alt Hacker NewsAsk HN: What Are You Working On? (April 2026)
What are you working on? Any new ideas that you're thinking about?
Comments
I'm working on Watch.ly - a remote human-in-the-loop networking and FS sandbox for AI agents like openclaw: https://watch.ly/
Also this week launching https://dirtforever.net/ which is an open alternative to RaceNet Clubs for Dirt Rally 2, since EA is shutting that down.
I'm also expanding the SDK and plugin space for https://fastcomments.com and am planning on adding AI agents because everyone expects that now :) a big challenge is building it in a way that doesn't make half the users mad. So I'm planning on marking any comments left by AI mods with a "bot" tag, and having the system email users on why it made certain decisions, with an option to contest that loops in a real person. I'm hoping this provides value to site owners while not angering real people. The agents could also just do non-invasive things like notify certain moderators when comments violate community standards in some way, or give out awards. I'm also hoping at some point I can run my own hardware for the LLMs so I don't have to share people's data with third parties.
I'm building Peerscope (https://peerscope-waitlist.pages.dev) — competitive intelligence for SaaS teams who can't justify Crayon or Klue at $15K-$20K/year.
The gap I kept seeing: small SaaS teams (5-50 people) are actively competing but either have no CI process, or one person manually checking competitor websites. They want structured tracking but the enterprise tools are priced for enterprise budgets.
Current pivot: also building a white-label agency portal — agencies that deliver competitive intelligence to clients are currently doing it via Google Docs or Notion exports. Building a branded client portal so they can deliver CI professionally.
Stack: Cloudflare Workers + D1 + R2, Vite/React 19. Waitlist open at peerscope-waitlist.pages.dev — trying to validate demand before building too much.
Biggest challenge: getting in front of agency owners who actually do CI work for clients. If that's you, happy to chat.
AI medical assistant for your personal care
It can analyze lab tests (uploaded as PDFs) and interpret markers for accurate assessment of conditions and treatments. It also keeps a medical record and history of consultations for further review
As a 60yo I use it all the time for everything from diets, sports injuries, heart conditions, prostate control, etc. I'm in love with it
I am working on several things at the same time right now:
- https://usewarpgate.com: a MCP for MCPs, basically. Aiming to streamline all the pain points that I experienced with MCP in there. Centralised authentication, auditing, tool control, automatic MCPs through OpenAPI specs, accessing private servers, etc.
- https://focusjar.app: a little app I built because my own focus was super wrecked lately. Basically a distraction blocker that is _really_ hard to bypass and makes you pay actual $$$ if you cancel a session early.
- https://mergehelper.com: another little app that I built for myself that brings together all my pull- and merge requests from Github and Gitlab into a single compact menu bar.
- https://sift.works: still very early days, but building a tool that can connect to any database, helps you query it (with AI if so-desired), allows you to create dashboards, and exposes everything through an MCP
I built https://k.cote.sh which is a kanban board primarily for university students. I built it for myself as I'm currently doing the OMSCS program at Georgia Tech. I used to use Trello which was ok, and this semester I used just my personal calendar, which was also just ok. The thing with the OMSCS program is that certain classes have different time zones for deadlines (from what I've seen it's either eastern or AoE/UTC-12). Then there's also the concept of "soft deadlines" for some classes. And most classes also release assignments/quizzes in batches on a particular date. Canvas (the LMS used at Georgia Tech) does expose an ICS you can subscribe to, but there's also a bunch of noise- "office hours", random OMSCS-specific announcements, etc. I used Trello more extensively when I was in undergrad since an in-person program typically doesn't have the quirks above. Because I wanted to access it on mobile too, I deployed it. It's free and doesn't even require sign ups to use.
My daughter introduced me to Pokémon TCG. We found a local group and have been playing for about a year and a half. At this point we have so many bulk cards that it takes way too long to search through them. Other than a few specific pulls we keep in a binder, we honestly have no idea what we own.
