alt Hacker NewsAsk HN: What are you working on? (February 2026)
What are you working on? Any new ideas that you're thinking about?
Comments
https://sfbapt.com/routes.html
Lots of work left to do, but happy to have a working version up. It's an interactive map that currently shows all the routes and stops for SF Muni, BART, Caltrain, samTrans, and VTA. There are many more agencies (official and unofficial) in the bay, so I'll be adding those throughout the next few days as I sort out the data.
Finding the data and cleaning/normalizing it is a real pain, so if anyone knows a good place to find them (and normalize them), please do share
I'm working on a macOS app that identifies music playing around you, constantly. You can scrobble them to Last.fm, ListenBrainz, LB-compatible sites, a Mastodon account, local Shortcuts workflows, and/or or local files. It also support Apple Music and Deezer.
I built it because most scrobblers only keep track of Apple Music or Spotify plays, but I have streaming stations on all day long, or the radio is on, and those never make it into my playing charts or recommendations etc.
Early days, more to come, public beta is ongoing, and I am looking for testers :)
Built myself a silly little menubar pomodoro timer tamagotchi thing for mac. I’ve been slowly going through and building highly personalized versions of my day to day apps. This is the first one I polished up enough to share. Free if anyone’s interested. https://apps.apple.com/us/app/time-flies-focus-timer/id67582...
I spent about a week, writing this[0].
It’s a special UIKit map package, designed to be integrated into SwiftUI, as map support is a big SwiftUI Achilles heel.
I have integrated it into another package (a SwiftUI admin tool, that isn’t public), and it works exactly as I planned.
I often do this kind of thing. If I can break some module out of my work, and publish it, I spend the extra time, polishing and documenting it. No one else really cares, but it forces me to do a really good job, so I get an extremely high-Quality component, that I don’t have to worry about, later.
A metacritic like website but for any product.
It analyzes thousands of professional critic reviews to find the best of the best.
I started building this because I adore how metacritic analyzes professional movie/game/tv show reviews and calculates a meta score for each title. In my experience it’s the best way to discover new things to watch or play, and I’ve often wished something like this would exist for when I want to buy a product.
This year, I decided to start building it myself and Criticaster is the result.
For a given product category we collect all professional reviews of a given product, analyze each to assign them a score and then calculate an average critic score.
The goal is to become the most trustworthy source to make product decisions.
Very curious what y’all think!
I am building HNSignals: HNSignals transforms Hacker News comment discussions into executive-level intelligence briefs.
It synthesizes comments into structured reports in a Chief-of-Staff style for tech leaders.
After 15+ years in web development — now diving into game development with Three.js / React Three Fiber (R3F). Keeping AI usage minimal where possible, but it’s been invaluable for complex geometry and math-heavy problems.
Game idea: DroneCraft is a third-person drone exploration game where players scout the world for parts, craft powerful upgrades, and trade strategically to evolve their build.
Whats coming: Core mechanics are up and running. First playable version planned within a month, alongside open-sourcing the full codebase.
I'm working on Kanji Palace (https://kanjipalace.com). It's a web app that help people learn Japanese language with AI.
Right now, I'm working on the OpenClaw-like feature. So, you can learn Japanese via Telegram. Keep track your progress. Practice conversation with your AI assistant. Etc.
Been dedicating some free time from my weekends to sending PRs the DarkRadiant project: https://github.com/codereader/DarkRadiant
DarkRadiant is open source, cross platform level editor for The Dark Mod and other idTech4 games like Doom 3, Quake 4 and Prey. It is a lot of fun to add features that make level designers more productive. I would like to invite anyone that is curious about wxWidgets, CSG and game dev to join us.
If you are a designer, we could definitely use a new icon set.
Semi-academic attempt to build semantic-level standard system for data type classification. Useful for metadata semantics data lakes / warehouses beyond basic physical level like text/number, with assumption that it may be useful for AI-targeted metadata, your text to sql cases. https://github.com/jaakla/semantic-field-types
I'm building Banger, a work email platform for builders (service + apps). It helps you to:
-> host your own custom domains and plug existing email sources with privacy;
-> share mailboxes with teammates and AI agents without having to share pwds as in the 80's;
-> git review like UX to collaborate with humans and AI agents; local AI to save money on mundane email tasks;
-> native apps + cli/mcp/api with guardrails for external AI agent access
waitlist on https://bangermail.com
A chant synchronization app for large protests and events https://wordsunite.us
I built a website where kids can practice reading comprehension and learn new words while staying up to date with the latest global news each day. I originally built it for my own kids to help them maintain their reading skills while we travel. They loved it—and even suggested adding gamified features like streaks and badges!
