alt Hacker NewsAsk HN: What are you working on? (May 2026)
What are you working on? Any new ideas that you're thinking about?
Comments
Working on trying to give some guidance on solar panels.
Plug in solar became legal here in the UK
Still sussing it out but started shipping something
Finding the pitch direction of the roof is kinda hard
Uses data from the house to try and get a rating
I continue to write. Unsure the end goal but in my recent dispatch I learned to get out of my way, stop intellectualizing life and let it be.
Employee benefit plan analytics. Had a huge dataset long ago as a consultant to the industry and finally vibecoded up a decent frontend. All public data but if you know the data there is a bunch of analytics you can do. Just about to launch and do some marketing in a few weeks, so saw this and thought I'd throw it in!
I'm working on JRECC, a Java remotely executing caching compiler.
It's designed to integrate with Maven projects, to bring in the benefits of tools like Gradle and Bazel, where local and remote builds and tests share the same cache, and builds and tests are distributed over many machines. Cache hits greatly speed up large project builds, while also making it more reliable, since you're not potentially getting flaky test failures in your otherwise identical builds.
It's been three months now of building Crit - https://crit.md - local first, open source tool for reviewing markdown and code output from your favourite AI agent.
It's inspired by GitHub PR review workflow, only with quick iterations and local.
It's been great! I found some dedicated users, dogfooding it every day with Claude and starting to get more contributions from the little community. We just got accepted into Homebrew core which was my target.
I'm expanding the team features now as I've got a few users keen to get the sharing service deployed in their private networks!
A small, generic Go library for retrying fallible operations with exponential backoff and pluggable jitter strategies. https://github.com/nodivbyzero/try
I’m working on a ground-up implementation of RADIUS with everything running on stateless compute. It’s a beast with many problems to solve but I have EAP-TLS, TTLS and PEAP all working. I’d love to connect with folks interested in this kind of thing.
I'm making a surf forecasting site a la surfline.com, I started mostly to have an API to use for my tidbyt, but I figured I might as well make it a full thing and built my own features! It's on quickswell.com but it's only Socal at this time (fewer spots to compute)
I've been going through Nora Sandler's Writing a C Compiler book and writing a compiler in Python. I'm excited to start the chapters on optimization - those seem like the most fun algorithm problems.
I recommend the book. It certainly isn't easy (maybe 3x harder than Crafting Interpreters), but I've learned a ton (eg how to deal with operations on different sizes of types, or the trick of using pseudoregisters to avoid having to figure out registers up front).
I'm working on a tool to let sfotware developers write well-formatted ebooks and printed novels: https://frequal.com/epublish/
Example book here: https://www.amazon.com/dp/B0GYCZJVGX
muster and muster-pattern-library. (https://github.com/azide0x37/muster)
an agentic coding scaffold/framework you can reference when building out your next random raspi project. prefer to build around systemd units first; make an idempotent installer script, then put as little as possible custom coding around that.
`impl muster` comes down to: /build out this tool wiring together `patterns` like: C3.dropfolder-trigger; R2.device-binding; C4.lazy-resource-gate
or composite patterns like:
T2R4.device-triggered-conveyor "Bind a physical device event to a bounded ingest job that waits for hot-storage capacity, proves cold-storage capability, stages local work, and hands output to a hot/cold conveyor."
I need to back up a couple hundred DVDs, so with muster I get out:
dvd-ingester T2R4.device-triggered-conveyor
Architecture DVD media becomes ready -> udev rule adds SYSTEMD_WANTS=dvd-rip@%k.service -> systemd runs /opt/dvd-ingester/current/bin/dvd-rip-one /dev/%I --apply -> dvd-rip-one proves DEST_DIR and waits for HOT_DIR capacity -> completed rip moves to HOT_DIR/<run-id> -> dvd-publish-one.timer drains HOT_DIR to DEST_DIR -> publish writes DEST_DIR/.incoming-<run-id> and atomically renames final output
Pipelined; ejects after rip completed. Monitors local disk capacity, retries after NAS comes back online; resumes after random reboot; etc.
https://brettkoonce.github.io/lean4-mlir/blueprint/
https://github.com/brettkoonce/lean4-mlir
I (w/ Claude) have built a framework for writing neural networks in Lean 4 that compiles to StableHLO MLIR and runs on GPU via IREE.