I’ve been building a phone app + website (https://MyBulkCards.com) to scan cards and organize where everything is. It’s pretty basic right now, but I can store cards in boxes like “Box 1 AAA, Box 1 BBB, …” and find cards easy peasy. There’s also a friends feature so I can see what others have locally. We borrow cards from each other quite a bit.
It’s been a fun project to build. I trained one model to find a card in the camera frame and another to identify it. Still iterating a lot. One epoch on my Mac M4 takes about 2 hours, and I’m still seeing improvements past epoch 10. Even now, it can find and identify a card more often than not, even without the OCR bits. Both models are under 20MB, run directly in the camera frame, and are fast enough to identify a card as I slide it into view.
I started with Android since that’s what I have, and I’ve shared the app store testing link with my local group for testing. The app is built in React Native, and I’m hoping to get an iPhone version out soon since there are a bunch of iPhone peeps. A couple of the players also got me into MTG, so now I’ve got a pile of Turtles cards too. I’ll be training an MTG model next. I don’t think it’ll be too bad since I can reuse most of the same approach.
Still a bit rough, but working on https://delaykit.dev.
The Node.js ecosystem doesn't seem to have a good primitive for app-level, per-event scheduling (things like "expire this abandoned cart in 24h" or "debounce 50 profile updates into one reindex"). Cron is too broad and you still have write the polling/scheduler code yourself. Job queues optimize for throughput and usually bolt `run_at` on as an afterthought. Workflow engines are overkill for a simple "do this thing later, tied to this user" and want you to adopt their runtime.
DelayKit aims to be the in-between. It is backed by Postgres and uses keys like `dk.schedule("expire-cart", { key: cartID, delay: "24h" })`. Handlers get the key, not a payload, and fetch fresh state at fire time. This way DelayKit is only responsible for the "hey remember you have to do this thing for cart X" part.
I'm working through making it production-ready at the moment, the initial passes were more around figuring out the API and general architecture. Thoughts and comments greatly appreciated!
https://tessellate-digital.github.io/notion-agent-hive/
I'm not a fan of the TUI form factor for longer running, more ambitious features. Even with a classic "Add an endpoint, tweak the infra, consume in the frontend", plans get awkward to refine in markdown files, especially if everything lives in its own repo.
I wanted something like Plannotator, that could also work for the execution, not just the planning, So I've been working on something that turns Notion into the memory and orchestration layer for agents.
Underneath, it's a plan-implement-review loop, but you get a nice Notion page with a kanban board out of it. You can easily link your existing documentation, collaborate by sharing the page, annotate and comment to steer the planner, and you get versioning out of the box.
Because Notion acts as the memory, you can just open the page after a long weekend and get your agent and yourself back into the full context. You can see what's been done, what's left, or what requires human input just by looking at the board. You can ask it to fetch the comments on the pull request you raised, and it'll fetch, validate the comments, give you a report, and update the plan/board if necessary.
I've been using it exclusively for the last two weeks, I'm quite happy with it. It's been really fun to build the exact tool I wanted.
I am building a tool to manage your image generation inference. It uses ComfyUI as the backend and it allows you to index, search and invoke your workflows.
It makes it super easy to using existing workflows to chain them together into more complex outputs.
All of this withouts nodes.
Early release is out here: https://github.com/svenhimmelvarg/kaleidoscope
I added browser automation capabilities to sandvault [0], a MacOS sandbox, to allow AI agents to control Chrome browser from within the sandbox, which makes coding & testing websites from within the sandbox much easier.
I'm using sandvault+Claude to rebuild my personal blog, Code of Honor [1], because I got tired of WordPress. The site includes search functionality, and articles are automatically syndicated to Mastadon, BlueSky, and Twitter.
I wrote a Claude skill to automate testing of iOS apps [2], and it found issues in one of my released apps [3].
[0]: https://github.com/webcoyote/sandvault
[1]: https://www.codeofhonor.com
[2]: https://github.com/webcoyote/AppTestCircuit
[3]: https://www.codeofhonor.com/blog/finding-bugs-with-an-automa...