I am trying to make a cheaper, straightforward easier to use observability platform with a good core offering without going overboard on features. Based in Europe.
I'm currently working on a RethinkDNS-like (Android) and DNSCrypt-Proxy-like app built with Tauri + Svelte. It will include DNS blocklists, a custom WireGuard proxy, and potentially cross-platform device support. I'm using Mullvad's GotaTun implementation. I wanted to learn more about these networking concepts. If I finish it, I'll open-source it.
Small lenticular holograms of math equations: https://gods.art/store.html
Building, "Workloom ( https://workloom.so )" - a free-to-use project management tool for freelancers and small teams.
What's Included: - Kanban boards - Time tracking - Documents - Team chat - Schedules - Analytics - Voice notes - File storage - Task templates - Client reports
Cross-platform book management and reading application: https://github.com/Eoic/Papyrus.
This isn't a serious project by any means, but rather a prototype. Also, I really got carried away using Claude Code, although my initial goal was simply to glue a quick proof-of-concept and see how it could look like.
I am researching go-string-concat-benchmark [1]:
A performance comparison of four common Go string building methods.
I recently updated my go-stats-calculator to include many more stats [2]:
CLI tool for computing statistics (mean, median, variance, std-dev, skewness, etc.) from files or standard input.
I also created claude-image-renamer [3]:
AI-powered image renaming script that generates descriptive filenames for screenshots.
[1] https://github.com/jftuga/go-string-concat-benchmark
I'm still working on my Web Server Library .NET Core, the rewrite is almost finished (beta version)
The past few weeks I've been building Blackbird
https://github.com/jbonatakis/blackbird
At a high level it's my take on how the execution aspect of spec-driven development should be handled. Where as most tools that are popular right now break a spec down into a task list and instruct your agent to work through it in a single session, I am treating agents as stateless. By this I mean a separate (headless) session is started with selected context for each task. This avoids context exhaustion, compaction (and the resulting confusion that can occur), and means that Blackbird can work through effectively an arbitrarily large task list.
Right now it's BYO-spec, but then it:
* breaks the spec down into a dependent-aware plan (DAG) composed of parent and child tasks
* executes tasks one at a time based on their status (ready to execute if all dependencies are marked as completed)
* allows you to (optionally) pause execution after each task to review, approve and continue, approve and quit, or reject the changes altogether
* (soon) treats parent tasks as an automated reviewer for all child tasks and optionally auto-resume those sessions to address the feedback
* and more
It's entirely bootstrapped, and so far I'm quite pleased with it. I also wrote a post[1] today about some of the concepts I had in mind as I was defining the architecture.
Month 2 of building the SQL client I've always wished I had.
One feature I'm especially proud of is the visual query builder. Drag & drop to build SQL queries.
There's also an entire SQL tutorial section for anyone who wants to learn or refresh SQL knowledge.
Been working on TenantSaas, a .NET library to make developing multi-tenant apps safer. Wanted something that prevents background jobs or admin scripts from accidentally running across tenants by refusing to run when tenant context isn’t clear. Comes with contract tests teams can run in CI. Still early, so be gentle.
Struggled to find the best priced meat from UK butchers keeping up-to-date in my spreadsheet so built a comparison site with multi platform scraper (and a taxonomy matcher to allow “apples” to “meaty apples” comparisons).
UK only for now, and very much a “solves my problem” side project, but easily scalable to other countries of the need is there!
A free interactive SQL tutorial that will get to the level of being a data analyst. It's Alice in Wonderland themed.
I used to be a coding bootcamp instructor, TA and guest lecturer. I've noticed more and more people need to learn SQL for various different reasons. I'm mostly concerned about lesson scaffolding since most SQL courses don't do it that well. I'm hyped about AI but they're not great with lesson scaffolding.
I'm 33% to 50% done. I've already noticed the way I scaffold the lessons is unconventional. For example, for the first 50%, I don't want students to know what tables are. It's too much all at once, everything should be small bites before the big concepts get introduced.
If anyone is interested in testing the beta version, let me know. It will be up within the next 2 weeks probably. My email is in my profile.
I an building Bloomberry, an alternative to tools like Wappalyzer.
It helps sales teams find companies that use any product/technology, but not limited to frontend tecnologies.
So for instance, you can find companies that use Canva: https://bloomberry.com/data/canva/ or Cursor: https://bloomberry.com/data/cursor/
I have been working on "scratch milling"[0] PCB's at home using my vinyl cutter/plotter and a engraving bit.
Inspired by Robin Debreuil his process and videos (see video the article and several forum posts). He is using a CNC but I figured regulating pressure is more important then depth, therefore my experiments with the plotter.
Currently dialing in the pressure/speed and amount of passes on a single layer copper board.