Main project is a deterministic .NET runtime (https://github.com/Smaug123/WoofWare.PawPrint). Today I upgraded it to net10, which has naturally caused dozens of regressions which Claude is beavering away at.
Side project is my own agent harness, https://github.com/Smaug123/writ , which is being built sandbox-first and with Nix as a first-class citizen. Obviously everyone has to write their own agent harness as a rite of passage.
I'm working on Isola (https://github.com/isola-run/isola), a passion project of mine. It allows you to easily create and control sandboxes for executing untrusted code on any Kubernetes cluster (one helm install). To support some features I wanted (like on demand snapshotting of specific containers' filesystem in a sandbox pod or limiting egress rate from the sandbox I am working on now) I contributed some changes to gVisor. Happy to chat about the design and implementation of such a project if anyone interested!
Currently working on `imgsrv` which is basically a container registry but it holds disk images. Enforces versioning, allows attaching multiple formats for a single release, prioritizes immutability, etc. Intended to build fully automated image release pipelines. I use a PXE setup for my homelab, so having a common place to manage image release lifecycle is helpful.
Right now I intend to make it compatible with Incus as a remote. So it's just a matter of adding it as remote and then you can consume all of your versioned images.
I'm working on Repple (https://repple.sh)! It's a modern spaced repetition x incremental reading/PDF library app with a few (tasteful!) QOL AI features.
I've been using Anki for 10+ years and love it but always wanted something with a cleaner UX and a reader view. The recent Anki ownership change pushed me to finally make something, and it's seeing some traction :)
Right now I'm focusing on getting the reading and note-taking view to be nice. I used to use Polar Bookshelf (RIP) but that went away, trying to make something better.
The flashcard side also has a REST API btw!
I have been experimenting on using AI for hardware development. I showed some experiments on HN a couple of weeks ago (https://news.ycombinator.com/item?id=47801255). I am now trying to make my approach a little bit more comprehensive and structured, instead of several disjoint MCP servers, a single platform that connects lab instruments to AI assistants: https://teasel.tools/
As a demo, I repaired an old Philips PM5190 function generator (about 40 years old) and connected it to Claude Code. Lots of fun. Going to post a follow up video the next couple of days.
I’m building a small map application that allows me/friends/family to explore data overlays about morel mushrooms phenology and habitat (ground temp/moisture/terrain/aspect/tree species/etc) in our area. There’s some lightweight forecasting and timing models to help guess at near-term fruiting. I had a big push about a mouth ago to tighten things up, and initial experiences in the field this year have been very promising.
I’ll keep chipping away at it this year, and probably expand beyond morels to other seasonal natural phenomena that my people enjoy like smelt/salmon run, wildflower blooms, etc.
I'm building DB Pro, a modern, beautiful, and now fully self-hosted database client for desktop and web.
Just launched Studio, which is the self-hosted version of DB Pro.
I also keep a devlog. #9 was just published to YouTube.
Self-Host Your Own Database Client | DB Pro Devlog #9 https://youtu.be/MJvSrJGtk70
I'm continuing to focus on ways to set strong code quality guardrails in an era where most code is not handwritten.
The result is http://getcaliper.dev.
It has a number of mechanisms that help substantially:
1. It can extract deterministic quality checks from your CLAUDE.md text; these checks then get executed after every agent turn.
2. It performs a lightweight ai-powered review at every commit; feedback goes directly to the agent, which can then make corrections.
3. It performs a more 'traditional' deep AI review at merge, or on-demand.
Free to use, just bring your own API key. Any and all feedback is welcome!
Building a small voice journaling app, hoping to ship in the next few days https://turquoisehexagon.co.uk/anima
TestFlight link, good for 10 users: https://testflight.apple.com/join/9VREtXzq
I am building Akariq which provides data eSIM data plans to 185+ countries and regions across the world. I am 2-3x cheaper than big brands in the same space. I also prefer local data routing i.e. I don't route traffic from US or EU to Hong Kong, they stay in your country / region.
Still working on ghidra-delinker-extension, trying to wrap up the OMF object file exporter at the moment. Then I'd like to implement generation of debugging symbols (at least DWARF and CodeView, maybe STABS and CTF), although lately I've received a PR for PowerPC and an issue for delinking shared objects.
I'm also thinking about writing the Necronomicon of delinking at some point. The extension keeps spreading by word of mouth and there's only so much UX improvements I can do, for something that requires throwing everything you've learned in CS 101 into the trashcan before you can "get" it.
