Ask HN: What are you working on? (October 2026) | Hacker News Reader

Ask HN: What are you working on? (October 2026)

10 pointsmeerita30 comments

What are you working on? What have you been curious about lately?

I got introduced to pico8 during Covid and absolutely loved it. It piqued interest in me to create sprites and that led me to create a very basic minimalistic sprite editor for pico8.

I first published the app a few years ago in 2023 and lately updated it in the past few months. I am planning to add animation next but still struggling with to keep the UI as simple and minimalistic.

Here is the App Store Link: https://apps.apple.com/us/app/pixel-edit-lite/id1665723146

I've been planning to get back into flight simulation but was dismayed to find that the latest Steam client is a Chromium Embedded Framework that runs at all times (which I find great irony in considering Gabe Newell's famous "service problem" quote), so I used Claude to yak shave my way to a native Win32 frontend (and when I'm done tinkering with it, my Flight Sim). It's also (dubiously) portable.

https://ibb.co/ccPf9Yqq

This is the kind of personal task I think LLMs are the best at solving.

I've been working on the MinPin platform (https://minpin.me) with Claude Code, trying to get my self-hosted server rack to at least partially pay for itself. Right now I've got a reusable framework I can use to sell various services:

* IPFS pinning

* Photo storage and backup with Immich

* Monitoring and Alerting with a status page

* Personalized browser history search with Hister

All the apps SSO against my identity provider so switching between them is pretty seamless, and billing is unified: instead of a bunch of independent Stripe charges, one subscription can grant access to multiple applications. One app is $5/month and four or more are $8/month, billed as a single charge, so Stripe's fixed 30¢ fee is paid once rather than four times.

I'm going through the "one (to|two|too) many" problem right now: picking the next app and working out how to present a growing list of products without the site turning into a junk drawer. Curious how others have handled that.

I’m working on Lexicon (https://lexicon.zone), a vocabulary builder built around fun word puzzles.

There are eight games to sink your teeth into, including my own take on a daily challenge and original games like Cascade and Flipside.

It’s for people who enjoy a daily word puzzle, and proficient English speakers who want to discover new words or feel more confident using ones they already recognise. All eight games draw from a carefully curated and edited collection of words and definitions, so the more you play, the more familiar they become.

When you come across a word that's interesting or useful you can save it to your own Lexicon and revisit its meanings, example sentences, synonyms and antonyms later.

I am working on a new kind of sorting "algorithm" for when the cost of comparisons is high and heterogeneous. The ideal use case is as a methodology for minimizing the number of pairwise human evaluations: in AI, marketing, search evaluation, clinical trials, etc... It works by keeping track of what pairs are solidly established, which parts of the ranking are still ambiguous, what alternatives are still possible/impactful and from this deduce the fraction of the sorting that can be reused during updates.

Think of it as reversible incremental ranking with information-theoretic guarantees and updates that scale linearly with the number of items to rank. From this I realized there was a direct application in machine learning: differential sorting and ranking. I've been exploring this space for the past couple days and it seems I could be flirting with state of the art performances. To sort 8000 items, softsort takes 1.9 secs to complete one forward and one backward pass. My implementation takes 27ms. 488Mo for softsort. 0.18Mo for me. Sorting 10^6 items costs about 100ms and 15Mo. Softsort can't run at this scale. Alas softsort isn't SOTA, fastsoftsort is. We both run in O(n log n) in time and O(n) in space, and have very similar raw performance profiles except that my thing takes 3x as much memory ... which allows me to update the ranking in linear time on the next runs. Mmmh!

I'm working on a book starring the 6 Simple Machines. I already did one book called "Simple Machines Made Simple." I'm now doing a follow up book that I'm thinking will likely be titled "Simple Machines Made Complicated" (or maybe Complex). Each page is a working mechanism. Kind of like a book full of fidget toys. So I want to have one page for each simple machine, and some of the progressions to complex are easy i.e. (wheel -> gears), but the one I'm currently stuck on is the pulley. My first book had a compound pulley, but now I'm struggling to figure out a more complex pulley example that I can recreate with less than 5mm for the page thickness. My first idea was to do like a roman blind mechanism with a pinch-stop type thing for the cord, but the more I played with it, the more I felt like this wasn't any different from the first book. I also like the idea of a chain hoist type mechanism, but I'm worried there isn't enough visual pay off for what you put into it. Like you move this chain thing around a bunch, but you only get a very small movement on the other end. Anyway, that's I'm working on. It might take me another 6 months, but I'm sure I'll figure something out ;)

I'm building Eigakan Guide (https://eigakanguide.jp/en), a cinema showtime aggregator for Japan. It allows searching for showtimes all over Japan by area, cinema, date, time, and features like IMAX and Dolby Cinema.