[0] https://hackaday.com/2020/07/10/making-pcbs-the-easy-way/
I'm going to do the hyper-literal take of what are you working on literally today, since I'm always working on the same old project Marginalia Search and I have been for going on five years now.
* Integrating website liveness data into the crawler to make more informed decisions about whether to keep or wipe data from a website if it can't be reached while crawling
* Working out why the liveness data gathering process isn't stopping as scheduled.
* Deploying a voluntary max-charge request header for the commercial API
* Making website URL elements searchable. They should be already, but are not for some reason.
* Maybe looking into an intermittent stacktrace I get on one of the index partitions.
No blockers.
I started working on https://ottex.ai three months ago just for fun to test Gemini 3 Flash as a voice-to-text model (by the way, it’s amazing, but a bit slow). However, it quickly transformed into my main project.
It's a free macOS app written in Swift that allows you to type with your voice. It supports local models and BYOK (Bring Your Own Key) with a bunch of providers.
You can assign different models and post-processing steps to polish the text. For example, I have a setup for Obsidian that transforms my voice into clean, formatted Markdown. Or, when I use it inside VS Code, it switches to the Parakeet V3 instant local model.
Hi HN,
I’m working on an API that provides heuristic signals about whether a piece of text is likely LLM-generated.
The system uses an ensemble of techniques, and on my internal evaluation set (details below) it achieves ~99.8% accuracy. I’m fully aware that general-purpose AI text detection is hard and adversarial, so this is meant as a probabilistic signal rather than a definitive classifier.
There’s a simple demo site: https://checktica.com
And public API docs (no API keys required): https://api.checktica.com/v1/docs
I’d really appreciate critical feedback, especially on:
- Failure modes you’ve seen in similar systems
- Whether this framing makes sense at all
- Where such a tool might actually be useful (or useless)
Happy to answer technical questions.
I left my job as a PM a couple of years ago to start acquiring small e-commerce brands that sell on Amazon. I'm currently running those, and mid-acquisition on one.
Because they're relatively low-effort (Amazon is terrible for sellers in many ways but man do they provide an incredible amount of infrastructure), that leaves me plenty of time to play with AI, and it just so happens that the business serves as a giant, practical eval as new models come out.
I've been vibe coding apps for internal use and using Nano Banana for listing images and whitebox photos, and more recently I've started to lean on Claude Code heavily as an assistant. It's got API creds for my Amazon account, so I use it for everything from figuring out when I need to reorder to filling out spreadsheets for companies that safety test my product.
And of course I am writing a Substack that I must shamelessly self promote that goes into the practical use cases of AI in my business: https://theautomatedoperator.substack.com/
Flora Carta - https://floracarta.com
A tool to "Design and keep track of gardens, orchards, and landscape projects."
It's very alpha right now, so its quite ugly, but people are already using it. I'm committed to releasing early and often.
You can see some people's designs here: https://floracarta.com/browse
I'm working to figure out new auxetic geometries for 3D lattices. The arrowhead is cool and simple, and gyroids are very effective, but I'm trying to discover if there's something simple, printable, and maximally effective. Tough problem. There's no general theory for auxetic lattices, so it's a matter of reasoning from the desired mechanism to find patterns that fit, almost like alchemical trial-and-error.
I’m building an app that facilitates discovery and eases payments for roadside stands that sell produce, honey, maple syrup, eggs, firewood, crafts, etc. The concept is that any roadside vendor can sign up for free (forever, no add-ons or upsells) and they have an online home for their home business. The vendor can list up to 3 stands and show off the products they sell in each stand. Users can discover stands near them by list, search, or map, view the vendor and stand details, ratings, payment methods accepted, etc. When arriving at a stand the user can scan a QR code which opens a web cart, allowing them to add products they are going to purchase and then “check out” using one of the vendor’s stated payment methods like Venmo, CashApp, PayPal, Apple Cash, Zelle, or good old hard currency. We make these payments easier by standardizing the check out experience but we do not facilitate payments at all - these stands have always been and will continue to be self-serve on the honor system. Once you’ve paid, you get a receipt and take your goods. The vendor gets an alert that a sale intent was started and by which method so they know where to look for their revenue. In the future we may help with some basic reporting and very light inventory management if vendors ask for it. We allow users to alert the vendor if a stand is out of stock, which is also reflected in search so other users are informed as well. Users can then ask to receive re-stock alerts as the vendor restocks. Then of course users can favorite stands and products, share them, rate them, and create shareable collections of stands they curate (The Honey Trail or Summer Sweet Corn All-Stars, etc.). Eventually we will be adapted for events like farmer’s markets, craft fairs, and christmas markets. I built this because I am a maple syrup producer (tapping starts in a few short weeks from now) and I’m starting to get into mass sales of my syrup. I felt like people who produce and sell these products put a lot of hard work into the process and deserve a legit discovery tool as well as a basic stand management system that does not make them change their process or get in their way. An app like this costs basically nothing to run and I will ensure it is free to use as long as I am in charge. I’m testing this week and likely soft-launching in the next couple weeks - the goal is to be online around March 1. It was just going to be web-only (Supabase with a Svelte front end) but after Claude put me in timeout last week I tried Antigravity and now have 80% of an iOS app and will scaffold my Android app in the next month - so native apps will follow a web release pretty quickly.
