I'm building my own cloud (I actually typed claude instead of cloud there... wow). There's no IaaS or PaaS; it's much simpler. I wanted my own way of connecting to machines and the TCP services on those machines without having to install Tailscale (not allowed on a locked-down corporate PC) or pay for Azure or AWS or GCP or even Hetzner or Linode. I've got 10gbps fibre and a huge workstation at home, and I've got lots of laptops and VMs and other outboard stuff that I want to work in concert with that workstation, so I started building something I call Tela (Filipino for fabric; I was sitting in Ninoy Aquino International Airport waiting for a flight when I had the idea, and it's implemented as a network fabric).
Working on an audio streaming platform powering an indy internet radio. Looks like Icecast & friends show its age and a similar product can be easily built with the functionality cast down to simply robust streaming & handling "timed playlists". I enjoy every bit of knowing exactly what happens in the code. It's not open source atm, but will be. It's in Go, is a pleasure to write and the deployment takes minimal amount of resources.
Other project is to continue a bit stalled progress of a configuration language BCL - add functions, more structures and fix some hidden scoping issues. Making languages is an endless fun.
https://github.com/wkhere/bcl
My friend and I are building Strandfall, a highly physical outdoor larp (live action role playing game) that uses custom spatial computers: https://strandfall.com
Players are survivors of a global disaster that has unleashed mysterious, deadly storms. For three hours, they investigate the origin of the storms and make fateful decisions about their future as individuals and as a community.
We received Immersive Arts funding, which means we can run it in Edinburgh later this year. Here's an excerpt from our 2025 grant application about exactly what those are:
--
Our “storm sensors” are novel spatial computers designed for outdoor usage over long distances. They will house ePaper displays, LoRa (long range) radios, accelerometers, gyroscopes, and GPS chips in a 3D-printed enclosure to provide a low-tech way to augment the reality of the park. These computers will be cheaper, more rugged, longer-lived, and more capable than smartphones, deployable to locations with zero cellular service and no battery charging options.
The sensors will be mounted on top of camera tripods for deployment. Runners will carry them through the park, then position and aim them in the correct direction, as co-ordinated by “operators” using walkie talkies. This will let players feel like they are really setting up important equipment, scanning historical sites for clues (like surveyors), and establishing laser communication links. Lacking colourful touchscreens, the sensors will be less distracting for runners, helping them focus on their surroundings. Essentially, they are a highly tactile and deeply realistic way of immersing players in a post-apocalyptic setting, since such devices – not smartphones – are the most likely to be used.
The idea came after I finished a permanent piece for a museum using MaxMsp and python. I always had this thought in the back of my mind that "I could express this so much easier in a few lines of code.."
JDBC does not allow pipelining (a Postgres only feature).
It can reduce the number of db round-trips a lot, especially when using Supabase+RLS (or other systems that require frequent setting of configuration values that are basically fire-and-forget).
Meet Bpdbi, a library with first-class pipelining, which provides a Postgres db driver (that's binary only, as the legacy text-based protocol is no longer needed, it just takes up space) and exposes an API that's more close to Jdbi's that to JDBC's (developer friendly).
So great to see how much stuff there is going on outside of AI!
I've been working (on and off for over a decade) on a way to manage my unwieldy photo and video collection. I experimented with so many tools before but ultimately nothing ticked all my boxes so I wrote my own. Admittedly lots of niche stuff like proper support for UTC time, stereo photos and partial dates and things like 100% offline face detection and custom tag hierarchies and querying without a database. Each time I add a new feature I'm having a blast :)
I’ve been working on an open-source tool to help visualize JavaScript/TypeScript codebases (React, Next.js, Node.js). The core idea is building a dependency graph at the function/component level and exploring relationships across the codebase.
One feature that turned out to be particularly useful is what I think of as “blast radius”, If you change a function or component, what other parts of the system are affected?
This has been especially helpful when thinking about PR reviews, not just what changed, but what the change might impact indirectly.
It’s still early, and I am working on the cloud version which will include LLM interface to interact with graph and other team collaboration feature and github connection but I’d really appreciate feedback from others who might find this tool useful.
I literally said this exact thing verbatim with someone who was worshipping ai for code. This is an insane idea, really cool.
One bit of feedback, I'd love if this could be a little more aesthetic in the visualization. The method names are a little hard to read, white on light green.
