Trying to bring modern typed dependencies context DX to Python https://github.com/alexdrydew/inlay. Had a lot of fun with implementing efficient solver for this. It has rough edges (e.g. supporting only 3.14+), but I already migrated my other project to use it and love how it turns out, threw out huge piles of boilerplate
The State Department's CEAC portal times out constantly. Session expires, you lose progress. Every tool that solves this charges some money, or bundles you into an attorney service.
So I built a free interface to fill your DS-160 comfortably at your own pace - then a one-click bookmarklet autofills the CEAC portal for you. No account. No data leaving your browser. 100% free.
~10M+ nonimmigrant visa applicants file this form every year.
I continue to work on my general-purpose object store; which is a 'multi-modal' data management system. Using it, you can create and manage millions of files in a traditional hierarchical folder structure; create relational tables that can be queried and analyzed quickly and easily; and manage semi-structured data normally handled by nosql systems.
I'm working on a toolkit for building "full-stack" libraries. I.e. libraries that bundle frontend hooks, backend routes, and a database schema into a single package.
On top of this I built a library for email that listens to Resend webhooks and stores the email (with thread information) directly in the user’s database. This is a more complete package than what Resend themselves could provide with their backend-only library.
Working on a little DSP (digital signal processing) macOS menu bar app. I've got one of those typical bedroom producer setups (Focusrite Scarlett 4i4, Shure SM57), but no hardware for real-time stuff like compression, eq etc that would help me polish my audio in calls. I used to run OBS to host free VSTs (audio effects) and then hijacked the monitor output with a loopback driver to route it into Zoom etc, but it was a janky workflow. So I made a little app, with AI assistance. I made it just for myself, but it's open source - https://github.com/JyrgenSuvalov/klaar/.
Building a website that helps patients to find clinical trials that are actually looking for patients right now. 70% of contacts listed on governmental website for studies that should be active, do not respond to patients. It’s AllClinicalTrials.com
Currently Working on https://devlens.io which is an Open source Intelligent Codebase Visualizer for javascript, Reactjs, nextjs and nodejs for easy onboarding and easy PR review.
Also the main MOAT of Devlens is not the visualizer but the blast radius feature. Which essentially tells you if I change this file/function where it will affect my Codebase which is specially useful for PR review.
A non interpretive non authoritative temporal recording of institutional publication. I wrote a grammar to build long horizon , non founder dependent, capture resistant systems.
I bought $5 worth of Claude API tokens the other day because Newspeak [0] is implementing an API interface in the IDE.
I work with Newspeak every night building all kinds of crazy stuff, from the raycasting tutorial to an IndexedDB interface. Currently, I have the IDE running as an Isolated Web App for access to TCPSockets [1][2][3].
I'm implementing ancient TCP protocols bringing them to the web.
I used claude to rebuild a simple mobile math game I had previously prototyped in GDevelop. the rebuild is ai assisted but the original game had been created entirely in GDevelop with the use of its visual programming.
I have been working on making streaming cheaper/more private/lower latency (via WebRTC) in OBS. Been working on this site to get people excited https://webrtcforthestreamer.com/ after it is done want to record a YouTube video for it.
After that I want to spend the weekend just closing out Pion bugs/relaxing :)
Good timings on this. I think a lot of people are quietly building useful things and just not posting about them because it doesn't get traction next to AI announcements. would love to see more of these threads.
Good to see this post, a simple non AI tagged post would make it more easier to find useful projects that are not focused on AI, while keeping the community together.
Thanks! Do you mean a [Non-AI] tag or something on the answers to the WAYWO post? That could make it easier to search indeed, if it's somehow standardized.
working on a crossword tool that leans on qxw [https://quinapalus.com/qxw.html] for the grid filler and borrows some UI ideas from it, but adds integrated clue management, as well as experimenting with some UI features that are perhaps too idiosyncratic for qxw to add. I will contribute features back to qxw wherever that makes sense too - largely, the latter focuses on being rock solid, stable, and bug free, and I want to take more of a "let's toss in a bunch of features and see what ends up being useful" approach.
Vibe coding allowed me to build a couple of map-related things I geek out about:
- a map with all direct train destinations from a given station https://trainconnections.ai-tigers.com/
- a navigation app which avoids traffic jams even it takes longer (I like to stay in flow state)
- a bicycle navigation app which avoids traffic lights (for the same reason as the traffic jam app)
yea trying to get beta out for an online game. using deepski and its cheap af rn discount 75% im still at $4 usage after a week of it chugging, and ppl say its not as good as the other models but we are at point where they all distilled off each other and are within margins such that u just pick cheapest one and it works ok as the others, eg none of them are gonna give u anything novel by default if u dont know what ur doing anyway but they can just autocomplete stuff pretty well and explain / find bugs answer questions etc well enough so idk what that means
I recently added a links page to my site for links that use the human.json protocol[0,1] to denote actual human created (non-AI) content, and added metadata parsing that adds title, description and fediverse link where available[2,3]. I'm also working on some game projects. Yes I know my site is terrible and half-assed and my Python skills leave something to be desired.