Just tried out Opus 4.6 to make a ground-up new version of a perennial side project: static site podcast player.
For this one I focused on loading speed and reducing interaction with repo. So it processes the images (converting to webp) and loads the feed list from a Gist. Also used the "frontend-design" skill. From brief to ready-to-use took about a couple hours.
An exi decoder/encoder (goal is to have modes for spec conform and interop, which right now doesn't seem to be the same thing). Afterwards I also want to try to use it to encode huge XML datasets in precomp mode with good encoder (maybe ztsd). Should be pretty useful for large repetitive datasets. I also want to build a tool to visualize XML to exi de/encoding in the browser.
An on-call runbook execution engine - being able to take plain text runbook steps like - look at logs for $service, check for dependency failures - look at so-and-so dashboard.
and execute them when an alert fires.
We are at https://www.relvy.ai
I have been working on chemistry arena designed to benchmark the current SoTA LLM's on drug discovery tasks. My plan was to then focus on getting this annotated to sell this data for post training of scientific reasoning models. https://github.com/deepakorani/chemistry-arena.
Also trying to find a co-founder who I can work on projects and solve problems faster to be honest.
Working on Postkit - auth, permissions, config, metering, and job queues as pure SQL functions inside Postgres.
I've been using Claude Code to spin up apps quickly, and I kept needing the same infrastructure every time - user auth, permissions, usage tracking, job queues. So I pulled it all into one SQL package that lives in Postgres. Now when I start a new app I just tell Claude to use Postkit and all that stuff is already there, no external services to set up. I can focus on the actual product and iterate fast.
It was also a good excuse to actually use stuff I'd studied for system design interviews - Zanzibar-style ReBAC for permissions, a double-entry ledger for usage metering, transactional job queues with SKIP LOCKED. ~15k lines of SQL across five modules, with a Python SDK. The SQL works from any language though.
With LLMs, the cost structure is credits/tokens all the way down. This month, I built a tool to help you price your service credits or plans at https://aitokenpricing.com.
I also did a write up of PRDs and using Railway to build apps on my blog at https://aftercompsci.com.
I just released Configmesh this week. It's a macOS app (with CLI companion) for e2e encrypted syncing and backing up of dotfiles and application configurations. You can sync for example stuff from ~/.config/, Application Support, *.plists, and so on, and add config sync to apps that don't support it natively
Fresh off the press
Personally:
- The Laravel Artisan Cheatsheet - https://artisan.page
- Cachet, the open source status page system - https://cachethq.io
Professionally:
- Laravel Cloud's "Private Cloud" offering - https://cloud.laravel.com/enterprise
Maintain my blog
All kinds of things! I work with AI every day to do various kinds of work. Coding. Research. Brainstorming. I write up notes nearly every day and then I post a summary of each week on Patreon. https://www.patreon.com/cw/aiconfessions
One of the projects that features in these notes is the attempt to build a programming language using AI. https://github.com/xvandervort/graphoid
Since I left my last job, I do a lot of writing. I also have a couple substacks. One is a humorous weekly look at science and tech (https://technoscreed.substack.com/ ) and the other is a monthly exploration of history (https://historyroad.substack.com/)
Building https://lenzy.ai - helping conversational AI products (think Lovable or Cursor) reduce churn and prioritize product improvements by analyzing their user's chats. I started about 4 months ago, made my 2 paying customers happy. Now trying to onboard more and more companies!
I finally got to rework my html form forwarder[1]. I realized how cheap SES is and thought I might as well make the service free.
I hope open sourcing it could be a way for me to get some mentorship from more experienced devs as well. Unfortunately, my work doesn't really do code reviews so I feel like I am not improving much on that front.
Building Pasture (https://www.usepasture.com)
Pasture takes each signup, enriches it (title, company size, funding, tech stack, and more), and scores it 0-100 against your ICP. Alerts go to Slack with full context. You can also track which channels bring quality vs. junk over time, which has been the most useful part so far.
I’m building lightweight-css to teach kids aged 8+ the fundamentals of programming using the web and JavaScript.
Link to Github: https://github.com/joshuamabina/lightweight-css