It was just the first look...My primary focus on was on the Engine. I will upgrade the UI and will surely change the UI you mentioned. You can also contribute if you wish as it is open source (built in nextjs)
A bunch of square panels with a grid pattern to mount hold on, with the panels hanging on french cleats (with a locking system, #TODO) so the panels are easily removable so I can hang something like planters on the wall as well with the same french cleats.
No AI, a bit of computers to draw things out in CAD, but otherwise just manual building stuff.
Question, do you "build LEGO" as in "make what's on the box it came in", or do you use LEGO to build things yourself and not following instructions?
(I am old, but as a kid, I remember building what the kit was supposed to build once or _maybe_ twice, but using the parts from the kit and various other kits I had to make things purely out of my head pretty mush all the time. I _think_ that's not how people "play with LEGO" these days?)
Both, I enjoy the process of building, so following instructions. But then it's done and boring. I.e. LEGO supercars sets are nice, they're so nice until you marriage chassis with drivetrain - you can play with all the gears and see them, after the marriage it's just boring. So now I'm working more towards reusing parts and building things of my own.
Edit: LEGO Architecture Set (not sure if they still make it) is super nice for building random stuff - it's a set of pure white bricks, super fun.
I recently built Cranki [0], a free little PWA that generates crosswords using your Anki flashcard lists. It's aimed at language learners (who find flashcards boring and crossword fun!). To be honest I built it just for me and then decided it might be useful for others.
It's all local, no server, no database, etc. Mobile and desktop friendly.
I did build it with the help of AI, but no AI inside the actual thing.
I'm working on ghidra-delinker-extension [1], which is a relocatable object file exporter for Ghidra.
The algorithms needed to slice up a Ghidra database into relocatable sections, and especially to recover relocations through analysis are really tricky to get right. My MIPS analyzer in particular is an eldritch horror due to several factors combining into a huge mess (branch delay slots, split HI16/LO16 relocations, code flow analysis, register graph dependency...).
The entire endeavor requires an unusual level of exacting precision to work and will produce some really exotic undefined behavior when it fails, but when it works you feel like a mechanic in a Mad Max universe, stripping programs for parts and building unholy chimeras from them, some examples I've linked in the readme. It has also led to a poster presentation to the SURE workshop at ACM CCS 2025 in Taiwan as a hobbyist, an absolutely insane story.
I'm building a good intermittent fasting app for iOS using SwiftUI and native Apple APIs for the most "Apple" experience possible. The app is 100% free and without ads.
still working on https://stringscales.com - fun sideproject to visualize guitar scales on a configurable fretboard, with interactive note highlighting to a backingtrack.
The backingtrack is what I'm actively improving right now. It's just a pad running now, but it will turn into a full track with bass/drums/piano/... and will feature a comprehensive chords based editor so you can add and save your own progressions with a logged in account.
I’m reimplementing the Nix parser. I don’t know if it’ll morph into another formatter, or an LSP server, or what. I guess I’d like to try and improve on error messages and build it from scratch. Rust has really brought ergonomics to systems programming, and I wrote very few parsers for real languages, and never one with good errors.
I'm going with my fourth (or fifth) attempt to create a digital twin of my apartment in a game engine.
This time it's in Godot and i want to add more interactive stuff in the virutal version, rather than just redraw the rooms with some lights synced with Home Assistant.
Also working on a Korean guide, so far it introduces Hangul (한글), the Korean alphabet: https://tolearnkorean.com/
Also making an app / web app for memorizing Korean words, takes inspiration from Anki and Duolingo. Words go through 4 stages: 1) matching up pairs 2) Multiple choice answer 3) Writing word through blocks 4) Free-form writing.
I've been building the games I've always wanted to play. It used to take months to make something like Medieval Building Simulator.
Now you tell some idea of what you want and AI gives you a better thing than what you originally imagined. Then you go into a rabbit hole... simulate productivity, sales, customer, cash flow, materials for the building, make it curvy, make it pop up in 3D, populate it with NPCs, give your NPCs wages and background, write a dialogue with the ghosts of Christmas. It ends up being a game in itself.
It's good fun. It costs me $20 and I can do it and deploy it from my phone.
A crossword puzzle generator, just for fun. Grid generator in Python because it's easy to hack around and grid generation doesn't need to be particularly fast. Go for populating the grid with words because with large grids there are combinatorial explosions, and Go's speed is beneficial.
Not source-available yet because it's a bunch of hacks (particularly the Python) but maybe one day.
I'm scratching my own itch: building a yet another audio signal generator for smartphone. I need some extra functionality that is not available elsewhere, and impose some limitations "just because": it must be bare minimal PWA.
