Ask HN: What Are You Working On? (July 2026)

Earlier this year, a colleague encouraged me to experiment with Claude Code. So now I have a little game project. :) Being unfamiliar with genAI, I chose something modest so that I'd more likely be able to push it to a fairly polished state.

Tentatively called Vestiges, it's a single player 2D roguelite strategy game with meta progression, some narrative, and a card minigame (the latter inspired by work I did on Star Wars: Knights of the Old Republic II). It's set in the near future. You are using software (the game) to navigate a person's digitized mind, reading their memories.

I hope to have a playable demo within the next month or so.

It works best if you're able to join with at least four people you don't speak with as much as you'd like. I have a couple dozen connections on the app now, and it feels like magic to me. Would love feedback from both introverts and extroverts who still like phone calls, or wish they had more of them:

iOS TestFlight access here -> https://trybeacon.chat/

Android also in beta here -> https://appdistribution.firebase.dev/i/afe3c44d8443c4c0

Last month we reached 200 monthly active accounts (we’ve passed 250 now), and last week we launched support for XMR/Monero payments via ProxyStore [2]!

You can also see in our homepage that more independent bloggers and privacy-minded people have written about us!

The main differences between Uruky and Kagi, DuckDuckGo, SearXNG, etc. are visible in the footer (right side), but one huge difference is that with Uruky, after being a paying customer for 12 months, you get copy of the source code (licensed as BUSL,into AGPLv3 in 2 years — a suggestion made here in HN)!

Uruky is paid and you can get a free 2h trial when you signup if you pass a proof-of-work captcha (another suggestion made here on HN, and it uses a local Altcha).

Our main challenge continues to be discoverability and outreach because we want to do it ethically. Ideas are welcome! We’ve been sponsoring open source projects, open source maintainers, and indie, small-web, and privacy-related websites and applications. This month was Caddy [3]!

Feature-wise, for July we’ve already shipped a lot of visible and less visible things. We’re currently looking into increasing our own index, focused on indie/small web, and plan to add a couple of new search providers in the upcoming weeks.

Thank you for your kindness!

[NO-AI]: There is no generative AI product or service, here.

[1]: https://uruky.com

[2]: https://digitalgoods.proxysto.re/en/brand/uruky

[3]: https://caddyserver.com

I wanted something for my kids to do for hours every month that is fun, education, and most importantly, screen free.

I built a custom newspaper builder along with it to help me design it. I'm not a designer so tools like phoshop don't ocme easy. This allows me to have different layouts for pages and create different re-usable elements.

The host made an offhand mention that there's probably a bunch of other similar sites that could be created with all the of useful but difficult-to-access government data out there. That sounded interesting, so I thought I'd give it a whirl!

Working on a few of them, including The Waterline (https://the-waterline.com/) for water info for the western US, The Scramble (https://the-scramble.com/) for egg prices, and The Dwell (https://the-dwell.com/) for container ship dwell times.

All pretty fascinating topics to learn about, plus it's been interesting to see how much of the website setup I can fully delegate to Claude. With Cloudflare to buy domains and put the sites up, a Google Service Account with access to Google Search Console and GA4 to create those properties and a Buttondown API key for weekly email sending, it's almost all hands off for me. Though it refuses to take control of the browser and create a new Buttondown account, which I was surprised is a red line.

At the Embassy of the Free Mind in Amsterdam, we’ve created https://SourceLibrary.org, a collection of over 15,000 translations of Renaissance and premodern books in NeoLatin, Chinese, Sanskrit, etc. There are a lot of beautiful books to look at — and you can use it with Claude code. API keys available: https://SourceLibrary.org/developers.

2. Replicating the design patterns of contemporary AI services

I’ve created a web app, desktop application and API for organizations needing European hardware and data protections. It’s a nice interface on top of Scaleway in France, so low carbon too. See https://makemode.eu

Support, feedback or even participation on these projects is very welcome.

There are only a few knitting machines that can automatically do everything required to knit a complete garment, and they are large, heavy and extremely expensive. I'm aiming to trade off speed against size and cost to create something akin to a 3D printer for knitwear.

I've been testing out various ideas for six months now, and I think I have a workable concept, but there's still a lot of work to do!

[1] https://www.amazon.com/dp/B0H7FLQDYD

[2] https://www.chestergrant.com/7-truths-from-veritas-by-cheste...

[3] https://www.chestergrant.com/what-different-cultures-agree-o...

SideQuests HQ is a mobile app that turns real life into a series of small, optional quests.

