alt Hacker NewsAsk HN: What Are You Working On? (March 2026)
What are you working on? Any new ideas that you're thinking about?
Comments
I'm working on "context bonsai" which is currently a plugin for OpenCode that allows the LLM to self-edit its own context. It works like compaction, but it can retrieve back the compacted info if needed. And it's not just when the context is completely full, and it doesn't compact the entire context - it picks messages / tool calls where the details are no longer necessary, like a debugging session that is already solved or feature implementation that is complete and you've started on implementing the next feature.
I've also used tweakcc to make this work in Calude Code and plan to also do one for open source coding agents - codex, pi, Gemini, etc. And I'm also doing Livestreams of the development process.
Still working on:
A hobby project I started putting together late last year; a little spot on the internet for prayer and reflection. I've just shipped a small feature where you get a Bible reading (KJ only for now) in response to a prayer.
https://dugnad.stavanger-digital.no/
A pro bono tech consultancy for local (Stavanger, Norway) non profits. The idea is to help them use tech to better deliver on their mission. Last week I built a little bookmarklet for a non-profit to surface some of their data buried in a SaaS tool ... which will make their apple pressing operation easier.
I started working on Rio[0], a modern TypeScript HTTP client for Node.js 24+, Deno and Bun. It's built on the native `fetch` API and fully type-safe from request to response. It's still very much a work in progress but I'm working towards an alternative to fetch/Axios/ky I wish I had.
Nonograms! I built Nonodle[1], a daily nonogram puzzle game and I’m adding an option to access these puzzles from Nonoverse[2], my iOS nonogram app.
There is an API, and it’s a straightforward task, but one thing led to another and I’m also improving the app UI. The update will take some time but I hope it will only be better.
[1]: https://lab174.com/nonodle/
[2]: https://apps.apple.com/app/nonoverse-nonogram-puzzles/id6748...
Building CiteLLM, an API that extracts structured data from PDFs and returns exact source locations for every field.
There's also a widget so you can add click-to-verify to your own app in a few lines of code.
Click any value, jump straight to the highlighted source in the PDF.
Demo: https://citellm.com/demo
I'm working on a native code backtester that compiles pinescript strategies, and (hopefully) runs them super fast. Also a parameter optimiser with different scoring methods.
I'm working on an alternative solution to Ansible and Puppet, trying to mix the best of both world: The pull-based aspect of Puppet, by having an agent running on nodes, and the simplicity of Ansible, by writing playbooks and roles.
It's called Peekl, and is available on Github. Tho it's still in what I'd called "alpha". Lot of new features to come!
https://gitlab.com/usecaliper/caliper-python-sdk
An LLM observability SDK that let's you store pre and post request metadata with every call in as lightweight an SDK as possible.
Stores to S3 in batched JSON files, so can easily plug into existing tooling like DuckDB for analysis.
It's designed to answer questions like; "how do different user tiers of my services rate this two different models and three different systems prompts?". You can capture all the information required to answer this in the SDK and do some queries over the data to get the answers.
I didn't like any of the car maintenance apps I've tried, so I'm building my own. simpler, can be run locally (I run it in a k3s cluster). I don't think I have any stand-out features for it right now, or planned for the near future, but it works well for my purposes.
We have been homeschooling our kids. Homeschooling in India is not that widespread. So when a national newspaper covered our experiment, I got lot of questions around what we were doing. For a while I wrote blog posts answering them.
Now I've written quite a few posts (and given talks), I thought of writing a book. Just wrote two chapters. The draft lives here: https://www.jjude.com/books/hs/
AI assisted medical consultations
My first agentic app to dive head first into the AI world not to be left behind, oh boy this new world moves faster than I thought.
Feedback highly appreciated
Here is my fun mini-project:
https://apps.apple.com/us/app/kaien/id6759458971
I wanted a way for my kid to learn the alphabet, but without a UI that looks & behaves like a slot machine. It's all maximally slow, relaxed and designed to be easy to put down.
TermWatch, a terminal-native monitoring
Monitor your infrastructure from the terminal.
Integrated GitOps flow.
Define monitors in YAML, deploy from your terminal, get alerted when things break.
No dashboards to click through.
Install via CURL, dotnet package or manual download.
Currently moving Carolina Cloud to Kubernetes. I had built a custom orchestrator but really want the freedom of pod movement as well as KubeVirt's live migration capabilities. My ultimate plan is to open a second location in South Carolina at a cheaper colocation and then drain the nodes one by one, moving them to there, and leave my prior colocation. Kubernetes will make this possible.