But actual app does not matter, the main take away for me is: it is easy and fast to write bloatware (esp. with AI), but not that easy to distill to what is really needed. And what looked like a weekend project, a couple of hours max (with help of AI), now lingers for 2+ weeks on-and-off on evenings (of manual effort).
I am building a web application for learning math. I want it to be something between Khan Academy and Math Academy. Here is a demo of fourth grade https://demo.numerikos.com/ Currently the best part about it is one of my kids is using it. I have some more lessons ready, waiting to be released and I am currently working on Trigonometry.
I started looking into AT Protocol recently and find it very interesting, so I started collecting a list of decentralized products built on top of AT that are alternatives to mainstream (popular) products.
One question, I tried sharing your site with a friend and the share URL is coming up as http://localhost/ which is strange and I've never seen before. Is this something that's possible to customise via meta tag, or is this a bug in my browser (Vanadium 147 on GrapheneOS)?
I've been working on a super-straightforward volunteering platform -- https://handsup.barryvan.com.au/ . Mainly because I got frustrated with my kids' school and sailing club activities being (mis)managed across WhatsApp, Google Forms, pieces of paper, and the like.
I'm working on identifying mental disorders from speech patterns. The classifier itself is not really the point, though, but rather finding specific disruptions that we can correlate with areas of the brain and, eventually, use to identify their root causes.
I'm currently working on a small tool that helps me setup new personal computers in a purely descriptive way.
A little bit like ansible, but then totally not like ansible.
It generates SSH-Keys. It clones repositories, installs uv and rust. It removes Snap from my Ubuntu machines and installs firefox from the mozilla repositories.
MusicLibrarian, a macOS app that deep-cleans Apple Music libraries. iCloud sync duplicated my playlists up to 540x (77,542 ghost entries). Built it in Swift, uses ITunesLibrary framework + AppleScript. https://thevaultdj.com
I'm building the CAD app I have always wanted. Free, Parametric, STEP export for 3d printing. The input is just a javascript editor. It uses OpenCascade under the hood. Lots of fun.
i’m currently avoiding committing/releasing a bunch of changes i did last week because people are actually using the library already (the curse of writing something useful lol)
I am building something similar to Windows Task Manager but it also works on Linux, launches/monitors web servers, and provides a visual dashboard for docker containers.
Took some good ideas of Pascal and making it more modern. Minimal runtime, manual memory management, single (small) executable, no dependencies. Compiler itself is written in Swift and I am using QBE as a backend ATM.
I recently made a 'notification-driven RPC' app for Android to help me help others:
Notifactor (Android App)
A lightweight native Android app that intercepts notifications on the device and triggers actions based on configurable rules.
The app uses Android's NotificationListenerService API. Once granted notification access it receives a callback for every notification posted on the device. It then checks each notification against configured rules and runs the configured action.
Why would you want this?
I don't know why you would want this but I can tell you why I made it: to make it easier to control some functions on Android devices used by people I often help using them in some way. My mother's Android TV (which I use to communicate with her through Linphone), phone and tablet sometimes stop doing the right thing. I live about 1300 km to the north of where she lives so I can't just hop on my bike to fix things. Thus far I relied on a set of Termux scripts on these devices to keep a reverse ssh tunnel open to an endpoint on my server but this has a number of drawbacks: the tunnel is not always there when I need it due to WiFi dropouts and other similar problems and the constant connection uses battery power on the phone and tablet. If only I could cause the tunnel to be created when I need it and brought down when it is not needed... Well, that is possible using Notifactor by sending a notification on a specific channel (ntfy refers to these as topics) whereupon Notifactor runs a Termux script which manages the tunnel (etc.).
Since the moment "AI" became a thing, I've been rolling my eyes and looking the other way since I know it's just a half-assed hype buzzword. With that in mind, pretty much everything I work on is not AI related. As a hobby(and to a degree to be prepared for WW3), building a ton of drones and drone equipment, flying them and finding more and more ways to push their range to the absolute limit.
I don't think it qualifies as AI in the modern day and age but NLP in general. It's truly amazing how easy it is to spot troll farms online and no one is doing anything about it, be it individuals, private sector or even on national level, given that those should be considered a risk for national security.
This seems a little bit of a duplicate of [1], but I can repeat my answer here as any views help!
My wife and I continue to work on Uruky, a EU-based Kagi alternative [2].
Since last month we finally got our production API Key for EUSP/STAAN (it was certainly the slowest and most complicated search provider to adopt, so far), and that brought us to 5 search providers you can choose from and sort as you prefer.
