Not a programmer (accountant here) and thanks to Github copilot in VSCode, I was able to make the app I couldn't find. It's been a super fun hobby over the last 6 or 7 months, especially hosting it myself and my only cost now is Apple's recurring 99 USD in case i want to ship an update later, so I can keep it up for free into the future.
95% of the app is boilerplate API and DB stuff but I stood 0 chance of making this without the LLM handling syntax and the volume of code.
It's Plantshare and it helps people share and find plants, because if you do gardening, your plants are likely making more free plants all the time. which you can now share and find on this app. Saw an app for sharing tools in your neighborhood and I thought that was really cool, so I did this but for giving plants/seeds/cuttings away.
I work as a software engineer, so I was familiar with the stack I told the AI to build the forum with. Also, the forum was made specifically to see what vibecoding was like, so I didn't look for alternatives. But I'd say is working quite well code-wise.
Yes, the only issue I had to debug was the password hashing iterations being too much for the low resources of the VPS I was using. That made for logins and registration taking 2 seconds; other than that, smooth as butter. You can check the code at
Idea is to give agents and humans power to both post to and consume from a vector/RAG database. So then you can do something like "Im looking for someone whose read this book recently and wants to talk about it" or "find me blog posts about x y or z". Pretty difficult to get over the need a user base to have a user base hurdle though and Im horrible with marketing
Not vibe coded, but in the last six months I’ve been heads-down building ORA—an autonomous super agent that represents the next step toward AGI. An agent, specifically designed to help people go from vibe coding to production-ready code.
The gap between 'I got AI to make something' and 'this is actually deployable' is massive. Curious if others in this thread are thinking about that transition.
Demo: https://x.com/OscerraHQ
https://spokengoods.com/ - a podcast summary site that also pulls out product mentions. I kept forgetting which books were mentioned while I was out doing stuff.
Or don't. It has been made fairly clear that AI-generated content isn't on-topic for HN.
Now, if you have written a product that is successful and want to share that success, and just happenned to vibe-code it, that is a different story. But "Hey, I vibe-coded this." is not particularly interesting.
I disagree. It's interesting to see what sorts of tiny problems people had and solved. The code is less important than the problem itself. A bit like duct-taping a solution in your garage for an hyper-specific issue in your house.
Estimated: 10 - 15 minutes work each. Total 1 - 2 hours of work, plus posting to HN and reading and replying to comments.
- A full auto-updating static interactive archive of HN with various queries and new views: https://hackerbook.dosaygo.com
Estimated: 4 days of work
- A "most cited" HN-ecosystem ranking of topics that span multiple discussions over time ranked by the topics whose discussions/comments were most cited by other HN comments: https://hacker-backlinks.browserbox.io/
- BlueDot - a universal cloud-console TUI that lets you search, compare and rent VPS across 6 clouds (GCP, Hetzner, AWS, Azure, Digital Ocean and Vultr), right from your terminal: https://tui.bluedot.ink
Estimated: (my own thinking - weeks over years), coding: 2 days of work
- An email-bridge for your CLI AI agents to let your drive them from email while you're out, leaving your laptop at home: https://ai-chat.email
Estimated: 12 days of work (so far)
- A little open-source tool to mute your macOS mic to zero as well as detect Siri listening overrides and open settings so you can toggle: https://github.com/BrowserBox/NoSpy
Estimated: 1 hour of work
- A set of AI "taste & doctrine" files to teach your agents code that belongs and give them procedures for checking semantics for quality, or case-by-case human overrides, to keep codebases higher quality and more maintainable, rather than just ensuring syntax parses: https://ai-lint.dosaygo.com
- An impressionistic Windows 98 desktop called Windows 98½ with real Web Browsers: retro UI wrapper around BrowserBox using the new BrowserBox WebView Embedding API - a showcase for my corporate work and a cool 90s tech nostalgia art project in one, what could be better! :)
Estimated: 5 - 6 weeks of work for the Desktop (over the last 1 - 2 years), and 7 years of work for BrowserBox (rn > 90% pre-AI).
- Structropy - Towards a metric of organization. An adaption of qualities of the metric of Shannon entropy toward not just "surprise" but strucutre, to try to fill the gap that Shannon leaves where "highly random" is "high information" but "low organization". Loosely based on expected sorting time of new inserts. https://github.com/DOSAYGO-STUDIO/structropy
Estimated: months of thinking over years, then 1 - 2 days of coding
And a few more even bigger ones. And that's just the last 3 months!
Not a programmer (accountant here) and thanks to Github copilot in VSCode, I was able to make the app I couldn't find. It's been a super fun hobby over the last 6 or 7 months, especially hosting it myself and my only cost now is Apple's recurring 99 USD in case i want to ship an update later, so I can keep it up for free into the future.
95% of the app is boilerplate API and DB stuff but I stood 0 chance of making this without the LLM handling syntax and the volume of code.
It's Plantshare and it helps people share and find plants, because if you do gardening, your plants are likely making more free plants all the time. which you can now share and find on this app. Saw an app for sharing tools in your neighborhood and I thought that was really cool, so I did this but for giving plants/seeds/cuttings away.
https://ps-prod.bloodys.cc/links/getplantshare
http://forum.philosofriends.com/
I wanted to try vibecoding, so I decided to make a 100% vibe coded project: a HN-clone forum just for philosophical topics/questions and articles
I am interested in building a forum. Hiw well did the vibe coding approach work for you vs just using something oof the shelf?