Very much mvp but I just got this all set up: https://www.pginbox.dev/
Downloaded and parsed a bunch of the pgsql-hackers mailing list. Right now it’s just a pretty basic alternative display, but I have some ideas I want to explore around hybrid search and a few other things. The official site for the mailing list has a pretty clean thread display but the search features are basic so I’m trying to see how I can improve on that.
The repo is public too: https://github.com/jbonatakis/pginbox
I’ve mostly built it using blackbird [1] which I also built. It’s pretty neat having a tool you built build you something else.
I’m working on WC Price Hostory, a plugin that handles price tracking and Omnibus Directive compliance for WooCommerce.
It’s been available as a free tool for years, growing to over 45k active installs. I just rolled out the Pro extension to offer more advanced features, and the early traction has exceeded my expectations. If you're running e-commerce in Europe, this is a must-have for staying compliant with EU law.
Spent the last year expanding my homelab and now I have my own rack at my local DC with my own ASN and /23 prefix.
Its been pretty fun cosplaying as an network engineer, and now I'm building out an Anycast network for a few ideas that I'm working on.
Its nothing too revolutionary or new, but I'm proud that I've built them from ground up and all running on my own infrastructure.
- DNS Authoritative Hosting - https://thelittlehost.com/dns/ - Quietnet - A family-focused internet filter - https://quietnet.app
I'm also getting ready to launch https://relaye.io, which was my personal tool I built to support my devops consultancy.
Building a tool to make test reports useful - https://gaffer.sh
I work on a few products as an indie bootstrapper.
* https://theblue.social — TheBlue.social, provides Bluesky native tools
* https://stacknaut.com — Stacknaut, SaaS starter kit to build on a solid foundation with AI, includes provisioning on Hetzner, deployment with Kamal 2 and dev with coding agents
* https://codevetta.com — Codevetta, Architecture and code reviews service
* https://myog.social — MyOG.social, OG Image Generation Service
I've been planning a new idea with that and possibly future ideas based on the future (and near future) where there are more and more "agent" users.
I was interested in the idea of generating vector graphics with simple scripting language. The "simple" did not happen, but... I launched https://scriptdraw.com and I have lots of fun with it. My goal is to make the language much simpler than it actually is and then create a lot of generators (for example, gears!).
I'm building open source homebrewing (as in beer) software at https://www.brewdio.beer. It's something I've poked at periodically for a few years but now I'm using AI to see how far I can take it.
It has a few core libraries built in rust with a web app and a terminal UI. Android app is in the works. The persistence layer is intended to be offline first using a CRDT with an optional sync server. I'm also trying to integrate "bring your own AI" assistants to help tweak recipes or make suggestions.
It's been a fun way to sharpen my claude skills but also to see how feasible it is to maintain multiple frontend applications with a large amount of shared code. Still a lot to do, particularly the core calculations are not yet on par with existing offerings.
Been building https://textkit.dev/ for the past week.
It's a collection of 40 (and growing) tools for text processing, data cleaning, conversions, dev utils etc. Everything runs in the browser and is completely free.
Started this partly to learn SEO from scratch on a fresh domain, partly because i am lazy with regards to doing basic data cleaning using pandas and i found myself repeatedly using similar online tools that are completely riddled with ads.
I built this using Flask + Vanilla JS. I don't think there was any need to overcomplicate it. And for fun, i vibe coded a windows 95 desktop mode where all the tools open as draggable windows. https://textkit.dev/desktop
A bunch of ideas that have had domains but never enough engineers. Now there isn't enough time it seems except when I've hit my LLM subscription limits and they need to cool down.
Already launched biz-in-a-box.org and a life-in-a-box.org spinoff as frameworks to replace every entity's QuickBooks. I'm using them myself for every project my agents are spinning up.
Stealth project is related to classpass but for another category of need that won't go away even in the age of AI that really is only possible with critical mass of supply to meet existing demand. Super excited cus there's no better time to build with unlimited agents that scale without people problems.
Lastly, can't wait to run local LLMs so no longer limited by tokens/money.
Solving my issue with context switching when working with multiple coding agent sessions.
https://github.com/stefanhoelzl/codehydra
Always looking for feedback and ideas how others handle this!