We already have got over 40 paying customers (excluding family and friends, we’re guessing these paying customers came from some privacy listings and HN comments) and have exited beta last month!
Customers seem to really enjoy the simple UI (search can be used without JS) and search personalization (from choosing the providers to the domain boosting and exclusion). We also have hashbangs (like "!g", "!d", or “!e”) when something doesn’t quite give you what you’d expect, though.
You can see the main differences between Kagi and Uruky in the linked page, but one huge difference is that with Uruky, after being a paying customer for 12 months, you get a copy of the source code!
One thing we’re struggling with is outreach because we want to do it ethically, and it’s hard to find communities or places to sponsor which are privacy-focused and don’t require €5k+ deals. Ideas are welcome!
Because of bots there isn’t a free trial easily available, but if you’re a human and you’d like to try it for a couple of days for free, reach out with your account number and we’ll set that up!
Thanks.
P.S.: Because people have asked before, our tech stack is intentionally very "boring" (as in, it generates and serves the HTML + bits of JS to enhance settings and such), using Deno in the backend (for easier TypeScript), PostgreSQL for the DB, and Docker for easier deploying.
P.P.S.: Because this has been also brought up before, the name has no special meaning but we read it like "Euro-key" in English. Names are hard, and we’re aware it can remind people of Uruk and Uruk-hai. That’s OK.
P.P.P.S.: Another frequent question here is “how does it work?” When you search, we query the first search provider on your list, and if it yields less than X results (only Mojeek really gives us a total count, we have to try + estimate for the others), we try the second, and so on. We then merge the results in a round-robin fashion (first of first, first of second, second of first, second of second, and so on). There’s a bit of more nuanced logic to also properly rank the results with the pin/exclude/raise/lower preferences, because it works differently across providers and not all of them support that, for example.
That’s a good point, thanks for mentioning it. I think we’re safe since Uruk can’t be copyrighted as it’s an ancient city’s name and the name of a time period, and there are no intentional Tolkien references on the product.
I'm building my own cloud (I actually typed claude instead of cloud there... wow). There's no IaaS or PaaS; it's much simpler. I wanted my own way of connecting to machines and the TCP services on those machines without having to install Tailscale (not allowed on a locked-down corporate PC) or pay for Azure or AWS or GCP or even Hetzner or Linode. I've got 10gbps fibre and a huge workstation at home, and I've got lots of laptops and VMs and other outboard stuff that I want to work in concert with that workstation, so I started building something I call Tela (Filipino for fabric; I was sitting in Ninoy Aquino International Airport waiting for a flight when I had the idea, and it's implemented as a network fabric).
https://github.com/paulmooreparks/tela
Working on an audio streaming platform powering an indy internet radio. Looks like Icecast & friends show its age and a similar product can be easily built with the functionality cast down to simply robust streaming & handling "timed playlists". I enjoy every bit of knowing exactly what happens in the code. It's not open source atm, but will be. It's in Go, is a pleasure to write and the deployment takes minimal amount of resources.
Other project is to continue a bit stalled progress of a configuration language BCL - add functions, more structures and fix some hidden scoping issues. Making languages is an endless fun. https://github.com/wkhere/bcl
I made this one-page little tool to help me split bills when going out with friends:
https://trianguloy.github.io/githubPages/SplitBill/splitBill...
The UI is horrible, but I really liked doing the coding. I'm also aware of other similar sites, but this one contains the features I need.
My friend and I are building Strandfall, a highly physical outdoor larp (live action role playing game) that uses custom spatial computers: https://strandfall.com
Players are survivors of a global disaster that has unleashed mysterious, deadly storms. For three hours, they investigate the origin of the storms and make fateful decisions about their future as individuals and as a community.
We received Immersive Arts funding, which means we can run it in Edinburgh later this year. Here's an excerpt from our 2025 grant application about exactly what those are:
--
Our “storm sensors” are novel spatial computers designed for outdoor usage over long distances. They will house ePaper displays, LoRa (long range) radios, accelerometers, gyroscopes, and GPS chips in a 3D-printed enclosure to provide a low-tech way to augment the reality of the park. These computers will be cheaper, more rugged, longer-lived, and more capable than smartphones, deployable to locations with zero cellular service and no battery charging options.