I work as a software engineer, so I was familiar with the stack I told the AI to build the forum with. Also, the forum was made specifically to see what vibecoding was like, so I didn't look for alternatives. But I'd say is working quite well code-wise.
Same. We have a very similar use case then. Did it handle user flows well? I'm talking about account validation, deactivation and recovery.
Or better, if the code is on GitHub, I would love to have a look.
Yes, the only issue I had to debug was the password hashing iterations being too much for the low resources of the VPS I was using. That made for logins and registration taking 2 seconds; other than that, smooth as butter. You can check the code at
https://github.com/nomai-tech/forum.philosofriends.com
Thank you! And django too! How convenient
I've just finished this project: https://partysnap.fun/
It's a kind of an AI Photobooth for events and celebrations. It's not 100% finished (need to add payments), but already pretty usable.
P.S.: Mostly done with Cursor
Been working on mistro.sh
Idea is to give agents and humans power to both post to and consume from a vector/RAG database. So then you can do something like "Im looking for someone whose read this book recently and wants to talk about it" or "find me blog posts about x y or z". Pretty difficult to get over the need a user base to have a user base hurdle though and Im horrible with marketing
Founder here.
Not vibe coded, but in the last six months I’ve been heads-down building ORA—an autonomous super agent that represents the next step toward AGI. An agent, specifically designed to help people go from vibe coding to production-ready code. The gap between 'I got AI to make something' and 'this is actually deployable' is massive. Curious if others in this thread are thinking about that transition. Demo: https://x.com/OscerraHQ
https://twokindsof.com
Just a silly site I built in a weekend.
Nice, pretty cool site!
https://spokengoods.com/ - a podcast summary site that also pulls out product mentions. I kept forgetting which books were mentioned while I was out doing stuff.
https://seaquel.app/
It's the SQL client I've always wished I had. It's a desktop app, but I made it work in the browser too thanks to DuckDB Wasm.
Every vibe coded website always has the same vibe
It was an easy project, but managed to hit frontpage here :
https://news.ycombinator.com/item?id=46345897
https://tendayweekcalendar.com
I took a python library for generating posters from maps and wrapped it up as a web UI:
https://maptoposter.penk.in/
I mean AI did all the work for me with some minimal guidance. All and all it took about 3 hours to do with PaaS hosting
Or don't. It has been made fairly clear that AI-generated content isn't on-topic for HN.
Now, if you have written a product that is successful and want to share that success, and just happenned to vibe-code it, that is a different story. But "Hey, I vibe-coded this." is not particularly interesting.
I disagree. It's interesting to see what sorts of tiny problems people had and solved. The code is less important than the problem itself. A bit like duct-taping a solution in your garage for an hyper-specific issue in your house.
A couple I've done lately:
- UDP777 - a simple UDP pager in Go: https://www.udp7777.com/
- FIPSPad - a simple Notepad app with encryption-at-rest that only runs if your system is FIPS-compliant in Rust: https://fipspad.browserbox.io - A suite of AI-written alternative HN front pages (serious joy and laughs): https://dosaygo-studio.github.io/hn-front-page-2035/ - A full auto-updating static interactive archive of HN with various queries and new views: https://hackerbook.dosaygo.com - A "most cited" HN-ecosystem ranking of topics that span multiple discussions over time ranked by the topics whose discussions/comments were most cited by other HN comments: https://hacker-backlinks.browserbox.io/ - A game where you have to do mental math for hexadecimal crossword puzzles: https://do-say-go.github.io/hexfiend/?hn=2 - A game where you factor RSA semiprimes by doing constraint propagation on a guessable tableaux: https://do-say-go.github.io/insights/ - BlueDot - a universal cloud-console TUI that lets you search, compare and rent VPS across 6 clouds (GCP, Hetzner, AWS, Azure, Digital Ocean and Vultr), right from your terminal: https://tui.bluedot.ink - An implementation of an idea I had from 2013, an "approximate matching" LZW algorithm: https://github.com/BrowserBox/LZW-X - An email-bridge for your CLI AI agents to let your drive them from email while you're out, leaving your laptop at home: https://ai-chat.email - A little open-source tool to mute your macOS mic to zero as well as detect Siri listening overrides and open settings so you can toggle: https://github.com/BrowserBox/NoSpy - A set of AI "taste & doctrine" files to teach your agents code that belongs and give them procedures for checking semantics for quality, or case-by-case human overrides, to keep codebases higher quality and more maintainable, rather than just ensuring syntax parses: https://ai-lint.dosaygo.com - git-prime - mine git commit hashes for large primes by fuzzing a nonce annotation in the message: https://textonly.github.io/git-prime/ - prime news - a view of all of HN that only selects items with an ID which is a prime number: https://dosaygo-studio.github.io/prime-news/ - A visual proof of the Pythagorean theorem that let's you play with the "squares to triangles to bigger square" mapping: https://do-say-go.github.io/insights/others/interactive_peri... - An impressionistic Windows 98 desktop called Windows 98½ with real Web Browsers: retro UI wrapper around BrowserBox using the new BrowserBox WebView Embedding API - a showcase for my corporate work and a cool 90s tech nostalgia art project in one, what could be better! :) - Structropy - Towards a metric of organization. An adaption of qualities of the metric of Shannon entropy toward not just "surprise" but strucutre, to try to fill the gap that Shannon leaves where "highly random" is "high information" but "low organization". Loosely based on expected sorting time of new inserts. https://github.com/DOSAYGO-STUDIO/structropy And a few more even bigger ones. And that's just the last 3 months!A couple?
[dead]