I’m building a UI design app similar to Figma or Sketch, but with a few differences:
1. Responsive artboards and flex-like layout engine
2. Deep support for design tokens
3. HTML/CSS previews and export
4. Multiplayer AI and human collaboration. Agents can connect to documents and collaborate like any other user.
Built in Swift and cross platform Mac, iPad and iPhone.
I’m designing and building the UI and implementing the underlying features with Codex. So far it’s going surprisingly well.
I’m working on a tiny terminal config / dotfiles / tool installation manager so I can keep everything in sync between my machines. Also includes profiles so I can tailor each machine how I see fit. https://github.com/phalt/pauldot
It’s intended just me for and follows a philosophy around hyper personal software that I’ve been developing: https://paulwrites.software/articles/hyps/
Working on Marmot https://github.com/maxpert/marmot recently added support for vector index. My local benchmarks show pretty decent QPS with less than GB of RSS on DBpedia dataset.
Interesting part is that I started off implementing a research paper for indexing and performance was not good enough. I ended up tuning things up for my own use-case and ended up with good enough replicatable RAG store.
I'm working on https://fdeploy.com
fDeploy is a self-hosted Windows deployment automation tool — a lightweight, on-prem alternative to Octopus Deploy. It consists of a Server (Windows service with a Web UI) that orchestrates releases, and Agents installed on target windows machines that execute deployment steps (IIS sites, file copies, scripts, etc.) across environments.
I vibe-produced a website for an (earnest?) pastiche of Final Fantasy games (i.e., for a game that does not, and probably never will, exist). I noticed that Nano Banana could generate reasonable facsimiles of Square Enix's promotional render style and ran with it. Next up: faking some gameplay or a few shots from an FMV-style cutscene.
I've been working on a AI app to replace Claude Desktop and Mobile for personal use.
The main goals are to own my data (memories, artifacts, chats), be able to switch AI providers at any point (if one is down or I want to try a new model), have the same experience between desktop and mobile especially when it comes to working remotely on code.
A bigger vision is to offer everyone a alternative to Claude and ChatGPT they can own just like OpenClaw but with a great app experience.
I hope to have the first beta published by the end of next week.
I've been iterating on the type system of a somewhat-kooky programming language I'm developing: https://github.com/mkantor/please-lang-prototype
The other day I got into some pretty weird territory with Claude, trying to map out what an immortal ASCII cat would look like. Basically an autonomous Tamagotchi.
The idea was to create a quine that runs forever on something like Akash network with its own crypto treasury to support and pay it's bills and try to replicate. It would then talk to an LLM for support and actions on what to do to stay alive.
It got pretty out there. Stored some of the ideas here.
A couple of fun random projects:
- https://shirt.cash - Vibe code your t-shirt ideas and sell them.
- This weekend was substack MCP (https://www.youtube.com/watch?v=jHARlcInLqU)
new ideas welcome lol
Working on a tool that lets you author in WordPress as usual (own Docker container, full editor + plugins) but exports the site to static HTML for the public version, so PHP doesn't run in front of readers. Deploy targets are Cloudflare Pages, a Git remote, or statichost.eu. Solo, just launched, currently grinding through hardening. Called Stelae if you want to have a look.
Adding support for the threads/atomics proposal to my Wasm to Go transpiler: https://github.com/ncruces/wasm2go
Since I started it a couple of months ago, it's been used by me to transpile SQLite to Go, and by some other folks to transpile other C, C++, Zig and even Perl libraries to Go.
I am working on mesh (https://growmesh.io) I started working on the topic of human development two years ago and I dived into the topic of how humans have been trained, manipulated, educated or brainwashed. My central idea which I am investigating is that whenever a person is interacting a highly tuneable ML model such as the X/Fb/TikTok feed or chat interfaces with LLMS, does the thinking and development of the human happens more times or less due to the new experience.
pod2book.com It converts podcasts to e-books for the deaf, hard of hearing or neurodivergent because transcripts are clunky.
A no-code platform packaged as an AI tool for building data-driven applications and serving as a data store for AI to tell it interact with your data; https://saasufy.com/ - Tested with Claude Code and pi.
I’m working on Beanback, my side project SaaS app for ‘effortless digital loyalty for cafes’.