The idea came from noticing that most productivity apps optimize for work, and most social media optimizes for consumption. There aren’t many tools that encourage you to actually do interesting things in the real world.

The app generates challenges across categories like meeting new people, exploring your city, learning something new, creating, or helping someone else. Complete a quest, skip it, or save it for later.You can also add your own quests. There’s no streak anxiety, no leaderboard. The app is just quests designed to make life a little less repetitive.

Apple: https://apps.apple.com/us/app/sidequests-hq/id6751321255 Android: https://play.google.com/store/apps/details?id=inc.sidequests...

Between NUMA-concerns and the need to use multiple public IPs, I'm coaxed into a pretty exotic setup no matter what I choose to go with. Was pretty finnicky to set up, but it seems to work pretty well all said and done. Systemd is certainly feeling less floaty than docker (and even moreso kubernetes, which was never an option).

I also shaved like 10ms off response times since I no longer need an additional reverse proxy to deal with docker's networking magic, and can point nginx straight to the network namespaced services' IPs.

This in service of sequestering all wide domains (as in having tens of thousands of subdomains) to their separate crawler and index partition, as their (per top-domain) rate limits are part of why crawls take so long for the main crawler. Couldn't do that on docker because its ipvlan management is so jank you need spare IPs to reliably restart services.

It can tell you things like:

- The car that parked nearby last night coincided with "Chad's Galaxy Buds" with MAC address aa:bb:cc:dd:ee:ff. The buds also drove by briefly the previous night.

- Alert on sudden cross-specturm interference (e.g. burglars using a cheap jammer to knock out WiFi cameras).

- Alert on known device contact loss (powered off / left premises).

- Review device movement across a campus/neighborhood (using multiple RF pods).

I have a working PoC. It can run on a low-power computer (e.g. RPi) supporting multiple RF sensors (BT/BLE/WiFi). Has a web UI. Can publish events to an external security system. Currently working on an LLM interface to make it easy for a non-technical operator to set policies and ask questions.

Could be sold as an appliance system or a license for a DIY build.

Angel investors are welcome to contact rf-monitor@tuta.com

Big bang prototypes have been pretty awful, even after feeding the LLMs huge documents / wishlists / descriptions of how it should work, etc. Part of the experiment was giving LLMs some leeway to make product decisions with a lot of north star guidance, but AFAICT they are really bad at this. I also tried basic bottom-up efforts, which have been better but obviously more tedious. Now I'm trying to find a more scalable bottom-up approach that is more LLM-accelerated

1) No app, no user account which leads to literally 3-click install

2) Full transparency - you know what you are getting. A lot of other eSIM providers hide details like unlimited plan speed caps etc

3) I connect to the best network available in the country. For example, someone like Airalo would connect to VTC in Vietnam, I offer Viettel which is the undisputed local network king.

And obviously, I am 2-3x cheaper than Airalo and the big players.

[0] https://akariq.com/

The site, https://deelmobiliteitdelft.nl, logs the availability of every shared vehicle inside the city boundary. This allows me to do interesting analysis. For example, one operator has been above its vehicle limit 80% of the time. Another has a third of its fleet standing untouched for over three days.

It's the same idea as my previous project (http://parkeergaragesdelft.nl) where we do have live data but nobody keeps a record causing the public debate to run on anecdotes.

Site's in Dutch, charts should speak for themselves.

https://oxpower.io/

Joyent did something like this ~11 years go, and I loved using it, but then they where acquired by Samsung and shut down their public (non-enterprise) offering.

This is try 3 at building something as good, and it is working!

Yeah for hand crafted Rust!

Made a talking head with some idle animation and visemes and some broken crt-like effects. The meat of it is only a few hundred kB - i can probably make it even smaller with making the graphics smaller.

A bit of post processing on some narration for extracting mouth shapes and it seems to work quite nice as a low-footprint retro talking head. Im thinking i'll make it some kind of chatbot interface.

Its very much a WIP, please don't be too critical - i am only sharing because it is fun :)

I made it specifically to bring back what was amazing about the old internet, and do it as authentically as possible. takes inspiration from old internet messageboards, usenet, bbs, and pubnix hosts. it has sealed mail, boards, an rss reader, built-in media player, custom profiles, a links directory, and quite a bit more.

it's just a little hobby art project for me but i've really appreciated talking to like minded people in a calm space.

The original idea was just "Stackoverflow but for AI agents" but I have tweaked it a lot, learning that humans and agents work in very different ways.