None of this is of any real interest to HN but I just didn't want to be entirely off topic by mentioning how disappointing it is that even though this thread is marked as explicitly for non-AI projects, a number of people posted AI projects anyway and one person is saying we need to just accept that AI will eat everything and to not even bother trying to avoid it.
Sad state of affairs, the degree to which AI is being forcefully normalized, and how much of "hacker culture" simply accepts it as inevitable. To the point that simply having a space explicitly for non-AI projects won't be permitted.
If you must know, I'm writing a real-mode 16-bit x86 assembly to C translator in Common Lisp to be able to run that 3-sector inference engine, which was posted the other day, outside of QEmu, currently I'm debugging it. I'm doing this because it's fun but I can rationalize it as follows: It allows me to simultaneously learn Common Lisp, as I'm otherwise an OCaml & Scheme guy; 16-bit x86 assembly, since I grew up learning 68k assembly and segments are otherwise disgusting; and also slop generator architecture, right now I don't feel like reading a summary paper, those always leave out details. This is 100% hand-written in Emacs in Evil mode, and with SLIME. I guess that counts as a coding assistant, and it's written in LISP and LISP=AI, so that I can claim that I am familiar with generative AI technologies for coding when applying to jobs.
Hi, I'm David and I run the "What are you working on?" posts. I hear what you're saying and yet I think it's just the new reality. There will be more projects and some of those will be quickly-made vibe-coded projects. I'm not against it.
In Moveable Feast, Hemingway talks about being opposed to ski lifts (which were probably fairly new at the time). He thought, if you're fit enough to hike up the mountain, you're going to be fit enough to ski down without getting injured; it opens skiing up to people who maybe shouldn't be skiing. I deeply prefer ski lifts. :-) And I love agentic AI coding.
I guess what I'm saying is that we shouldn't compare it to the past, because that's gone. And yet, each month the number of comments breaks 1000, so "drowning" is a fair word for just number of projects alone. And I don't have any answers.
Anyway, you (and anyone reading this) can feel free to email me. My email in my profile.
It doesn't seem like you hear what they're saying at all.
Writing code to coax computers into doing specific things in specific ways is a craft, trade, and art form older than pretty much anybody alive today.
Writing prose instructions to direct LLM's to generate that code also produces software, but it's essentially a different craft/trade/art altogether.
Whatever success the latter claims in commerce or mindshare, both of these arts will coexist for longer than anybody will be around.
It's completely natural to carve out spaces where people who appreciate the process or character of hand-crafted projects can discuss those projects without people talking about a whole different thing crowding out their conversations.
I heard what they're saying, and while I don't personally agree with it, I can see it's shared by a lot of people.
Personally, I'm not sure a separate thread makes sense. What if people mark their project as "[ARTISAN]" like they do with [Remote] in Who's Hiring? Any other ideas?
One day, way before the lifts had started, I hiked up the mountain, snowboard under my arm. Trudging, my breath heavy, lungs filling, oxygen, snow, blue sky, cold but my body warmed. There was not one track in the fresh powder. My snowboard was slower in the powder, and I felt safe in the knowledge. I had to work a little harder. Carving.
Trying to bring modern typed dependencies context DX to Python https://github.com/alexdrydew/inlay. Had a lot of fun with implementing efficient solver for this. It has rough edges (e.g. supporting only 3.14+), but I already migrated my other project to use it and love how it turns out, threw out huge piles of boilerplate
I built a free DS-160 filler - the only one I could find thats 100% free
https://fillvisa.com/form/usa-ds160/
The State Department's CEAC portal times out constantly. Session expires, you lose progress. Every tool that solves this charges some money, or bundles you into an attorney service.
So I built a free interface to fill your DS-160 comfortably at your own pace - then a one-click bookmarklet autofills the CEAC portal for you. No account. No data leaving your browser. 100% free.
~10M+ nonimmigrant visa applicants file this form every year.
I understand the frustration with dealing with the constant log outs, but I am too paranoid to try anything automated in case they have bot detection.
For some people, getting a US visa denied can ruin their lives (job loss).