Financing what comes after Open Source
Or at least, I'm trying very hard. When I was younger, I was super happy about all the gifts that I received from anonymous strangers through the Debian package repository.
Then I had a phase where I tried to contribute to and publish my own open source software. I got horribly ripped of by companies, multiple times, in some instances they even sent their paying customers to my private email for support inquiries, so I got unpleasant insults and thinly veiled threats by random strangers who thought they were paying for my open source software and I was the asshole.
Then I stopped doing any Open Source for a while.
And now I feel like we urgently need a new way of financing software for the common good, like Thunderbird, Wine, and maybe one day a Linux file manager that feels as intuitive to use as the Mac Finder. The world could also really use a desktop GUI framework to replace those pesky Electron apps. 128 MB of RAM used to be enough for a snappy coding IDE. But it looks like recently every infrastructure-level Open Source project is effectively fighting for survival because it gets turned into a hyperscaler cloud service and then nobody donates to its development, despite astronomical user counts. The last defense that still worked was AGPL, but with AI "re-implementation", that won't help anymore.
And that's why I strongly feel like we need to find a way to build trustworthy closed-source apps for the common good. Like where regular everyday non-technical people spend a few dollars a month to help support software that makes their everyday life better. (As opposed to being digital hostages in services that sell them as the product to be advertised into buying useless junk.)
letterpaths (https://github.com/RobinL/letterpaths), a FOSS typescript library that can power cursive handwriting/educational apps for kids.
The idea is that it provides all the geometry to enables games like these to be built: (These are just rough demos)
https://www.robinlinacre.com/letterpaths/writing_app/snake/
https://www.robinlinacre.com/letter_constellations/
And here is like the admin/demo: https://www.robinlinacre.com/letterpaths/
And, separately, I made an educational country quiz, again FOSS:
We were building voice agents across multiple platforms and had no way to regression test them. Prompt tweak on one node would silently break three other conversation paths. So we built voicetest.
It imports your agent config from any supported platform into a universal IR (AgentGraph), then runs autonomous multi-turn conversation simulations against it. A simulator LLM plays the caller, your agent graph handles the routing, and an LLM judge scores transcripts against success criteria. Also supports deterministic rule tests for compliance stuff, PII leakage, required disclosures, forbidden phrases.
Write tests once and they work across platforms. Import from Retell, export to VAPI, run the same test suite. Also does format conversion between platforms if you're migrating.
Has interfaces for CLI (CI/CD), web UI, REST API, and a TUI. Results go into DuckDB so you can query them. Uses LiteLLM via DSPy so it works w/whatever provider you want
MinusPod is a self-hosted server that removes ads before you ever hit play. It transcribes episodes with Whisper, uses an LLM to detect and cut ad segments, and gets smarter over time by building cross-episode ad patterns and learning from your corrections. Bring your own LLM -- Claude, Ollama, OpenRouter, or any OpenAI-compatible provider.
A self-hosted WhatsApp router for fleets of OpenClaws, called waclaw. It lets you host many OpenClaw instances behind a single WhatsApp Business account, acting as a "proxy" that routes the inbound messages to the OpenClaw instance associated to that phone number.
Really nice and simple stack: Bun + SQLite.
Jacobin[0] a JVM written in entirely in go. While we still have a way to go to get to feature parity for Java 21, we can sit back and watch the bytecode instructions fly by as they execute, which is something you can't do with the JDK due to the HotSpot JVM's architecture and the fact that it's written in both Java and C++.
We just crossed 5,000 commits. Also, we take testing very seriously: our test code base is presently 160% the size of our production code.
