Show HN for May 6, 2026

For a game I'm developing I needed an easy way to create hotspot areas on images. So I hacked together a small tool to do so and before I knew it, I had created an entire app :-)

It allows you to generate JSON or YAML from the coordinates you tack on the image. tack runs entirely in your browser, there is no server side component to it, so good in terms of privacy.

Hope this is helpful.

Hi HN!

I made an app that takes a photo of a paper drawing and, in a handful of seconds, creates a fully rigged character that can be used in an animation or little story. It doesn’t use any image-to-video generative AI models. Instead, I built it using the years of insights I’ve picked up studying children’s drawings and character animation.

Today we’re releasing a community beta. I respect this community and would value any feedback you offer. It’s easy to try- you don’t need to create an account to check it out. We’ve got several free stories to drop your character into, and a Mother’s Day eCard.

I’m also working on a tool, DoodleMate Studio, to easily allow people to author their own stories instead of using premade templates. But what form that takes is going to be highly dependent on the type of feedback we get from the community with this beta.

How this came to be:

I’ve worked in this space for a while. Here’s an old HN post related to a popular tech demo I did ( https://news.ycombinator.com/item?id=30469321) and another one from when I open sourced the data and code ( https://news.ycombinator.com/item?id=35561203). I also wrote a SIGGRAPH paper about the methodology (https://dl.acm.org/doi/full/10.1145/3592788).

I’d moved on to other things, but had always felt like there was such potential in this space. Last year I decided I was over big tech and, with a lot of encouragement from my family, finally decided to pursue this seriously. Since then, my wife and I have been building this together. We’re bootstrapping at the moment, trying to give ourselves time and space to make sure DoodleMate turns into something wonderful and wholesome.

Thanks, Jesse

Hi, HN. I read about 60-100 books a year and wanted a frictionless ereader for Project Gutenberg books, so I built Guten.

Users can search the PG catalog within the app or Share a book's page from their browser to Guten for import. The app includes the usual ereader features like search, highlight, bookmarks, and notes. I like my apps privacy-focused: Guten doesn't require an account, and there are no ads or tracking.

The reading experience is totally free, but there's also a one-time premium upgrade available for extra features like TTS read-aloud and note/highlight exports.

Play Store: https://play.google.com/store/apps/details?id=com.bhunt.gute...

Would love feedback while it's still in open testing.

Hey HN! I enjoy writing, but find the act of driving a piece to completion taxing and occasionally paralyzing, and I'm unsatisfied with the quality of work that LLMs produce alone - they're okay at summarization but don't always preserve intent or voice.

Prospero is a Claude Code plugin inspired by Superpowers - it focuses a great deal on interrogating your argument and helping you build a spine that's staged into an outline, which is then refined and drafted into a final form using your voice. Prospero also includes research and critique phases: a critic runs in a subagent, does independent research, and tries to find the holes in what you've written. The critic is pedantic to a fault but useful. Give it a poke.

I built gpu.fund as a small live price board for GPU rentals. It tracks cloud inventory and hourly prices across providers so you can quickly sanity check what H100s, 4090s, 3090s, and smaller cards cost right now.

Mostly made it because I got tired of opening six tabs just to answer one question: what is the cheapest usable GPU today?

Feedback welcome, especially from people renting GPUs for training, inference, or weird weekend experiments.

A few weeks ago there was a thread about using AI to finish abandoned projects, and a comment from avereveard https://news.ycombinator.com/item?id=47905088 on building their 100x100 grid battleship got my spark. I didn’t build that, but it pushed me to finally finish a combat idea I’ve been wanting to play for ages.

Introducing Naval Strike!

It’s a simultaneous turn-based fleet vs fleet on a grid played in the browser. There’s no account or signups needed. Both players plan their moves, then the turn resolves at once.