I continue to work on my general-purpose object store; which is a 'multi-modal' data management system. Using it, you can create and manage millions of files in a traditional hierarchical folder structure; create relational tables that can be queried and analyzed quickly and easily; and manage semi-structured data normally handled by nosql systems.
https://www.didgets.com
I'm working on a toolkit for building "full-stack" libraries. I.e. libraries that bundle frontend hooks, backend routes, and a database schema into a single package.
On top of this I built a library for email that listens to Resend webhooks and stores the email (with thread information) directly in the user’s database. This is a more complete package than what Resend themselves could provide with their backend-only library.
https://github.com/rejot-dev/fragno https://fragno.dev/fragments/resend
Working on a little DSP (digital signal processing) macOS menu bar app. I've got one of those typical bedroom producer setups (Focusrite Scarlett 4i4, Shure SM57), but no hardware for real-time stuff like compression, eq etc that would help me polish my audio in calls. I used to run OBS to host free VSTs (audio effects) and then hijacked the monitor output with a loopback driver to route it into Zoom etc, but it was a janky workflow. So I made a little app, with AI assistance. I made it just for myself, but it's open source - https://github.com/JyrgenSuvalov/klaar/.
Building a website that helps patients to find clinical trials that are actually looking for patients right now. 70% of contacts listed on governmental website for studies that should be active, do not respond to patients. It’s AllClinicalTrials.com
Did you measure the 70% via sampling by contacting listed coordinators yourself or pulling from an aggregator?
What I've enjoyed recently: someone building a mechanical watch movement from scratch. I find it facinating and so skillfull.
My startup is a 3D editor that turns any indoor place to FPS immersive game. In one hour.
https://free-visit.net
Ps : never had an interview at yc but, it's ok.
An instant remote for shared spaces. Always wanted to innovate something.
https://www.kickstarter.com/projects/openinfrared/openinfrar...
Helping a friend repair the internal oscilloscope control board on an IFR-1200 Super-S communications service monitor.
There's a new undocumented version of the board with surface mount parts that rhymes with the old known through hole version.
Currently Working on https://devlens.io which is an Open source Intelligent Codebase Visualizer for javascript, Reactjs, nextjs and nodejs for easy onboarding and easy PR review.
The open source version is live at https://github.com/devlensio/devlensOSS
You can star the repo if you like the project. I am still working on cloud version which will have lots of features.
Also the main MOAT of Devlens is not the visualizer but the blast radius feature. Which essentially tells you if I change this file/function where it will affect my Codebase which is specially useful for PR review.
A non interpretive non authoritative temporal recording of institutional publication. I wrote a grammar to build long horizon , non founder dependent, capture resistant systems.
I bought $5 worth of Claude API tokens the other day because Newspeak [0] is implementing an API interface in the IDE.
I work with Newspeak every night building all kinds of crazy stuff, from the raycasting tutorial to an IndexedDB interface. Currently, I have the IDE running as an Isolated Web App for access to TCPSockets [1][2][3].
I'm implementing ancient TCP protocols bringing them to the web.
[0] https://newspeaklanguage.org
[1] https://developer.chrome.com/docs/iwa
[2] https://developer.chrome.com/docs/iwa/direct-sockets
[3] https://github.com/GoogleChromeLabs/telnet-client (example IWA)
I used claude to rebuild a simple mobile math game I had previously prototyped in GDevelop. the rebuild is ai assisted but the original game had been created entirely in GDevelop with the use of its visual programming.
https://www.mathabito.com/
I have been working on making streaming cheaper/more private/lower latency (via WebRTC) in OBS. Been working on this site to get people excited https://webrtcforthestreamer.com/ after it is done want to record a YouTube video for it.
After that I want to spend the weekend just closing out Pion bugs/relaxing :)
Good timings on this. I think a lot of people are quietly building useful things and just not posting about them because it doesn't get traction next to AI announcements. would love to see more of these threads.
Good to see this post, a simple non AI tagged post would make it more easier to find useful projects that are not focused on AI, while keeping the community together.
Thanks! Do you mean a [Non-AI] tag or something on the answers to the WAYWO post? That could make it easier to search indeed, if it's somehow standardized.
I am working on npm supply-chain security. My tool is called @lateos/npm-scanner You can install it via npm -I @lateos/npm-scanner . The npm readme page is at https://www.npmjs.com/package/@lateos/npm-scan and open spruce is at https://github.com/lateos-ai/npm-scan . I appreciate all input and feedback
working on a crossword tool that leans on qxw [https://quinapalus.com/qxw.html] for the grid filler and borrows some UI ideas from it, but adds integrated clue management, as well as experimenting with some UI features that are perhaps too idiosyncratic for qxw to add. I will contribute features back to qxw wherever that makes sense too - largely, the latter focuses on being rock solid, stable, and bug free, and I want to take more of a "let's toss in a bunch of features and see what ends up being useful" approach.