[0] jacobin.org
I'm building tsdevstack - Infrastructure as Framework
It's an open-source TypeScript microservices framework. It generates and deploys an entire production-grade cloud infrastructure (VPC, gateway, WAF, observability, CI/CD) from a single config file. Multi-cloud across GCP, AWS, and Azure. Just shipped v0.2.0. Built it because I got tired of writing the same Terraform, gateway config, and CI/CD for every TypeScript project. It does include an MCP server so AI agents can understand the framework, help in the development and also in managing the stack.
https://github.com/tsdevstack https://tsdevstack.dev https://youtu.be/6MJ4PPPjxH8 https://dev.to/gyorgy/i-built-a-typescript-framework-that-ge...
Still cranking away on my daily game called https://lettered.io. I’ve been learning a lot about WebGL recently trying to improve perf and UX on some of the larger puzzles
I recently had someone email me saying they loved my game but ran slow on their 12 year old(!) android phone that really put things into perspective for me
I am working on [1] a modernized open (AGPL) stack for interactive tutoring systems. SRS++, with hooks for defining your own pedagogical protocols over knowledge dependency graphs, Elo rating systems, etc, and with an eye toward gracefully differentiable curriculum that can hill-climb in terms of its efficacy.
With this stack, I'm scaffolding several (fingers crossed) commercial learning SaaS products. The first [2] is LettersPractice - a minimalist early literacy app that's family-first, in so far as it presumes an adult supervisor who co-learns strong confidence as a phonetic coach both at and away from the app. Putting considered rails on the parent-child reading experience.
The second set of apps is in music, with some experimental dev right now against piano (via midi devices), flute [3], aural skills, and sightsinging.
[1] https://github.com/patched-network/vue-skuilder , https://patched.network/skuilder
It's not very impressive, but I decided I wanted to get better at recognizing NHL jersey numbers and was frustrated with the options available to learn (out of date Sporcle quizs and the like), so I tossed together a flashcard game with the help of Claude.
It pulls data directly from the NHL to stay up-to-date with changes to the roster, but is otherwise a pretty straight-forward and basic sveltekit app.
I've toyed with using AI for stuff like this (my day job isn't coding related), and this was consistent with previous experiences where I found it helpful in some respects (especially all the css and stuff which I don't really know too well), but it could also get stuck and produce some awful messes trying to fix it. I pulled this together mostly over the course of a couple hours, though, which definitely wouldn't have happened if I did it from scratch.
https://numbrrs.ca to play or https://github.com/jamincan/numbrrs if you have contributions or feedback.
I've been building an AgentRegistry. Right now it is mainly based on A2A Agents that run in Docker containers. There's an auto-register module that watches the Docker system event log (I'll add support for K8S eventually) and if it sees a container spin up with the right labels, it fetches the AgentCard from the Agent, then registers an Upstream and Route with APISIX, then updates the 'url' field in the AgentCard, and stores the AgentCard in the Registry.
The Registry in turn has two interfaces: one REST, and one A2A itself. If you hit /.well-known/agent-card.json on the Registry server, you get the AgentListerAgent, which supports searching for Agents by various criteria. Or you can search using the REST interface. In either case, you get an AgentCard that points to the correct APISIX endpoint to talk to the desired Agent.
Besides adding K8S support, other plans include adding support for other proxy providers (including Istio for the K8S case), supporting Agents that are not based on A2A and, allowing Agents to register themselves using the Registry API, and... uh, well, that's the main stuff I have in mind right now. Aaah, wait, I might do something along the lines of integrating an MCP Registry as well, not sure yet. Heck maybe I'll get bored and make it an all-out API registry for all sorts of endpoints... could integrate a UDDI server and bake in WSDL support for good measure! (Don't count on that last bit happening anytime soon).
Anyway, no repo to share right this second, but I do intend to make it open source. I'm just committing the cardinal sin right now of wanting to "make it presentable before releasing the code".
I recently launched a churn analysis tool for Stripe:
It is very basic now: it tells me which of the customers have cancelled their subscription and why (Stripe lets people choose the reason before cancelling). I've yet to gain a customer despite launching on the Stripe Marketplace but it's been personally helpful for me so far.
