What Are You Building? Share Your Projects

Non-Profit based ACME CA in Europe.

We want to establish an alternative to Let’s Encrypt that is taking the core features and values from ISRG. That’s not based on “bad US!1!”, more then a alternative would strengthen the ecosystem. This also means to create an alternative ecosystem like boulder.

The current challenges are mostly about incorporating the non-profit and structure it right. So that it’s as open as possible.

Motivation? To help shape the security landscape and bring much wanted features that are not viable for Let’s Encrypt to implement. Viable describes that boulder would require major rewrites for it to get implemented.

Specially we want to provide SMIME and .onion certificates.

Inspired by Andrej Karpathy's excellent YouTube video on tokenizers, I used Llama3 to analyze all of GPT-4's 100.000 tokens and today I wrote a blog post about it! https://koenvangilst.nl/blog/analyzing-gpt-4-tokens

I’m building a crawler that crawls over 50k company websites directly for remote jobs.

It’s currently a mailing list you can sign up here: https://bloomberry.com/remote-jobs/ . You can pick and choose from any role you want including ML engineer, backend, frontend, as well as other non-engineering roles.

I started it to scratch an itch of mine but along the way others were interested in it so I decided to build something people could sign up for :)

Me and my wife are working on https://nativi.sh

It's a tool to help you improve your writing in your second language. We're still tweaking the input pre/postprocessing and prompts at the moment and would love some feedback!

Motivation for the project: my wife is French, I'm still not fully fluent and I sounded like a goof on text in her family's group chat so this helps sanity check my texts without bugging her for review.

I have a small side project called AutoPTT [1]. It's a Windows app for voice communications, mostly aimed at gamers. It can press and/or release your push-to-talk key automatically based on voice activity because not all software/games support voice activation natively.

The reason I made it was the reason I make most things: because I needed it myself. Making it was not that difficult, though I must admit I've never been too fond of Windows C/C++ APIs.

I have a feeling that the website could use some work. It's certainly not flashy, and mostly just text. Would appreciate any feedback :)

[1] https://autoptt.com

I'm building a Java IDE for education in the browser:

https://reportmill.com/SnapCode

I'm mostly doing it as a labor of love, though I have no real users so far. Not sure how to properly get the word out.

I'm building a language learning app for intermediate language learners.

https://www.langturbo.com

I'm learning Swedish and I found out that there are not many intermediate to advanced resources. Most language learning apps focus on beginners (rightly so, that's where the money is) so I built this first for me, but now I also have a paying users and people seem to like it.

Some new features I'm planning are SRS integration and pronunciation training with the "shadowing" technique.

Challenges: focusing too much on trying out new tech and finding a fair price point (still figuring this out)

I'm creating a hobby operating system in Nim (early stages). After an initial attempt[0] I decided to start over[1] and document my journey in building it[2]. The focus for me is on learning low-level systems programming and enjoying the journey, rather than building something production ready.

[0] https://github.com/khaledh/axiom

[1] https://github.com/khaledh/fusion

[2] https://0xc0ffee.netlify.app/osdev/01-intro.html

A language learning desktop application aimed at serious learners (especially at B1-C1 level) No gamification, no silly chat bots. Built from first principles after learnings from writing https://thehardway.guide

Main challenge: not be too perfectionist. I hate the hustling culture where each MVP has to be rushed in 2 weeks. It might be risk-effective, but it can't possibly be meaningful. I want to do the best I possibly can with the greatest care however long it takes, but I also don't want to waste (too much) time on irrelevant details and over-perfectionism

Git's subcommand aliasing, but for arbitrary tools. Git has the capability to configure aliases, distinct from shell aliases, for its commands. E.g. 'git cm' = 'git commit'

I had a fairly naive, limited Bash script wrapping docker to provide something similar, but for fun I decided to rewrite it in Golang with arbitrary command support. Turns out that Go is not my cuppa so I rewrote it again in C# (my go-to), using the AOT compilation as well. Something like this probably already exists but writing it is about the journey, the fun of problem solving.

It will ultimately be a simple translator from a convenient config format (e.g. something like below) to shell commands, so nothing special. Just a useful tool for my day to day.

[command="git",exec="/bin/bash"]

        cm = commit

        cm <CommitMessage:alphanumeric> = commit -m "$CommitMessage"

        cm pop <CommitCount:int> = reset HEAD~$CommitCount

B2B News & Media site:

http://elnion.com/

B2B Podcast platform:

http://podnion.com/

B2B Video platform:

http://vidnion.com/

Creating a simple SaaS and library for doing blocking scheduling. “Find me the next open hour for Y” - sort of like calendly but for developers with a really easy/neat callback mechanism.

I’m building it because I’m tired of being oncall for lousy delay queues, which seem to always be built in house with no docs or tests. I think I’ve seen it homebrewed at least 6 times now.

It’s the smallest possible idea I could see myself getting across the line while on paternity leave. I guess project number #1 is baby boy #2!

Working on Honcho[0] a platform for personalizing LLM based applications to individual end users.