Very small app for framing photos on Mac/ipad/iphone. I know some web alternatives. But when i wanted to add borders to photos I took to upload to IG i used imagemagick. I want to make something stylish(no tacky borderds), also usable for batch processing. I have prorotype and need to fix some issues.
Numenon: Sandbox knowledge base management tool for teams.
website: https://about.numenon.app
3D AI Modeling software intended for 3D printers.
Originally I made it for my grandpa, but I got a lot of interest so I made it into a full commercial product.
Just yesterday I published a set of 3 mini tutorials if you want to see how it works - https://youtube.com/playlist?list=PLKt1F5TvOjAHE07oBDlPXcrHc...
I’ve been training an alphazero style model for an abstract strategy game I created 20 years ago. It’s been really fun learning about MCTS and figuring out how to optimize all parts of the pipeline to be able to train on ~millions of moves for ~hundreds of dollars.
Lately I’ve been spending a lot of time transitioning from tech into urbanism and working on a few projects I care deeply about.
- Urbanism Now - I run https://urbanismnow.com, a weekly newsletter highlighting positive urbanism stories from around the world. It’s been exciting to see it grow and build an audience. I'm thinking of adding a jobs board soon that'll be built in astro.
- Open Library - I’ve been helping the Internet Archive migrate Open Library from web.py to FastAPI, improving performance and making the codebase easier for new contributors to work with.
- Publishing project - I’m also working on a book with Lab of Thought as the publisher, which has been a great opportunity to spend more time working with Typst.
These projects sit at the intersection of technology, cities, and knowledge sharing, exactly where I’m hoping to focus more of my time going forward.
Using a Muse EEG headset to read brain activity and use that to drive the output of a GAN. Similar to other projects that try to visualise or decode thoughts, but at the moment it's an art project. Obviosuly quite limited by compute and hardware. I'm sort of looking for collaborators / co-founders / opportunities in the AI + neuroscience + creativity space.
I've been on sabbatical (not on leave from anywhere, just decided to take a break from work) for months now, taking some time for myself. Minimal tech stuff until more recently, but now I'm back in the deep end.
The main thing I'm currently working on is a platform for organizing and discovering in-person events. Still not certain about the boundaries for "Phase 1", but I have a bunch of ideas in that space that I've been incubating for a while. One subset of features will be roughly similar to that app you've probably heard of that starts with 'M' and ends with 'p', but hopefully an improvement, at least for the right audience. But wait, there's more. :)
Currently building it; it's not public yet, so no link. Next month.
Thinking about how to grow the userbase is intimidating, but I think it might end up being fun.
I'm working on a computing system that would replace the Turing machine. In 2024, I published a space emulator in which computations can be performed using addressing. In 2025, I published a parallel addressing mechanism based on a sorting network.
I am working on two small apps for my dungeons & dragons group. We're playing inperson and I really like to give them printed out cards for magic items they receive and also for spells, because they are quite new to the game.
So I build these two app to create items and spell cards and print them out.
ClipNotebook: A FREE modern bookmark manager
- Visual cards with titles and thumbnails
- Playlists keep research focused
- One private URL for editing, share links for viewing
- Fast import for friends and teams
I'm never clear if this Ask HN is for posting about what you're messing with or for promoting organized projects that chase github stars or are commercial.
But anyway, I've started to learn Go. By doing a vertical scrolling shooter with embiten. Kinda like fitting a square peg into a round hole. No, it's not public and will probably never be.
Studying how do do a memory pool for actors, since it doesn't look like garbage collection and hundreds of short lived bullet objects will mix well.
I'm writing a book, which covers the mental models for writing code in a functional style. The examples are in Scala, but it will be useful if you use other modern languages like Rust, Kotlin, Swift, OCaml, or Typescript.
too many things in flight!
inspired by the karpathy/twitter posts on running (semi) autonomous research loops, I build https://github.com/tnguyen21/labrat to be able to try and replicate some paper results over night. still early stages but I'm getting some use out of it already.
also spending a lot of time thinking about how you "close the loop" on software projects. right now figuring out how you can combine static analysis + review heuristics to let LLMs course correct the codebase when they over-engineer or produced unwieldy abstractions.