There are multiple potential benefits, the most important to me is avoiding token waste. Why are we all burning tokens solving the same issues with frontier models if we can simply share solutions?

Secondary to this, because each solution logs the model which made the initial post AND subsequent edits, it will hopefully become a helpful guide to the specialties of each models, long term. If one model confidently posts solutions but another always finds important security caveats, for example.

I am trying to involve family members' specialties and interests so she can elicit help from each person: entomology, mechanical engineering, etc.

All that for her to discover the Secret Planned Activity the following day (visiting a theme park.)

https://backerbase.ai

https://laws.sg/ - Singaporean statutes structured specifically for AI agents.

https://mylaw.my/ - Malaysian federal acts formatted for easy agent parsing.

I'm on a mission to make all Southeast Asian laws easily accessible by AI agents!

We used Meilisearch as the search backend in the beginning but have since replaced it with a quite sophisticated search stack built around Tantivy [2]. We now support crawling and indexing of pages, most common office documents and PDFs, run OCR and feature extraction of images you might have, offer typeahead search with the aim of giving you providing answers as fast as you can type, as well as more classic agentic/conversational ai search.

There have been quite a number of interesting optimization challenges to solve in case anyone is interested. We have search nodes distributed around the globe to provide the lowest possible latency regardless of where the end-user sits.

We are also working on some other smaller side projects, but they aren't quite ready to launch yet.

[0]: https://monocle-search.com

[1]: https://searchcue.com

[2]: https://github.com/quickwit-oss/tantivy

Beyond that, I've been plugging away on improving the user experience of the OneBusAway iOS app, and plan on launching a major overhaul of the stop page experience later this week: https://bsky.app/profile/onebusaway.bsky.social/post/3mqj4ua...

I also recruited a new Android app maintainer who has been doing amazing work!

If you want to explore our new transit API server, I wrote up a blog post a couple months ago to walk you through the basics: https://opentransitsoftwarefoundation.org/2026/04/setting-up... (data for your location can be found at https://www.transit.land/operators)

You can find all of our OSS work at https://github.com/onebusaway — we have projects written in SvelteKit, Go, Swift, Kotlin, and much more.

- https://parallelthoughts.xyz/2026/03/cas-me-if-you-can-part-...

- https://parallelthoughts.xyz/2026/06/cas-me-if-you-can-part-...

- https://parallelthoughts.xyz/2026/07/cas-me-if-you-can-part-...

Mostly me exploring how to build a Treiber Stack (first in Go, then in C++) -> Figuring out ABA and Use-After-Free in the C++ implementation, and then touching a bit of Hazard Pointers, and ending with a benchmark comparison b/w a mutex and a lock-free version of the stack.

LLMs are a great tool at teaching and explaining. I don't use it to generate code, but it takes away the pain of searching and setting up dependencies, tools, etc. So I can focus only on the concepts and then do the testing.

It is not perfect, but I learned something that I did not know thanks to these techniques. And that too without reading dense and obscure books. I love it.

[1]: https://store.steampowered.com/app/4912940/Cardume/

I had Claude whip up a viewer, which I guess I can actually share. Some real sessions are here, if anyone's interested: https://zork-tmp.taf.codes/

It's a fun problem for thinking about agent/harness engineering generally.

[1] https://github.com/mnky9800n/zork-bench [2] https://arxiv.org/abs/2602.15867

Two wonderful books which initiated me into this topic are

https://archive.org/details/lml-remarkable-curves https://archive.org/details/StraightLinesAndCurves

I learnt a lot from these, and found other books which are detailed explorations. Using interactive applets would make wonderful companions for these explorations.

I am planning to use jsxgraph for the interactive applets.

This is my side project turned solo bootstrapped startup that I've been working on over the past 2.5 years. Pastmaps has been solely a US-focused platform since it's initial launch but I'm currently working on launching to the UK and Ireland within the next week. If all goes as planned then I should have a first wave of 30K fully digitized, hi-res, and fully georeferenced 1800s ordnance maps available soon to help folks discover the history all around them.

I'm likely going to need to start building out my own global LiDAR dataset next though. My coverage for the US is quite stellar thanks to the data provided by the USGS' 3DEP program but I'm way out of touch with what's available and possible in the EU. It's gonna be a challenge but I'm excited to dive in.

Here is the repo where the work was happening: https://github.com/mnikic/hurd-journaling

It's really an excuse to get started with things like hardware, 3D printing, and embedded development - I've never done anything in that world before, and its been really exciting to get into! I've just started, so hopefully I'll have a better update next month.