The sensors will be mounted on top of camera tripods for deployment. Runners will carry them through the park, then position and aim them in the correct direction, as co-ordinated by “operators” using walkie talkies. This will let players feel like they are really setting up important equipment, scanning historical sites for clues (like surveyors), and establishing laser communication links. Lacking colourful touchscreens, the sensors will be less distracting for runners, helping them focus on their surroundings. Essentially, they are a highly tactile and deeply realistic way of immersing players in a post-apocalyptic setting, since such devices – not smartphones – are the most likely to be used.
A programming language to hack music, and anything else really https://github.com/audion-lang/audion
The idea came after I finished a permanent piece for a museum using MaxMsp and python. I always had this thought in the back of my mind that "I could express this so much easier in a few lines of code.."
here's the language spec: https://github.com/audion-lang/audion/blob/main/docs/LANGUAG...
I really liked how objects came out, I don't think it needs any more since I can do object composition.
There are some nice functions to generate rhythms and melodies with combinatorics, see src/sequences.rs and melodies.rs
Its a WIP but you can use it now to create music with whatever you want: hardware/daws/supercollider
supercollider is tightly integrated but not required. I havent had time to develop userland libraries yet but I'm working on it
A replacement for CMake/Ninja using golang.
https://github.com/bradphelan/nuke-engine/blob/trunk/USERGUI...
After a day of hating on CMake generator expressions I just wanted a proof of concept that something better is possible.
An example build is https://github.com/bradphelan/nuke-engine/blob/trunk/example...
JDBC does not allow pipelining (a Postgres only feature).
It can reduce the number of db round-trips a lot, especially when using Supabase+RLS (or other systems that require frequent setting of configuration values that are basically fire-and-forget).
Meet Bpdbi, a library with first-class pipelining, which provides a Postgres db driver (that's binary only, as the legacy text-based protocol is no longer needed, it just takes up space) and exposes an API that's more close to Jdbi's that to JDBC's (developer friendly).
https://github.com/bpdbi
It has an extensive benchmark that shows it's on par or faster compared to other db connectivity stacks.
I am working towards a big new release of my web scripting language, hyperscript:
https://hyperscript.org
Hoping to release next Monday
So great to see how much stuff there is going on outside of AI!
I've been working (on and off for over a decade) on a way to manage my unwieldy photo and video collection. I experimented with so many tools before but ultimately nothing ticked all my boxes so I wrote my own. Admittedly lots of niche stuff like proper support for UTC time, stereo photos and partial dates and things like 100% offline face detection and custom tag hierarchies and querying without a database. Each time I add a new feature I'm having a blast :)
https://gitlab.com/geekysquirrel/memo2
Ive been working on the HN Arcade :) https://hnarcade.com
Its always fun to see what games people are building - and some of the lesser known ones are amazing!
I enjoyed this! Thanks for the index
I’ve been working on an open-source tool to help visualize JavaScript/TypeScript codebases (React, Next.js, Node.js). The core idea is building a dependency graph at the function/component level and exploring relationships across the codebase. One feature that turned out to be particularly useful is what I think of as “blast radius”, If you change a function or component, what other parts of the system are affected?
This has been especially helpful when thinking about PR reviews, not just what changed, but what the change might impact indirectly.
It’s still early, and I am working on the cloud version which will include LLM interface to interact with graph and other team collaboration feature and github connection but I’d really appreciate feedback from others who might find this tool useful.
Github: https://github.com/devlensio/devlensOSS Landing page: https://devlens.io/
I literally said this exact thing verbatim with someone who was worshipping ai for code. This is an insane idea, really cool.
One bit of feedback, I'd love if this could be a little more aesthetic in the visualization. The method names are a little hard to read, white on light green.
It was just the first look...My primary focus on was on the Engine. I will upgrade the UI and will surely change the UI you mentioned. You can also contribute if you wish as it is open source (built in nextjs)
Building a bouldering wall at home, for the kids.
A bunch of square panels with a grid pattern to mount hold on, with the panels hanging on french cleats (with a locking system, #TODO) so the panels are easily removable so I can hang something like planters on the wall as well with the same french cleats.
No AI, a bit of computers to draw things out in CAD, but otherwise just manual building stuff.
LEGO. I build LEGO, because I like it and it puts a smile on my face :)
Question, do you "build LEGO" as in "make what's on the box it came in", or do you use LEGO to build things yourself and not following instructions?
(I am old, but as a kid, I remember building what the kit was supposed to build once or _maybe_ twice, but using the parts from the kit and various other kits I had to make things purely out of my head pretty mush all the time. I _think_ that's not how people "play with LEGO" these days?)