It was inspired when working on a tutoring service and realizing that the key problem of AI tutors was that they didn't know the student well enough. As we started tackling the problem other builders around us mentioned similar concerns. It's been pretty hard to explain the value to those not experiencing the same problems, but it's slowly catching on.

[0] https://github.com/plastic-labs/honcho

I'm working on a gemini web browser called astro (https://github.com/blmayer/astro), so far it has 50 stars, I'm super happy.

Also I have a SMPT server featuring automatic encryption using the WKD standard: https://dovel.email the idea is that people could self host their email easily. Maybe I'll make a SAAS for this.

My webring is mostly for fun and testing stuff: https://derelict.garden

Now they are hosted on my raspberry pi zero on my living room!

Working on Cardamom, a compiler that transforms trained ML models into C libraries for easy, simple embedding in applications.

I have an example use case of embedding a snake AI player (trained in Python) in a JavaScript version of snake for distribution on the web.

https://www.cardamom.ai/snake.html

I have a similar demo building for iOS, I’ve deployed ML as a lambda layer before, and I’d like to continue embedding in fun places.

I am building a site where people can ask for feedback on how they did in their interviews. I plan to extend it to all sorts of code reviews and/or interview feedback.

https://www.interviewblindspots.com/featured

People will post the question and the code they wrote in interviews. Experienced developers will give feedback.

Contributors/Cofounders welcome

Just released foldMation[0] earlier this year. It's both a virtual step-by-step origami creation and learning app.

- Inspiration - There should be a way to digitally fold origamis online. There really isn't one that works well so maybe I can?

- Challenges - Many. Understanding the math, working with 3.js, how to find points, edges and faces to cut so the user doesn't have to specify them, coming up with logic to deal with layering... The list goes on, almost endless. It's the hardest project I've tackled and one I underestimated the most.

- Motivation - Honestly, sunk cost fallacy. Always wrongly believing how close I am to where it needs to be. There's still many bugs, but at least there is no rush as they aren't showstoppers anymore.

Follow the tutorials in /create if want to know it more and /blog if want to read more on it.

[0] https://foldmation.com

Turbo - Bitly + Stripe for easier payment acceptance.

Copy, Paste, Paid.

https://tur.bo

Inspiration - Receiving invoices without a clear alternative to mailing a check.

I had need for a special purpose lock that solved this problem: There are `N` task sharing a single resource. It is incorrect for the `n`th task to acquire a lock on it until after the `n - 1`th task has released its lock. When the `N - 1`th task releases its lock the `0`th task may acquire the lock again.

I couldn't find any lock off the shelf that solved this, or a clean way to hack it with a fair mutex. So I wrote my own, and I'm pretty sure it's correct (1).

It's like a ticket lock, except the total number of tickets is known up front so every task can get one ticket. To acquire a lock is one CAS, where the lock is acquired iff the "current" value of the lock matches the ticket value of the task, if so we swap in a magic LOCK value. When the writer releases the lock it writes (ticket + 1) % num_tickets back to current so the lock can be acquired by the next task.

There's a deadlock condition though - if one task is cancelled while any other task is outstanding, no other task may acquire the lock, even if the cancelled task didn't acquire it!

To deal with that, the lock is poisoned with a magic POISON value. Then when any other task attempts to acquire the lock, but its value is POISON, then an error is returned and those tasks may be cancelled accordingly.

It's be possible to support adding/removing tasks concurrently by replacing the tickets with a circularly linked list of atomic pointers, so when one task is cancelled the "next" of the "prev" item can be atomically swapped, but I don't have a use case for that and it's annoying enough to write in Rust that I didn't want to bother with it.

(1) The code for this is here: https://github.com/m-hilgendorf/sequex/. I call it "sequex" for sequence-mutex. It could also be called a round-robin lock or something like that.

---

You could do this with just a ticket lock, where you take a ticket up front N times and then give one ticket to each task. When a task's ticket comes up it acquires the lock, but before releasing the lock, it takes another ticket. That's where I started, but I found this implementation cleaner.

I'm working on a personal hobby project of mine, written in MOO [0] of all things. Its a bridge to homeassistant - it connects to your local instance and lets you observe and control anything you have set up in homeassistant from within the MOO - toggle lights, set thermostats, send tts alerts, etc. It also receives state change events, so you can have objects that act like their real counterparts and display up-to-date information as long as the bridge is online. I run a local house MOO for my roomates and myself, and I thought it would be fun to control my smart home from it. Challenges... The first was reading the web socket rfc and writing my own implementation, in what is probably considered an esoteric language at this point. Next was discovering just how much of home assistant's web socket API is undocumented (read: pretty much all of it). There's a page which mentions a handful of APIs, and this got me started, but I really had to start reading homeassistant's code to see what other commands were available and being used presumably by the front-end to do some of what I wanted (listing users, subscribing to state updates for specific entities, listing areas, devices, etc)

[0] https://en.wikipedia.org/wiki/MOO

A small side project I wrote for my own use:

https://github.com/russellw/chatgpt-unpack

ChatGPT allows you to download an archive of all the conversations you have had with it, so you can do things like search for past conversations with your choice of search tool. The archive is in JSON format, not quite usable directly, so I wrote a program to unpack it to plain text.

Various GenAI experiments