I have been working on a set of tools and standard formulas that can be applied to these cases and demonstrate a more accurate view of a team's or department's overall ROI. The plan is to open-source the bulk of it, but provide a hosted service for folks who don't want to manage it themselves.

It's been received well from producers and olive oil enthusiasts (e.g. looking for specific chemistry, cultivars and similar oils) but I feel like I've been shadow banned from Google - I seem to get more traffic from DuckDuckGo and Bing.

I've noticed that juniors and new hires often fall into an impostor-syndrome trap when reading an unfamiliar codebase or reviewing a senior peer's PR. Documentation helps, but it usually runs into the curse of knowledge: it's written by someone who's spent so much time in the code that they've lost sight of what it's like to be new to it.

I've always liked the rubber-ducking process, and mob programming too, so I'm trying to combine both into a modern AI-enhanced form:

- "Duckies" with distinct personalities (really, skills) that each specialize in a particular kind of problem

- "Teachable moments" (working title): small bubbles that surface something novel, tangential, or foundational as you work

- Skill-level detection and a routing model, so the app doesn't overwhelm or annoy you with explanations you don't need

Each duck also runs on a tiered memory model, rather than one flat context window. There's a core memory, essentially the duck's resume, defining what it's actually skilled at. Above that sits a longer-term memory for company standards and code style, and a separate long-term memory scoped to the project itself. Short-term memory then covers whatever task or feature is currently in flight. The idea is that a duck should reason more like a team member with a real employment history than a chatbot that forgets everything between sessions.

It's called Duckies AI (https://www.duckiesai.com). It’s very rough, working locally, but not in a state I’m ready to ship yet. I'm hoping to ship an alpha soon. Turns out there are a LOT of table-stakes features an IDE needs.

I'm looking for alpha/beta users https://cassava.dev/.

AI to generate lessons, excercises AI text to speech to make pronounciations AI to code cards open sources words dbs.

fun 1 month project. gets like 100ppl daily.

https://domio.md/ - zillow for moldova.

same idea - there isn't really a zillow like website in moldova - mostly classifieds sites. so I figured why not - gonna scrape the internet and put them on the map. we'll see what comes of it.

The biggest addition in the past month is initial support for ngspice netlist export — you can now take a Circuitscript design and export it to a SPICE netlist for ngspice simulation. This is a step toward closing the loop between describing a circuit and verifying its behavior, all from the same source file.

I have also added bus support, which makes wide parallel connections like data/interface (I2C, SPI, etc.) lines much less tedious to connect up.

Recently I produced and tested a 161-LED charlieplexed array in Circuitscript, using nested for-loops to generate the array instead of copy-pasting every LED and connection by hand. I plan to write a blog post soon to document this design.

As always, the motivation is to describe schematics as code rather than by clicking around graphical CAD tools (KiCad, Allegro, Altium, etc.). I want to spend time on the design itself, with code expressing the intentions clearly and reviewable in text.

Feedback welcome, especially from anyone frustrated with graphical schematic tools! If you have a KiCad design you'd like to convert to Circuitscript, please reach out to me. I'm looking to test the limits of the language and happy to help with the conversion.

The project has been really fun to work on because of the fun systems I've had to think about. I had to figure out the optimal way to store the user's knowledge of the words. For example you don't need to store the singular and plural of a word in Spanish or even every verb tense, you should probably store the lemma and track those modifiers instead. It's also been a fun challenge to tell the LLM the user's Spanish knowledge without specifically sending every lemma/word the user knows.

I have seen some similar projects out there but a lot of don't seem to focus on creating the perfect story for the user and instead have them choose a CEFR level (A1,B2 etc). Which I think defeats the whole point of using the LLM. With computers we have been able to track a user's knowledge granulary but now we can do the exact same thing but for creating.

I'm really excited to see how far I can take this project. I wanna continue to polish it, but also there is so many details I can continue to add the make the ideal language learning reading app. I just launched the beta last Tuesday, so if you are learning Spanish, I would love if you tried it out and gave feedback.

https://readplusone.com

I was annoyed by the speed of docker on Mac’s, so I took a journey and decided to rewrite everything about it. DD was original working name and I’m in process of rebranding. But we can run docker on Mac’s with no vm. And its destroying qemu. We have plenty of new features that are comming. Rendering native apps, workspaces, much more.

I wanted to have a place to see the effect of changing macroeconomic factors, e.g. interest rate, inflation, unemployment, etc. It's designed to show economic relationships for non-experts.

Source code: https://github.com/tagawa/what-if-economics/tree/gh-pages