When I posted in the last "Ask HN: What are you working on?" thread I only had support for two cinema chains and no English metadata. Right now I've got 27 cinema chains, 373 cinemas, almost full English metadata coverage, and I've added English audio and English subtitles filtering.

So you can basically search for stuff like movies showing in Shinjuku on Saturday starting after 6PM, finishing before 11PM, with English audio in IMAX, and it'll give you every showtime from every cinema in Shinjuku grouped by movie.

I'm currently building a ChatGPT plugin which will allow searching showtimes through ChatGPT using natural language and later on I will add AI chat search features to the actual site as well.

My goal is to make finding movie showtimes as simple as possible and I would love to get feedback or suggestions for new features people want.

I am working on my very own Web App for Todos and notes. I wanted to take a real new take on todos, because most apps I used show a simple list and does not give me the option to see what is really important right now and how big a task is. Also tools are very different the way they work with subtasks and projects. Most tools have separate concepts for each (or none at all). So I decided to let the user decide for each task what it is: every task can have nested tasks layered as deep as it wants and there is a grouping feature to work smart with these tasks. It support also notes, links and files. I use it daily for work, it is in active development (new version coming out today) and it also has a Mac app. Feedback is more than welcome. You can try it here: https://aevo.so

I’m working on local GitHub-like app where you can comfortably review code written by your agents before the code will ever leave your computer. Once installed, your agent can send you a link to GitWarren each time it will finish a piece of work for you to make a review. The idea is to replicate the GitHub code-review experience.

Free and open source: https://github.com/klarluft/gitwarren-app

It works as a local WebUi or desktop app. Also works well with tailnet and ssh if you review a code on remote machines. If your phone is in tailnet you can make reviews on your mobile.

I've gotten quite far on my free, open source, language-learning tool for teaching Brazilian Portuguese. It's flashcard based, focused on verb/vocabulary-drilling, where you get the English version and have to produce the Portuguese. There's also a "Foco"-mode, with smartness in what flashcards to drill per day.

The site is completely static, no build step, and even offers optional cross-device progress sync between devices and browsers, through a tiny worker keyed by a sync code, totally account-less. Works on all platforms and browsers, no install required.

I've found it super helpful for my own PT-BR progress!

Website: https://falagringo.com/

GitHub: https://github.com/henrikhestnes/fala-gringo

I'm making a browser-based MMO experience of sorts for the 1999 Dreamcast game Shenmue - it's still pretty rough around the edges but people who like the game have liked it so far:

https://www.newyokosuka.com/

I'm also close to finishing an browser-based client for the MMOFPS Planetside:

https://terrasunder.net/play/

I'm getting reports of performance issues for TerraSunder from a couple people so if you run into them, submitting a performance report in the settings would help me a lot!

I have some other cool projects listed on my site here, including a browser-based version of EverQuest that plays itself: https://brynnbateman.com/

I keep working on

https://hn.fiodorov.es/ - an agentic search across HackerNews comments

https://climate.fiodorov.es/ - a climate agent with access to climatic data for many cities

https://lyrics.fiodorov.es - a way to search and translate song lyrics using an LLM

A Jellyfin Plugin: https://xn--teilifssen-04a4l.ie/

Will make your Jellyfin go from a Netflix style presentation to an old school TV channel surfing mode in the click of a button and back again.

It makes Jellyfin way more fun than Netflix as it removes complicated decisions from deciding what to watch. When you're not sure you click the Teilifís Seán button.

Im building serverless hosting for Wordpress (+ serverless mariadb) with the dev tooling I wish Wordpress had.. including mcp and cli that fully exposes the wp cli, mariadb, and all of the underlying infrastructure. So everything you can do through the dashboard, an agent can do via mcp or cli.

For example, if Im debugging a plugin, an agent can just connect to the site, read or modify the code, see what it's writing to disk or db, test it's changes, etc.

And this is standard WP -- this isn't a rewrite like emdash. So the fun part is figuring out how to make it all work without modifying WP, or requiring plugins to be installed, etc.

https://www.agiler.io

Working on an interactive explainer for code. When agents write most of the code, I find I'm losing the working theory of the code/architecture. Code review had typically served a couple of purposes, two of which are catching errors and disseminating codebase knowledge amongst the team. I find that adversarial agent reviews are good at catching errors, and that I prefer to learn about the architecture before/while I'm designing the next feature, rather than after.