Both, I enjoy the process of building, so following instructions. But then it's done and boring. I.e. LEGO supercars sets are nice, they're so nice until you marriage chassis with drivetrain - you can play with all the gears and see them, after the marriage it's just boring. So now I'm working more towards reusing parts and building things of my own.
Edit: LEGO Architecture Set (not sure if they still make it) is super nice for building random stuff - it's a set of pure white bricks, super fun.
I recently built Cranki [0], a free little PWA that generates crosswords using your Anki flashcard lists. It's aimed at language learners (who find flashcards boring and crossword fun!). To be honest I built it just for me and then decided it might be useful for others.
It's all local, no server, no database, etc. Mobile and desktop friendly.
I did build it with the help of AI, but no AI inside the actual thing.
[0] https://cranki.app
I'm working on ghidra-delinker-extension [1], which is a relocatable object file exporter for Ghidra.
The algorithms needed to slice up a Ghidra database into relocatable sections, and especially to recover relocations through analysis are really tricky to get right. My MIPS analyzer in particular is an eldritch horror due to several factors combining into a huge mess (branch delay slots, split HI16/LO16 relocations, code flow analysis, register graph dependency...).
The entire endeavor requires an unusual level of exacting precision to work and will produce some really exotic undefined behavior when it fails, but when it works you feel like a mechanic in a Mad Max universe, stripping programs for parts and building unholy chimeras from them, some examples I've linked in the readme. It has also led to a poster presentation to the SURE workshop at ACM CCS 2025 in Taiwan as a hobbyist, an absolutely insane story.
[1] https://github.com/boricj/ghidra-delinker-extension
I am building educational 16-bit game console for low-cost FPGA chips.
https://github.com/true-grue/Brus-16
I'm building a good intermittent fasting app for iOS using SwiftUI and native Apple APIs for the most "Apple" experience possible. The app is 100% free and without ads.
https://apps.apple.com/si/app/fast-slow-fasting-tracker/id67...
still working on https://stringscales.com - fun sideproject to visualize guitar scales on a configurable fretboard, with interactive note highlighting to a backingtrack.
The backingtrack is what I'm actively improving right now. It's just a pad running now, but it will turn into a full track with bass/drums/piano/... and will feature a comprehensive chords based editor so you can add and save your own progressions with a logged in account.
Does it let you visualise quarter tone fretting? (Or is it just me who's obsessed with Angine de Poitrine right now?)
Playdate games! Super constrained/lo-fi/retro and really a joy to program without AI.
Hyper Vector is dropping on Catalog next week!
I’m reimplementing the Nix parser. I don’t know if it’ll morph into another formatter, or an LSP server, or what. I guess I’d like to try and improve on error messages and build it from scratch. Rust has really brought ergonomics to systems programming, and I wrote very few parsers for real languages, and never one with good errors.
I'm going with my fourth (or fifth) attempt to create a digital twin of my apartment in a game engine.
This time it's in Godot and i want to add more interactive stuff in the virutal version, rather than just redraw the rooms with some lights synced with Home Assistant.
Also working on a Korean guide, so far it introduces Hangul (한글), the Korean alphabet: https://tolearnkorean.com/
Also making an app / web app for memorizing Korean words, takes inspiration from Anki and Duolingo. Words go through 4 stages: 1) matching up pairs 2) Multiple choice answer 3) Writing word through blocks 4) Free-form writing.
It's testable here: https://game.tolearnkorean.com/
Feedback is very welcome.
I've been building the games I've always wanted to play. It used to take months to make something like Medieval Building Simulator.
Now you tell some idea of what you want and AI gives you a better thing than what you originally imagined. Then you go into a rabbit hole... simulate productivity, sales, customer, cash flow, materials for the building, make it curvy, make it pop up in 3D, populate it with NPCs, give your NPCs wages and background, write a dialogue with the ghosts of Christmas. It ends up being a game in itself.
It's good fun. It costs me $20 and I can do it and deploy it from my phone.
A crossword puzzle generator, just for fun. Grid generator in Python because it's easy to hack around and grid generation doesn't need to be particularly fast. Go for populating the grid with words because with large grids there are combinatorial explosions, and Go's speed is beneficial.
Not source-available yet because it's a bunch of hacks (particularly the Python) but maybe one day.
I'm scratching my own itch: building a yet another audio signal generator for smartphone. I need some extra functionality that is not available elsewhere, and impose some limitations "just because": it must be bare minimal PWA.
But actual app does not matter, the main take away for me is: it is easy and fast to write bloatware (esp. with AI), but not that easy to distill to what is really needed. And what looked like a weekend project, a couple of hours max (with help of AI), now lingers for 2+ weeks on-and-off on evenings (of manual effort).
