and look at bills and energy burden. I work in energy affordability, and I wanted to make this information easier for people to access and understand.

MarketGround (https://marketground.io) lets you draw an area or choose a travel time around a location, then explore its population, income, housing, and nearby businesses. It also shows daytime versus nighttime population, which is useful when thinking about who might actually be around a potential store or restaurant during business hours.

MarketGround is a side project, and I’ve found getting it in front of potential users harder than building it. I’d be interested in hearing from anyone who has evaluated a retail or franchise location. What information mattered most, and what was difficult to find?

I've been building my own IDE that works the way do (and those in my social circles/professional life do). It allows for use of multiple agent harnesses (i.e claude code, codex, pi, opencode), and makes worktrees much easier to work with. Also has its own code review workflow. My coworkers and several friends have all started using it now. Its been an absolute blast to build, and has definitely increased our productivity vs VS Code, Zed, or Cursor.