ud0

I work at a FAANG and have spent the last 5+ years building video conferencing tools. I hacked together a Chrome extension to fix my video colour. I planned to add more features: local recordings, on-device AI transcripts and summaries, and clean screenshots of screen shares. Then my company shut down our in-house tool and moved to Zoom. Around the same time, I joined a new team and had to onboard, which is when I realised I truly needed these features. In technical meetings, I often want to capture a screenshot of a shared screen or record a brief explanation that transcripts miss, with permission, of course.

With Zoom, I'm unable to do any of it because of Admin controls. Many AI notetakers exist, but few run locally, and wiring a Chrome plugin into Zoom is messy. These things are easy & possible, so it frustrates me every time I join a call.

What do you find painful about video conferencing? And if you could design a Zoom/Teams/Google Meet alternative from scratch, what would it do for you?

Hi HN,

I’ve been working on a game idea inspired by Wordle, but with a unique twist: it uses your own photos to generate guessing words. Here’s how it works: the app picks a random picture from your gallery. It uses a small language model (SLM), running entirely on your phone, to identify a word from the image. The chosen word could describe an object, the mood, or any notable feature in the picture. You then try to guess the word, just like Wordle.

The app is entirely offline, private, and doesn’t require internet access. I’ve always been fascinated by the possibilities of small language models on devices, and I have more ideas I’d like to explore in the future.

I currently have a rough prototype ready, but developing this further is quite time-consuming as I also have a full-time job. Before investing more time into refining it, I’d love to know if this concept sounds appealing and if using your own gallery photos is something you’d find engaging.

Thanks in advance for your insights!

See screenshots here https://imgur.com/a/Rwsv7Kf

I am an experienced software engineer with decent knowledge of Python. I've used Machine Learning Models in the past to do stuff like background blurring etc.

What do I want? I want to be able to train models and use them for the applications I intend to build. I honestly do not care about the Math, statistics or theory behind them. I just want to know enough to be able to train a model, use a model and productionize it. Most of the resources I have seen on the web are bloated with information that I consider irrelevant and don't care about. Are there any resources where I can simply learn to do these things?