There’s three different play styles:

* Solo. Procedural maps and uses guided “AI” opponents (in the same way that 1990s games had “AI” opponents)

* Scenarios. Objective missions on procedural maps like "rescue the downed pilot" or "destroy the convoy"

* Campaigns. Historical battles on real-world maps (Sink the Bismark in the Atlantic, or escort the tankers through the strait of hormuz)

A few things I figured HN would ask:

I designed the architecture and Claude did implementation. While 90% of the decisions are mine 90% of the lines are AI-written. Lots of micro managing short bursts to get it to look/feel right. I think I consumed a week’s worth of tokens just to get the fog working how I wanted. Eventually I learned that having lots of small bursts of code with testing got me to where I wanted much faster than longer sessions. I built preview pages so I could test animations, sequencing etc without affecting the codebase so I didn’t chew through my tokens.

Stack. TypeScript/Canvas 2D, hosted on Cloudflare. The server is a tiny WebSocket relay, it pairs two players by room code and blindly forwards messages, with no game logic. So there’s no accounts and no tracking beyond default Cloudflare insights. Just open the URL and play. Opponent AI is guided scripted tactics with situational decision trees.

Art is a mix of AI/hand. I’m not artistic enough to do everything by hand, but there’s a lot of manual pixel by pixel editing on the assets. Assets are stored as JSON arrays and drawn directly to screen with a colour palette.

I ended up building a small map editor that lets me "trace" Google Maps screenshots to get the campaign maps geographically close to the real engagements. Sounds are from open source libraries - you can mute them with the little speaker button but they are on by default which might upset a few people.

Although this was largely coded by AI, I got heaps of enjoyment being able to focus on the UX, style, gameplay and UI to be just how I wanted. Being able to test (and throw away many!) ideas so quickly was awesome fun.

Feedback welcome. Especially on balance and the campaign design and gameplay, it’s hard to play-test every variable!

Every time I needed to inspect a database, I either had to wait for a bloated GUI client to chew through my RAM or struggle through psql, which isn’t great for actually visualizing data. So I built a TUI database client focused on one thing: connecting fast and letting you browse tables as datagrids and rows as JSON, with near-instant startup and vim-style keyboard navigation.

Still in the early days but already handles most of what you would expect from an SQL database client.

I've been wondering how I could use LLMs to help me write, without taking my own voice away. I arrived at a workflow where the AI has strict instructions not to give me any text, just to give me tips, but it was clunky to see which parts of the text the critique referred to.

To solve it, I made Lucid. I made it mostly for myself, but I added a "bring your own key" system for others to use it. I hope you like it!

Hi HN,

I'm Quentin, and I built something called Looties. Think Vinted, but for developers. It's a peer-to-peer marketplace where you can list and resell tech/company and conference swag. Hoodies, stickers, mugs, caps, onboarding kits… the stuff we all accumulate over the years.

This started pretty personally.

I've spent ~10 years in tech, the last 6 working closely with developer tools companies. Every onboarding, every offsite, every conference... I grabbed the swag. I'm a nerd at heart (gaming, DnD, the whole thing), so I genuinely liked that stuff. My laptop was covered in stickers. I now use a computer case to collect them and swap it based on my mood.

Honestly, there's something weirdly powerful about wearing merch from a company or ecosystem you admire. It's not just cotton. It's signaling. Belonging. "I'm part of this world." That's why I'm so attached to my K8S sweatshirt.

Then I left Strapi, and suddenly I had collector stuff from a brand I still liked… but wasn't part of anymore. That's when it hit me: some of this stuff probably means more to someone else than it does to me right now. And I knew community members (mostly web agencies) who would have loved to get their hands on that merch but couldn't. No public store, no secondary market, nothing.

I considered building something back in 2022. Didn't. Marketplaces are notoriously painful (payments, logistics, trust…), especially when users immediately compare you to mature giants.

Fast forward to late 2025. I was at PyTorch Conference in SF and joked to some colleagues: "I should build a marketplace for all this swag I've looted over the years." Instead of laughing, they said: "Honestly? I'd use that." What changed for me wasn't the idea, it was execution cost. AI coding tools + modern infra made it realistic to prototype and iterate on something like this solo without spending a year building plumbing.

