Live data from Hacker News

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

news.ycombinator.com

321–330 of 1001 posts

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

#321
I fell down the rabbit hole of voice transcription about a year ago, always had a love for utilising fine tuned LLMs so have put two and two together and built https://whistle-enterprise.com. The biggest challenge being it all running on CPU with the target device being your low to mid spec office laptop that's a few years old (I5, 8gb RAM). All nicely packaged together in a single completely offline selfcontained app that you just install and run (no environment setups, packages to download, models to download etc).

One of the hardest parts I've found is the diarisation (who said what) side of things. Trying to tune this and have it working in a way that doesn't absolutely grind the laptop to a halt or take forever to complete has been _hard_ but also extremely rewarding.

Another part has been the fine tuning side of the Phi-4 model, I'm on version 10 now, getting that pipeline down was a journey in itself, but I've got some great results. I wrote a bit about it in a comment here - https://news.ycombinator.com/item?id=48385906#48389625

I absolutely love working on this, I still wake up and the first thing I think about is voice transcription pipelines (sad I know), but I'm excited to see how much further performance and utility I can squeeze out.

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

#322
- attempting to vibe-engineer a new JS engine to replace the dying Duktape (almost done!)

- native library to build TUI apps without the 20-60MB bloat of node/bun/go

- terminal coding agent harness focused on orchestration/loops

- a small scripting language that looks like JSX but has signals and render optimizations built in

- open-source software and hardware smart doorbell for a community space

- teaching AI how to write games for the Nintendo Wii

- designing an arcade cabinet

All of this over the past 4-5 months. AI is allowing me to deploy my short attention span very effectively! This is more than all I’ve accomplished in the past five years.

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

#323
Building DB Pro [1]

A complete desktop app for browsing and editing your Postgres, MySQL, SQLite data, creating beautiful dashboards, and soon designing automated workflows for repeat tasks.

[1] https://dbpro.app

I've kept a devlog of the last 10 months of building DB Pro, which has been the best way to bring users to the product. I'd highly recommend folks starting a devlog if they can.

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

#324

I'm working on Hiring Method ( https://hiring-method.com ). After 1.5 years of development and two exhausting pivots, I’m incredibly happy to finally have our v1 live! While most of the HR tech is rushing to use black-box AI, I built the exact opposite. It's a transparent, math-driven fitness engine. It extracts objective data from CVs and calculates how well applicants match requirements, letting you see the reasoni…

I'm curious how you're addressing any legal aspects about this:

> No black-box AI. Every candidate gets a detailed match receipt explaining exactly why they scored an 85%, complete with contextual evidence from their CV.

HR teams like to play dead when they actually have a file with detailed feedback on a candidate. Yet, they choose to keep that to themselves out of baseless legal fear. I wonder how that works out when somebody proves a company's filter consistently proves a specific bias gets rejected systematically.

and

> Automated assignment validation

which is particularly troubling for devs: companies scaling assignments as first screen. How do you get around "AI evaluating AI" loops especially about assignments ?

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

#326
I thought that applying AI on 1v1 competitive pokemon would be a fun and educational experience on POMDPs and trying out reward free models on a problem that would be classically treated as a RL problem. This was only possible thanks to a lot of foundational work from the open-source community and last year's competitive pokemon NeurIPS track https://pokeagent.github.io/track1.html - they laid out the plugins to connect policy models to pokemon showdown for live play and evaluation.

I have already finished training the standard discriminative auto-regressive architectures by imitation learning on player actions, compared it with previous baselines set in the study. I want to match or exceed the best prior model Kakuna @ 142M params, but in a limited budget. JEPA style world models are showing promise when conditioned on actions [1] and frontier research on JEPA with trajectory straightening [2] shows that improved planning is natural outcome of improved representations.

If any good research ideas come out of this exploration then even better!

Currently fork with my models: https://github.com/sooham/metamon (under checkpoints) Orginal source for pokeagents: https://github.com/metamon/metamon

[1] https://arxiv.org/abs/2603.19312 [2] https://arxiv.org/html/2603.12231v1

A good primer on world models from Welch Labs - one of my favourite ML youtubers: https://www.youtube.com/watch?v=kYkIdXwW2AE

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

#327
I have a few projects on the run right now

Writing a sci-fi book, and it’s finally fleshed out to a point that it’s slightly readable, though more like a script. As this is all in markdown in one folder, with some text files as lists, I started writing a simple web project to keep track of it all.

Created a website for local community information, services, etc. Something that removes the reliance upon social media for this. It’s static, making hosting cheap, and in most cases free as it can run on vercel with contentful for blogs and github to store it. I’m sure there’s another project like it, but it’s always good to practice making something myself.

I was asked to show something for STEM week at my daughter’s school. Started a project to demonstrate AÍ to children. Uses very small training data set, you can write the beginning of a one sentence story, it can keep track of a configurable number of tokens, generates a given number of them. Allows taking steps through the process. This is the only one I’m vibe coding because I’m not entirely sure on how to implement it, plus I’ve added multiple models.

Have been working on a girlguiding page specifically for the division my wife volunteers with, as they relied upon the older district site that’s woefully lacking. Stuck waiting for approval.

Post reply on HN