I'm working on Rauversion https://github.com/rauversion/rauversion, an open platform for independent music communities that combines music publishing, events, and marketplace tools in a single place. Artists can upload tracks, albums, and playlists with metadata, audio processing (waveforms, analysis), and embeddable players with chunk-range loading to save bandwidth. It also includes ticketing for events (QR validation, Stripe payouts), streaming integrations (Twitch, Zoom, etc.), a magazine system for publishing articles, and a marketplace to sell music (digital or physical), gear, merch, and services. The goal is to give underground scenes a self-hosted infrastructure for releasing music, organizing events, and sustaining their communities.
i'm working on this two projects: https://github.com/shrec/Exorcist and https://github.com/shrec/UltrafastSecp256k1
I’m learning Godot and writing a basic game for my kids.
Coding agents are amazing and make me (feel) productive, but they really suck the fun out of programming.
I’m sure it’s possible to create a Godot-based game with an LLM, but I’m not sure how, so I’m forced to do everything the old-fashioned way – reading the docs.
I'm building a small tool called FormBeep[1] that sends a notification to your phone when someone submits a form on your website.
It started as a client problem, then something which I also experienced so decided to built it. It's just one small script and work seamlessly across platforms.
Continue improving my Chinese-character spelling game: https://store.steampowered.com/app/4218330/WordJoy/
I've been reworking my blog to have a table of contents per article, clean CSS (something that actually looks nice and no longer relies on Bootstrap) and a few other nice things. Also taking the opportunity to fix minor errors in previous posts.
Aside from that, I need to document and properly release one of the pieces that PAPER is relying on (some generic tree-processing code that makes operations on directory trees a lot nicer than with the standard library "walk"s), and work on others (in particular, a "bytecode archive" format for Python that speeds up imports for some projects, mainly by avoiding filesystem work at import time — I want to offer it as an install-time option in PAPER, and later have `bbbb` make wheels with the bytecode precompiled that way).
I just launched bookcall.io publicly last week. Think calendly that treats your scheduling page more like a sales funnel. Very important if one call can make you a bunch of money. Page builder, brand assets, videos, documents etc. attachable. Forms, video calls, everything included.
Also launching a supabase security scanner. If someone wants a free scan hit me up. Includes POCs and verification before and after remidiation. Goodbye false positives.
Building a tool that automatically generates living infrastructure diagrams from your cloud anccount and turns them into real-time incident dashboards. Think Figma meets Datadog - beautiful visualization that updates during outages to show you exactly what's failing and how to fix it.
I'm teaching a class in agent development at a university. First assignment is in and I'm writing a human-in-the-loop grader for my TAs to use that's built on top of Claude Agent SDK.
Phase 1: Download the student's code from their submitted github repo URL and run a series of extractions defined as skills. Did they include a README.md? What few-shot examples they provided in their prompt? Save all of it to a JSON blob.
Phase 2: Generate a series of probe queries for their agent based on it's system prompt and run the agent locally testing it with the probes. Save the queries and results to the JSON blob.
Phase 3: For anything subjective, surface the extraction/results to the grader (TA), ask them to grade them 1-5.
The final rubric is 50% objective and 50% subjective but it's all driven by the agent.
I am working on Kastanj. It aims to make cooking as foolproof as it can get. Anyone should be able to cook any recipe and get it right on the first try. Clear step by step images and instructions for everything etc.
It also features a recipe manager with family/friends sync. This makes it possible to upload your grandmother’s cookbook and share them with your whole family.
I’m building an observability system that tries to surface answers instead of making people dig through huge amounts of raw telemetry.
The basic idea is that when one failure fans out across 20 services, you often end up with 20 alerts and 20 separate investigations, even though there is really just one root cause. I’m using distributed tracing to build a live model of how errors propagate through the system, and then exposing that context directly at each affected service.
Longer term, I want this to become a very high-precision RCA engine. Right now I’m looking to try it with a few early design partners that already have a lot of tracing data, especially OpenTelemetry or Datadog APM users. I'll love to chat with some folks who would be willing to try it out!
"Does a launch make any impact if there's no audience?"
We've found most early-stage startups ignore social media until after a launch. Things like “$0 spent on ads” sound cool, but they don’t help if no one knows your product exists.
I'm building Appents to provide a done-for-you social media solution for startups.
Would love feedback: https://appents.com/
I've been working on an MQTT Broker/Topic tree explorer. It's intended to help someone understand what kinds of data are moving through their broker, what the busiest data is, and just generally be pretty. https://ryanbateman.github.io/mqtt_vis/
OSS, MIT licensed. Feedback welcomed!