So I built Looties as a side project (I still work at Pruna AI as their GTM lead). It's early. Inventory is still growing. It's live. I'm witnessing first transactions between the very first buyers & sellers

The launch, a couple of weeks ago, went better than expected. We hit #1 Product of the Day, #1 Product of the Week and #1 Product of the Month on Uneed (think PH but smaller. We even got a media interview from CMCM (France's biggest second-hand media outlet), and we're exploring a "Museum of the Best Loot" installation with Station F.

Right now, delivery is fully operational and battletested in Europe. Domestic US transactions is maybe 2 weeks away. It's built, currently in QA, looking for early beta testers.

Genuinely curious: have you ever had that feeling of "I have too much swag, this is becoming a problem", or the opposite, "I NEED this hoodie but there's no way to get it"?

Granite Switch is an open-source IBM Research project for composing several task-specific LoRA adapters into a single deployable Granite model checkpoint.

The idea is to get the accuracy benefits of multiple fine-tuned models without having to deploy and maintain a separate model for every task. It adds control tokens and a small switch layer that decides which adapter weights to apply, so different capabilities can be activated inside one model.

The composed model is designed to work with Hugging Face and vLLM, and the project includes ready-to-use adapters and pre-composed Granite Switch models.

Repo: https://github.com/generative-computing/granite-switch

I've written about a storage layer for sqlite before in my http://workdad.dev blog and had been using and improving it over time.

The result has been sqlflow, which at its core wraps sqlite transctions into Read() and Write() methods that takes care of the transaction life-cycle and avoids the most common issues with using sqlite concurrently (no SQLITE_BUSY errors for you anymore).

It pairs really nicely with https://sqlc.dev so if you use it, the query type it generates is compatible.

It also supports encryption and the multi-tenant pattern, where each user can get their own sqlite database file, both of which I use in http://trackm.net

For encryption I you can use the github.com/jgiannuzzi/go-sqlite3 fork to enable SQLCipher encrypted files.

A few days ago, I was looking for some obscure dev blogs and websites, and realized how completely useless normal search engines have become for finding "weird" content. Unless you already know the exact URL, almost every query just leads to SEO-optimized, commercial websites or AI-slop.

So I built Gulugulu to fix this for myself. It's a search engine that only indexes the old/weird web like digital gardens, Neocities pages, ASCII art, and personal projects.

There is no backend. It's a static site hosted on GitHub Pages.

You can try it here: https://cbrincoveanu.github.io/gulugulu/

The search runs entirely in your browser using Fuse.js against a single, flat index.json file. To get the data, I wrote a Python crawler that specifically scrapes curated indie webrings (like 512kb.club and Cloudhiker), extracts the basic metadata, and dumps it into the JSON array. Because it's completely serverless, there are zero analytics, no ads, and no cookie banners.

Obviously, loading a massive JSON file into browser memory has a hard upper limit. Right now, the index is small enough that client-side search feels instant. To scale it without melting mobile browsers, I'm working on a deeper crawler (depth=2) that runs URLs through an LLM to score their quality. If a site looks like commercial spam, it gets dropped before appending to the json.

I'd love feedback on the client-side Fuse.js performance. Also, the index is still pretty small. If you have a personal blog, a digital garden, or know any weird RSS feeds, please drop them in the comments and I'll add them to the crawler's seed list!

Most window managers and app launchers focus on moving/launching individual apps. As a person who switches context a lot, I wanted an app where I can save my entire workspace of apps and files, across multiple spaces and monitors and launch them in that exact configuration in one click.

Lattix can launch your entire workspace, including your apps, files and URLs, across multiple spaces and monitors, in one click. Spaces support was the most requested feature so far and with 2.0, Lattix officially supports launching across multiple spaces.

Please note that Lattix can’t create spaces so they have to be created before hand before launching your workspace.

Since Lattix is focused on becoming a workspace companion and manager, version 2.0 also includes two additional features:

Space Naming : Assign names to each spaces so it’s easy to remember and switch.

Space Hop: Jump between spaces super fast, using a modifier key and arrow keys/mouse movements.

Apart from that, Lattix supports hotkeys, custom layouts and one click workspace capture.