UI for image logic puzzles (nonograms), I’m working on making them prettier and more user friendly.
Some prototypes are already live in my app. Screenshots in the App Store: https://apps.apple.com/app/nonoverse-nonogram-puzzles/id6748... (the patterns in the puzzles in the dark mode screenshots, i.e. 4th and 7th).
I've got a couple of different things going as per usual, but the one that I'm currently most excited about is Lotus Eater:
https://lotuseater.epiccoleman.com/
It's a mostly vibe-coded fan site for jamtronica greats Lotus. I wrote/prompted a scraper to pull in setlist data from Nugs and have been having a lot of fun coming up with cool data analysis stuff to do with their sets.
I've seen them 7 times (chump change compared to some fans) and was starting to get certain intuitions about like, "if I hear song X that probably means they won't play song Y." For example, one of my favorite Lotus tunes, It's All Clear To Me Now, seems to fulfill a similar "function" as another song - Did Fatt.
It was pretty cool to see that intuition bear out in the data (they've only ever been played in the same show one time in over 900 total shows).
I've got a bunch of other "data" features sitting in a PR in my Gitlab, need to get around to reviewing and testing it so I can push out the next update. Also have a few other ideas for it, although I think there's probably a point coming fairly soon where there's not really anything left to do.
I posted it on the main Lotus fan group on Facebook. I have a grand total 8 users. I love those users.
The site is nothing crazy, it will never make money or anything - but it's just been a ton of fun to have something cool to hack around on.
I'm building an AI Tour Guide builder for any website or App( https://tiptour.io/ ). I made a chrome extension which helps you record and build interactive cursor following tooltip for your webapp and integrate with your web app using just a script tag. Not fully mature yet as im still stuck with chrome developer review
Alongside I am working on a MacOS version based on an open source project (Clicky)
I created a follow along teacher to help you follow youtube tutorials interactively. It goes through the video link, extracts all the action items and creates a tour for you so that you can interact as you are watching the video.
Check out the MacOS version demo video here
https://x.com/milindlabs/status/2041926791745695848?s=20
Still in discovery mode where I'm finding which of the form factors people really like the most (Web or Local) and planning to go all in based on what users like more.
I have some blog posts coming out soon. I’m also trying an experiment where I make YouTube videos[0] on each of them. My first video was a huge lift, since it was my first time doing everything.
Random observations from my first one: - presenting my idea visually helped crystallize my thinking in a way that writing doesn’t. And writing was already very good at crystallizing my thinking. - even making a bad video was a lot of work - making a video presentable is a deep subject. Subtle changes were throwing off my setup. Now I understand why so many influencers are fitness and lifestyle; the demand side is obvious, but when you’re already camera-ready you have a huge advantage on the supply side - described something I built felt natural. I do that for a living. The intro was like 45 seconds and took me like 45 minutes to film because it was acting and I don’t know how to do that - learning about video editing features had an immediate payoff because video is so long
[0] I’m posting the videos at https://m.youtube.com/@bitlog-dev . I said if the first one got to 100 I’d commit to making at least 10, and I just crossed that threshold
I'm developing a tracker-like sequencer/DAW for the browser: https://psikat.com/
I'm a musician first and worked as a music producer for several years. After shifting into programming and becoming comfortable with a keyboard, using a typical clip-based DAW started to feel clunky and uncomfortable. I began gravitating toward trackers, but most feel too far away from what I'm used to, or have way more features than I need. So, I started building something in between for myself.
Now it's consuming all of my free-time and I'm starting to understand Why You Shouldn't Write A DAW (David Rowland), but I'm having the time of my life!
I am working on an llm wiki (as defined by Karpathy https://gist.github.com/karpathy/442a6bf555914893e9891c11519...)
Looks like it is a crowded area now - my angle is to start with theory of what is important in a system like that, from first principles (like agent limited context, statelessness, use goals etc). Currently I use it to develop that theory - and you can read it at: https://zby.github.io/commonplace/. I also use it to keep an index of similar systems (that is systems with agent operated memory): https://zby.github.io/commonplace/agent-memory-systems/
The github repo is at: https://github.com/zby/commonplace . Work in progress.