I am building a web application for learning math. I want it to be something between Khan Academy and Math Academy. Here is a demo of fourth grade https://demo.numerikos.com/ Currently the best part about it is one of my kids is using it. I have some more lessons ready, waiting to be released and I am currently working on Trigonometry.
Recently started working on my app for self-tracking again, a customizable combination of habit tracker, health journal and diary.
It should be as powerful as a spreadsheet for self-tracking, but the daily usability should be more on par with a habit tracker app.
The website and waiting list for anyone interested: https://dailyselftrack.com/
I'm still working on my PDF search engine as an offline desktop application. It's at https://www.docgoblin.com/ if you're interrested.
No AI at all just plain old Java, JavaFX, Lucene and enough code to bring them together. I would love to get feedback if you try it!
I have been working on two opensource tools:
https://dhuan.github.io/mock/latest/examples.html
^Command line utility that lets you build APIs with just one command.
https://github.com/dhuan/dop
^JSON/YAML manipulation with AWK style approach.
“Ask HN: What are you working on? (April 2026) (Non AI)”
https://news.ycombinator.com/item?id=47679021
18 points | 1 day ago | 37 comments
Whoops didn't know about this, should have searched first
https://findsubstack.com
I'm building this mostly to scratch my own itch.
A newsfeed for Substack posts from the past 24h. Its helping me discover writers other than just what the algorithm gives me.
I’m building ATProto Alternatives (https://atprotoalternatives.com).
I started looking into AT Protocol recently and find it very interesting, so I started collecting a list of decentralized products built on top of AT that are alternatives to mainstream (popular) products.
This is very cool! I too have been very curious about ATProto after reading Dan Abramov's blog posts on it: https://overreacted.io/a-social-filesystem/
One question, I tried sharing your site with a friend and the share URL is coming up as http://localhost/ which is strange and I've never seen before. Is this something that's possible to customise via meta tag, or is this a bug in my browser (Vanadium 147 on GrapheneOS)?
Seems like accidentally they uploaded a local development build of the HTML:
<meta property="og:url" content="http://localhost/"/>
<link rel="canonical" href="http://localhost/">
Definitely something on my end. Thanks for the heads up! I’ll try to fix it tonight.
I've been working on a super-straightforward volunteering platform -- https://handsup.barryvan.com.au/ . Mainly because I got frustrated with my kids' school and sailing club activities being (mis)managed across WhatsApp, Google Forms, pieces of paper, and the like.
I'm working on identifying mental disorders from speech patterns. The classifier itself is not really the point, though, but rather finding specific disruptions that we can correlate with areas of the brain and, eventually, use to identify their root causes.
I'm currently working on a small tool that helps me setup new personal computers in a purely descriptive way.
A little bit like ansible, but then totally not like ansible.
It generates SSH-Keys. It clones repositories, installs uv and rust. It removes Snap from my Ubuntu machines and installs firefox from the mozilla repositories.
1. Reviving an old project that lets users sync their google docs as blog posts.
2. Writing a rich text editor library powered by pretext for cheap pagination
3. A layout engine that understands html/CSS subset for lightning fast pdf generation
Although AI is the main reason why I'm able to work on all these projects concurrently.
MusicLibrarian, a macOS app that deep-cleans Apple Music libraries. iCloud sync duplicated my playlists up to 540x (77,542 ghost entries). Built it in Swift, uses ITunesLibrary framework + AppleScript. https://thevaultdj.com
I'm building the CAD app I have always wanted. Free, Parametric, STEP export for 3d printing. The input is just a javascript editor. It uses OpenCascade under the hood. Lots of fun.
reverse engineered Elektron Octatarack binary data files and now writing libraries to interact with/modify them.
https://gitlab.com/ot-tools/ot-tools
i’m currently avoiding committing/releasing a bunch of changes i did last week because people are actually using the library already (the curse of writing something useful lol)
I am building something similar to Windows Task Manager but it also works on Linux, launches/monitors web servers, and provides a visual dashboard for docker containers.
https://github.com/prettydiff/aphorio
This project makes no use of AI.
New programming language.
Took some good ideas of Pascal and making it more modern. Minimal runtime, manual memory management, single (small) executable, no dependencies. Compiler itself is written in Swift and I am using QBE as a backend ATM.
A family. Hope to have a little one in the next year! Wish me luck!