It replaces paper stamp cards with Apple Wallet passes (Google Wallet coming soon) without the need for customers to download an app or signup. It’s still very work-in-progress (forgive the landing page) but I’m enjoying using Ruby on Rails. Please let me know your thoughts!
For occasions like birthdays or Christmas where people want to give you gifts, I have always wanted to ask them to make donations to charities of my choosing instead. So I built an app to enable this: http://donateyourgift.com/ It is very simple but I didn't find anything quite singly-focused like it, so I built it just to scratch my own itch.
I am working on Desiderata (https://github.com/github-of-NMI/Desiderata).
An LLM benchmark for open-weight models only, with secret questions.
The questions are asked multiple times to calculate a consistency score.
The results are available in JSON, containing the hash of the question with the number of correct and incorrect answers, the number of unique answers, and the number of times no answer is given. (Uses \boxed{})
Menu bar app that reduces your Claude Code token costs by ~50% so you get 2x more usage out of your plan.
People seem to like it so far :-)
I’m working on bomberman in ClojureScript, using no libraries, and making sure I write every line myself, it feels good to go slowly for a change having used a lot of LLMs in the past year.
After having it on my TODO list for a long time, I've installed EndeavourOS on another m.2 drive on my desktop. With the advancements in gaming support on Linux over the last few years, over 80% of my daily tasks, work, and games are well-supported by Linux. I've been using Linux in VirtualBox or WSL for many years, but it's been a long time since I've ran Linux directly on hardware. I'm excited!
A civilization clone, but going way deeper than is reasonable (inspired by dwarf fortress). I'm currently building the geophysics simulation that will allow for realistic terrain generation, powered by actual mantle convection and plate tectonics. Once this is finished, then an actual per-cell atmospheric and oceanic model. This should keep me entertained for at least a year, at which point I can start working on the actual gameplay part of things.
I’m working on Recoil, memory safe, compiled static language (yeah, I know, everybody’s doing it) with Rebol syntax and built-in candies like parsing, finite state machine and rich syntax (that’s given, because of Rebol).
It’s nice to see how well-thought language design can pay off years later, with lower token usage. From entropy POV, Rebol syntax is certainly close to optimal state.
https://github.com/xemantic/markanywhere
Incremental Markdown parser that emits streams of semantic events, plus tools to manipulate them - designed for real-time rendering of streamed LLM output.
I'm working on a web app that animates hand drawn human characters. You try it here without any login or anything:
It doesn't use generative AI, instead it auto-rigs the drawings in just a few seconds.
A small launcher for Claude Code, to make switching between different providers and configurations easier:
For example, if I downgrade from Max to Pro I'd still be able to use the subscription, but also run sessions with other models (less expensive/local) as desired:
ccode init-config # initializes a new config file for me to set everything up
ccode edit-config # opens it in my editor so I can change, can also include editor as argument e.g. vim
ccode # launches whatever my default profile is
ccode --deepseek # Using their API key, they have a discount this month
ccode --openrouter # Whatever OpenRouter model I have configured in the config file
ccode --openrouter-preset # Also supports OpenRouter presets e.g. if I don't want to use quantized models
ccode --deepseek --control # launches a Remote Control session, shows up in web/desktop app as a regular session
ccode --deepseek --auto # overrides the default permissions, --yolo also works
... (and so on, there's more examples on the website)
I finally got around to signing app for Mac, which is what this post originally was about: https://news.ycombinator.com/item?id=48075366 (the new versions will be out soon)
Also thinking that I might make it an Anthropic API --> OpenAI API proxy that allows talking to providers that don't support the Anthropic API directly, alongside allowing switching models dynamically during a session (Claude Code wouldn't even have to know about it, it'd just send requests to a local endpoint and the proxy would do the rest).
Early on, but Go is lovely to work with, mdBook is great for getting a site off the ground and I'm really surprised that more people don't use Itch.io for distributing software (or the pay-what-you-want model in general), it's dead simple!
working on https://heylife.ai, an app that gives you proactive advices based on your calendar/taste/hobbies/location
let’s say you are arriving in paris. it will send you advice on how to get to the city from the airport. big soccer game in an hour? will send you advice on prepare it.
you don't need to ask, it will give you before / when you need it.
now working on the sandboxing and scheduling of the advices. releasing it this week if anyone want to give it a shot. (it will be paid only)