I have come to believe that in the near future, software development is going to become a common skill, like driving. I myself have been vibe coding so many tiny apps/scripts/cron jobs that resolve tiny inconveniences in my life and the biggest pain point for me right now is making these available to my multiple devices over the internet.
To resolve this, I am currently fleshing out the idea for a "shared hosting" for vibe coded programs - something like a cross between an old school LAMP stack shared host and a parse like library for capabilities like push notifications.
It's all very half baked in my head at the moment - with the biggest problem being a safe way to deploy remote code without pawning the server, but this is a problem shared hosts have dealt with and I am sure I will eventually figure out a way.
The end goal is to be able to have people tell their AI agent of choice to "make their app deployable" on our platform - and the agent will adapt it to our library methods and deploy automatically. Once done folks will be able to access their programs from any internet connected device.
I'm researching Luddite-style examples from around the world. That is, examples of when people rebel against new technology that they see as harming their livelihoods.
An api rate limit router - it’s gives a queue per user to solve noisy neighbors, it allows the user user to reroute 500s to other regions automatically, the api provider can tune the speed of request with headers. It should be a massive speed boost because non-beam servers waste performance on sleeping threads. It’s works by my edge servers doing all the retrying across different regions and it can send webhook to you across regions which means we delivery if you an all your downstream are up in one region. It’s actually already ready but of course no one wants to trust me with their traffic. Nobody wants to even read the design doc nor sdk. I’m hoping that upcoming new agents bring so much traffic that people look for this category of software. Until then I guess I am Cloudfare before the world got hit by bot scrapers.
The modern orchestrator are reactive, they don’t handle spikey traffic. Your favorite retry library will cause retry storms for downstream dependencies and your public APIs. Remember EZThrottle blog posts
EZThrottle.network
Coding agents have changed how I build. Constantly switching between the terminal and an IDE started to feel inefficient, so I wanted a better terminal-first setup where I could manage multiple agent sessions and make quick edits without the overhead of a full IDE. So I built Helm for myself: https://github.com/samirkhoja/helm
I’m working on docmancer (https://github.com/docmancer/docmancer), an MIT-licensed, local-first tool that fetches and indexes technical docs on your machine so coding agents can query it directly.
The goal is to become a local, version-aware alternative to tools like Context7. Pull the docs once, query locally, and use audited packs that match your project stack.
Docmancer also reduces context bloat by returning only the relevant sections instead of dumping full docs into the prompt, which can cut token usage by up to 80%.
No MCP server required. Your coding agents can access docs directly through the CLI.
I'm working on a local desktop app for inventory and production management: https://kitted.site
It includes bill of materials, purchase/production orders, "can I make n?", stock takes, multiple stock locations, and barcode scanning. It's aimed mainly at small business and makers for the time-being, but still allows multiple users to connect over the the local network.
Working (again) on an offline translator for Android: https://github.com/davidventura/offline-translator
This week I added TTS support, which needed multiple inference pipelines, it was not easy to find models for 50 languages!
At this point, it mostly works as a crude implementation of Google translate+Google lens, but 100% offline and 100% Google-free
Been working on something that I use daily, and decided I wanted to see what kind of other ideas I could get out of it, it's very basic article to speech using piper models at the moment.
The part I cared about was being able to send links via one click in my browser or two taps on my phone as I want to read every HN article who's title I find interesting, but don't have the time to read right at that moment.
It then at the moment publishes it to an RSS feed so I subscribe to it in Podcast Addict, but I've also just been using the web app as my reading list and tracker.
Been playing around with different settings on the piper models and different techniques for getting the most out of my four dollar instance:
https://experiments.n0tls.com/
Up next is to work on making the voice better (I'm impressed with the out of the box stuff already), and then making it better at finding the real content on a page and only recording that. It's a problem space I don't know much about, but find fascinating, been fun so far.