I've also been designing a APL/BQN-like language that uses Korean as a way to compact and compose the meanings of the operations.

I’m building Habitnex, an iPhone AI habit coach that uses your calendar to help fit habits into your day. It’s still prelaunch: https://habitnex.io. How do you handle habits when your schedule changes?

I've been working on macOS RDP client that solve some of my gripes with the native tooling. Brings retina full screen scaling, auto reconnect, and measures network to downgrade if necessary.

https://kcirtapfromspace.github.io/mooring/

I'm building the Fillvisa: open-Source Alternative to Docketwise

So far, legaltech has mostly been closed-software. I'm building an open source alternative to Immigration tech software.

Github: https://github.com/athos2113/fillvisa-os

Live demo: https://fillvisa.com/demo

Docketwise: https://www.docketwise.com/

I've been working on a website monitoring service written in Rust. I had to port from using the reqwest library to libcurl in order to properly capture DNS/connect/TLS/first-byte timing. I basically ran the new engine side by side with the original for a few weeks, running 100k+ checks daily to get a 99.8% agreement. The mismatches were interesting because of the Cloudflare challenge that returned 403 based on the http/2 fingerprint, no matter the user-agent used.

I'm working on Aficionado (https://aficiona.do) - a scrapbook/tracker for media aficionados to remember, share and discover films, games, music, books and tv shows they enjoy(ed).

It's been a lot of fun working on it and using it myself, hopefully other might find it useful.

Working on an Apple Notes clone, but everything is a local Markdown file. Currently free, macOS (Silicon + Intel) and Windows. Has tags and priorities, can display many scientific syntaxes (mhchem, most TeX / LaTeX stuff, Excalidraw, PlantUML, Graphviz / DOT, Vega-Lite, and actual Markdown footnotes). Can show notes in vertical lanes like JIRA does. Free iCloud sync too. Still early alpha.

https://notesplus.io

A notebook for the interviews that leave you guessing. Record the conversation, get useful feedback, and prepare for your next round.

https://kcirtapfromspace.github.io/janus/

Laboratory.love, still a one-person side project, and the goal is still to make supply chains transparent with regard to plastic chemicals.

If you eat the same product every morning please consider kicking off a crowdfunding page to get it tested.

Check what's been tested so far here: https://laboratory.love/results

This is a really cool idea!

If you don't mind a little feedback:

- It took me a lot longer than it should have to figure out how to click into an item on the results list; I instinctually clicked on the card and then the photo of the item, assumed that they weren't clickable, then accidentally clicked into one before realizing that only the name of the item is linked. The grid cards on the main homepage don't have this issue.

- The visual design of the Child/Adult radio group makes me think that the white item is selected, but I'm pretty sure it's the black one (based on the fact that the Medium ratings turn to High when Child is rendered black). It's clear enough in the filter at the top that the item rendered black is the one selected, but it wasn't clear to me in the radio group.

Building the MCP for Voiden, the API IDE we open sourced a year ago almost. Added the ability to test APIs through AI also create MCP. Basically going headless

github: https://github.com/VoidenHQ/voiden

There are posted later in the month. Don't post a new copy on the 1st. The most recent is here: https://news.ycombinator.com/item?id=49686380

I think a biweekly post cadence versus monthly might be beneficial due to the thread locking after 14 days, which would ensure there is always an active thread for folks to share what they're hacking on (and they might not be ready or in a position for a Show HN thread). Otherwise, there is a period of time where you cannot share between the last thread when it closes and the next thread. 1st and 15th of the month, for example. Or perhaps allow these specific threads a longer active window before they close to new comments.

(n=1, ymmv)

looking at the post dates, there doesn't seem to be a schedule at all. it's posted randomly... posts on 9th through the 30th. Back in 2024, it was even posted regularly on the 1st.

IMO if there isn't a schedule, then posting it on the 1st sounds good.. since that's when the other posts like this are submitted.

I'd look deeper than the dates. There is a fairly consistent schedule posted by david927. Read the comments on the first few and you'll see some discussion about the whole thing with dang. Everything else is other people rando posting the same thing, whenever they feel like it.

So there is a schedule. But the noise from everyone ignoring it and posting their own copy masks it, which is exactly why I dislike these rando copies of the question.

> There is a fairly consistent schedule posted by david927.

My comment about the schedule is based only on david927's posting schedule. I don't see the schedule.. they look like random dates during the month. If you see a pattern there, feel free to spell it out for me.

I also couldn't find any discussion with dang. link?

Reply on news.ycombinator.com