Same here! Good luck
A Chrome extension that helps you efficiently declutter stale tabs.
https://SupplementDEX.com - helping people find out if their supplements work
I recently made a 'notification-driven RPC' app for Android to help me help others:
Notifactor (Android App)
A lightweight native Android app that intercepts notifications on the device and triggers actions based on configurable rules.
The app uses Android's NotificationListenerService API. Once granted notification access it receives a callback for every notification posted on the device. It then checks each notification against configured rules and runs the configured action.
Why would you want this?
I don't know why you would want this but I can tell you why I made it: to make it easier to control some functions on Android devices used by people I often help using them in some way. My mother's Android TV (which I use to communicate with her through Linphone), phone and tablet sometimes stop doing the right thing. I live about 1300 km to the north of where she lives so I can't just hop on my bike to fix things. Thus far I relied on a set of Termux scripts on these devices to keep a reverse ssh tunnel open to an endpoint on my server but this has a number of drawbacks: the tunnel is not always there when I need it due to WiFi dropouts and other similar problems and the constant connection uses battery power on the phone and tablet. If only I could cause the tunnel to be created when I need it and brought down when it is not needed... Well, that is possible using Notifactor by sending a notification on a specific channel (ntfy refers to these as topics) whereupon Notifactor runs a Termux script which manages the tunnel (etc.).
Since the moment "AI" became a thing, I've been rolling my eyes and looking the other way since I know it's just a half-assed hype buzzword. With that in mind, pretty much everything I work on is not AI related. As a hobby(and to a degree to be prepared for WW3), building a ton of drones and drone equipment, flying them and finding more and more ways to push their range to the absolute limit.
I don't think it qualifies as AI in the modern day and age but NLP in general. It's truly amazing how easy it is to spot troll farms online and no one is doing anything about it, be it individuals, private sector or even on national level, given that those should be considered a risk for national security.
This seems a little bit of a duplicate of [1], but I can repeat my answer here as any views help!
My wife and I continue to work on Uruky, a EU-based Kagi alternative [2].
Since last month we finally got our production API Key for EUSP/STAAN (it was certainly the slowest and most complicated search provider to adopt, so far), and that brought us to 5 search providers you can choose from and sort as you prefer.
We already have got over 40 paying customers (excluding family and friends, we’re guessing these paying customers came from some privacy listings and HN comments) and have exited beta last month!
Customers seem to really enjoy the simple UI (search can be used without JS) and search personalization (from choosing the providers to the domain boosting and exclusion). We also have hashbangs (like "!g", "!d", or “!e”) when something doesn’t quite give you what you’d expect, though.
You can see the main differences between Kagi and Uruky in the linked page, but one huge difference is that with Uruky, after being a paying customer for 12 months, you get a copy of the source code!
One thing we’re struggling with is outreach because we want to do it ethically, and it’s hard to find communities or places to sponsor which are privacy-focused and don’t require €5k+ deals. Ideas are welcome! Because of bots there isn’t a free trial easily available, but if you’re a human and you’d like to try it for a couple of days for free, reach out with your account number and we’ll set that up!
Thanks.
P.S.: Because people have asked before, our tech stack is intentionally very "boring" (as in, it generates and serves the HTML + bits of JS to enhance settings and such), using Deno in the backend (for easier TypeScript), PostgreSQL for the DB, and Docker for easier deploying.
P.P.S.: Because this has been also brought up before, the name has no special meaning but we read it like "Euro-key" in English. Names are hard, and we’re aware it can remind people of Uruk and Uruk-hai. That’s OK.
P.P.P.S.: Another frequent question here is “how does it work?” When you search, we query the first search provider on your list, and if it yields less than X results (only Mojeek really gives us a total count, we have to try + estimate for the others), we try the second, and so on. We then merge the results in a round-robin fashion (first of first, first of second, second of first, second of second, and so on). There’s a bit of more nuanced logic to also properly rank the results with the pin/exclude/raise/lower preferences, because it works differently across providers and not all of them support that, for example.
[1] https://news.ycombinator.com/item?id=47679021
[2] https://uruky.com
> Names are hard, and we’re aware it can remind people of Uruk and Uruk-hai. That’s OK.
What I would worry about is if it reminds whoever holds Tolkien's copyrights of Uruk and Uruk-hai.
That’s a good point, thanks for mentioning it. I think we’re safe since Uruk can’t be copyrighted as it’s an ancient city’s name and the name of a time period, and there are no intentional Tolkien references on the product.
Still working on my two decades old forum. Still going strong. https://www.tractorfan.app