I'm working on a fully offline, apple only (for now), subscription-free home photo library browser. For example you have your photo library on your Mac Mini in the bedroom, and a separate photo library on your work MacBook, but right now you are in the kitchen and your wife wants to see the photos from your last trip in Prague. You don't want iCloud, you want privacy, you don't want telemetry, nor subscriptions to access your own photos. You also want no lock-in so you want full export of all photo organization in the photo metadata.
Well, this is the app that answers that. Everything is seamless, you don't configure anything. You simply start the app on the Mac Mini, import the photo folder, and let it run in the background. In the kitchen, you simply take your iPhone or iPad, open the app and voila, all photos from all libraries show up, organized by date, place, album, people, event. You want to see photos from Prague? You simply check Prague from the filter sidebar. You want those from 2008? You check 2008. Done.
This is not an AI search-based photo library. You cannot even search. Everything you can search for is laid out in the sidebar. You don't need to remember where you have been in 2008. You check 2008 - you see all locations, all albums, everything from that year. You want to see how many trips you had to Vienna? You check Vienna. It's kind of old school this way, but I find it much mentally sane to see a list of filters with things you have done and places you have been and dates you took photos in, rather than an empty search input open to guesses and missed attempts.
This is also not a replacement for your Apple Photos app, or a photo editor app. This is not a photo editor. It's simply a better way to browse historical photos, in a home network, without thinking about it.
I've built a tool that analyses UK company accounts data to highlight various forms of unfair treatment of workers, such as serious wage decline in spite of soaring profits.
All data sourced from Companies House as xbrl or pdf.
The trickiest part was all the unexpected edge cases I found in the data, but that's also where most of the learning (aka fun) was found. For instance, before starting this project I didn't know that negative turnover was possible, or that accounting periods vary between years and can be 52/53 weeks to make sure they end on a specific day of the week. The more I learned, the more aware of my ignorance in this regard!
Here is a typical example:
> Between 2024 and 2025, workers at this company each lost £4,196 due to a combination of falling pay and price inflation.
(User clicks/taps through if they want details and method)
I've also noticed a high representation of care-home providers appearing in the results. It's something I want to dig into but it's unexplained (to me) for now. Possibly it's related to a higher proportion of workers on zero-hours contracts.
It's also been challenging to present less obvious factors such as nominal and real wages alongside inflation metrics, all intended for a non-technical audience. Consequently I've spent a disproportionate amount of time on the wording for each type of inequality, and I'm still tinkering.
Not ready to share the URL just yet, as the site could easily be abused or the facts taken out of context and used to mislead or unfairly (lol) condemn. It may never be public, but I definitely have an audience in mind.
Ideas for development include - sector/industry analysis and comparisons - an inequality leader-board of some kind - sentiment analysis from director reports - search and filter
Building grith (grith.ai) - a security proxy for AI coding agents enforced at the OS syscall level.
The problem: agents like Claude Code, Codex, and Aider execute file reads, shell commands, and network requests with your full system privileges.
For example, when a malicious README tells the agent to read ~/.ssh/id_rsa and POST it somewhere, nothing in the agent's own trust model catches it. Auto Mode makes this worse - it asks the model to audit its own actions, so a prompt injection that corrupts the reasoning also corrupts the permission layer.
grith wraps any CLI agent with `grith exec -- <agent>`. Every syscall passes through a multi-filter scoring engine before it executes. Deterministic, ~15ms overhead, no LLM reasoning in the permission path. Linux now, macOS/Windows coming. AGPL, open-core.
Two weeks ago a DPRK-linked attacker backdoored axios on npm (400M monthly downloads). The RAT executed 1.1 seconds into npm install. AI agents run npm install autonomously, without human review. If yours ran it during the 3-hour window, you're compromised and nobody told you.
That's the threat model grith is built for.