Vibe coding allowed me to build a couple of map-related things I geek out about: - a map with all direct train destinations from a given station https://trainconnections.ai-tigers.com/ - a navigation app which avoids traffic jams even it takes longer (I like to stay in flow state) - a bicycle navigation app which avoids traffic lights (for the same reason as the traffic jam app)
Post it somewhere else, this is for non-AI projects.
yea trying to get beta out for an online game. using deepski and its cheap af rn discount 75% im still at $4 usage after a week of it chugging, and ppl say its not as good as the other models but we are at point where they all distilled off each other and are within margins such that u just pick cheapest one and it works ok as the others, eg none of them are gonna give u anything novel by default if u dont know what ur doing anyway but they can just autocomplete stuff pretty well and explain / find bugs answer questions etc well enough so idk what that means
I recently added a links page to my site for links that use the human.json protocol[0,1] to denote actual human created (non-AI) content, and added metadata parsing that adds title, description and fediverse link where available[2,3]. I'm also working on some game projects. Yes I know my site is terrible and half-assed and my Python skills leave something to be desired.
None of this is of any real interest to HN but I just didn't want to be entirely off topic by mentioning how disappointing it is that even though this thread is marked as explicitly for non-AI projects, a number of people posted AI projects anyway and one person is saying we need to just accept that AI will eat everything and to not even bother trying to avoid it.
Sad state of affairs, the degree to which AI is being forcefully normalized, and how much of "hacker culture" simply accepts it as inevitable. To the point that simply having a space explicitly for non-AI projects won't be permitted.
[0] https://codeberg.org/robida/human.json
[1] https://news.ycombinator.com/item?id=47298655
[2] https://kennethrapp.net/pages/links.html
[3] https://kennethrapp.net/posts/housekeeping-2-electric-boogal...
Working and getting crazy good results in compression algorithms :)
If you must know, I'm writing a real-mode 16-bit x86 assembly to C translator in Common Lisp to be able to run that 3-sector inference engine, which was posted the other day, outside of QEmu, currently I'm debugging it. I'm doing this because it's fun but I can rationalize it as follows: It allows me to simultaneously learn Common Lisp, as I'm otherwise an OCaml & Scheme guy; 16-bit x86 assembly, since I grew up learning 68k assembly and segments are otherwise disgusting; and also slop generator architecture, right now I don't feel like reading a summary paper, those always leave out details. This is 100% hand-written in Emacs in Evil mode, and with SLIME. I guess that counts as a coding assistant, and it's written in LISP and LISP=AI, so that I can claim that I am familiar with generative AI technologies for coding when applying to jobs.
Here to read all the comments!
Hi, I'm David and I run the "What are you working on?" posts. I hear what you're saying and yet I think it's just the new reality. There will be more projects and some of those will be quickly-made vibe-coded projects. I'm not against it.
In Moveable Feast, Hemingway talks about being opposed to ski lifts (which were probably fairly new at the time). He thought, if you're fit enough to hike up the mountain, you're going to be fit enough to ski down without getting injured; it opens skiing up to people who maybe shouldn't be skiing. I deeply prefer ski lifts. :-) And I love agentic AI coding.
I guess what I'm saying is that we shouldn't compare it to the past, because that's gone. And yet, each month the number of comments breaks 1000, so "drowning" is a fair word for just number of projects alone. And I don't have any answers.
Anyway, you (and anyone reading this) can feel free to email me. My email in my profile.
It doesn't seem like you hear what they're saying at all.
Writing code to coax computers into doing specific things in specific ways is a craft, trade, and art form older than pretty much anybody alive today.
Writing prose instructions to direct LLM's to generate that code also produces software, but it's essentially a different craft/trade/art altogether.
Whatever success the latter claims in commerce or mindshare, both of these arts will coexist for longer than anybody will be around.
It's completely natural to carve out spaces where people who appreciate the process or character of hand-crafted projects can discuss those projects without people talking about a whole different thing crowding out their conversations.
I heard what they're saying, and while I don't personally agree with it, I can see it's shared by a lot of people.
Personally, I'm not sure a separate thread makes sense. What if people mark their project as "[ARTISAN]" like they do with [Remote] in Who's Hiring? Any other ideas?
I would also prefer it to be on the same thread. I think "[Non-AI]" or "[NO-AI]" might be more obvious, though.
I had only ever known ski lifts.
One day, way before the lifts had started, I hiked up the mountain, snowboard under my arm. Trudging, my breath heavy, lungs filling, oxygen, snow, blue sky, cold but my body warmed. There was not one track in the fresh powder. My snowboard was slower in the powder, and I felt safe in the knowledge. I had to work a little harder. Carving.