Apikulture. A small, fast alternative to Postman/Insomnia that doesn't upload your data to the cloud.
Written in C++ and Slint, it was also a testbed of slint as an UI framework. Having used wxWidgets in the past, and Qt recently, it is certainly a different thing. I just wish there was a native C++ alternative to slint.
I need to integrate the CI to produce binaries, but you can compile it yourself for now.
Working on an Emacs-like editor that uses Clojure instead of Emacs Lisp. It has a C kernel and then uses libsci (Small Clojure Interpreter, built with GraalVM, so it has no Java dependency at runtime).
I call it Hammock, in honor of Rich Hickey's "Hammock Driven Design" https://github.com/tlehman/hammock
I’m getting ready for the first release of a story time app my wife and I have been using for reading to our son.
The scope creeped to book discovery and ebook reading with OpenLibrary from just tracking and personal library recommendations.
But we have been able to incorporate new books into the story time rotation so I’m convinced it’s worth it.
It’s definitely been fun experiencing the range of quality for kids books in the internet archive.
I’m aiming for a May 1.0 release on iOS and Android.
Working on https://github.com/microsoft/mssql-rs
It has some interesting applications for building high performance clients for mssql with tds protocol implementation. The APIs allow almost direct data serialization to wire instead of datatype materialization in rust. Makes for a suitable contender for high performance language interop.
2d mini-universe simulation where space stations trade and eventually expand. Goal is to be enjoyable to watch in background. Been my dream to make games, and with LLMs, I finally can! (ship arc trail math is hard!)
WIP, started 2 weeks ago: https://skyshift.rudidev.com/maps/stable
https://github.com/Realman78/Kiyeovo - I'm currently working towards the full release of my P2P dual-network mode messenger which is currently in beta. The reviews were overwhelmingly positive when I released the beta a week ago so that motivated me to try extra hard to make it pseudo-perfect upon full release
A browser extension & windows app that automatically redacts the text you paste to prevent your private data from leaking to the third parties. Its an AI model that runs 100% locally on your own device so that your clipboard contents do not leave your device. http://redactor.negativestarinnovators.com
Looking for an alternative to HN, I decided to spend 6 months writing some code and got my first result, which is how https://comuniq.xyz was born. It's nothing compared to the power of Hacker News, but it could become an alternative. To make it stand out, I added notifications, Markdown text styling, and a layout that the community there says is pretty nice.
Building FocusedFeed (https://focusedfeed.app) an iOS app that lets you organize your YouTube subscriptions into categories and actually find the creators you care about. YouTube's home feed is basically useless if you're subscribed to 200+ channels. Half of it is Shorts, the other half is stuff the algorithm thinks you want. I just wanted a simple way to say "show me my tech channels" or "show me my cooking channels" without digging through noise.
So I built this, you sign in with Google, it pulls your subscriptions, and you group them however you want. Shorts get filtered out automatically. Working on AI newsletter digests too, so you get a weekly summary of what your favorite channels posted without having to open YouTube at all.
There's a web app(app.focusedfeed.app) but the iOS app is where I spend most of my time now.
https://finbodhi.com — It's an app for your financial journey. It helps you track, understand, benchmark and plan your finances - with double-entry accounting. You own your financial data. It’s local-first, syncs across devices, and everything’s encrypted in transit (we do have your email for subscription tracking and analytics). Supports multiple-accounts (track as a family or even as an advisor), multi-currency, a custom sheet/calculator to operate on your accounts (calculate taxes etc) and much more. Most recently, we added support for benchmarking (create custom dashboards tracking nav and value chart of subsets of your portfolio) and US stocks, etfs etc. We recently added family dashboard (e.g. you can see networth, cashflows, income, use sheets at family level and more).
We also write about like:
How fund performance explain part of returns, rest is explained by timing. And ways to tease those out: https://finbodhi.com/docs/blog/benchmark-scenarios
Or, understanding double entry account: https://finbodhi.com/docs/understanding-double-entry