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

Preparing https://www.dreamlist.com for the holidays. It's a private wish list and registry site that enables you to also group lists of friends and family into one page to share with grandparents or social media. So you can run family holiday giving + gift drives / toy drives / disaster recovery with no ads or spam or any other badness.

When I first launched it was the only wish list that didn't list your name and registry on search engines. It is has grown a lot since then, and added a lot more privacy features, and the best marketing has always been doing right by users (extremely rare these days). You can see by looking up any name + the name of a wishlist/registry site and see what shows up to check at any time.

I'm working on more products following the same rule - serve users as you would your own family.

I've have jumped back into maintaining and cleaning up OpenPaint, a project I originally started about 17 years ago to bring a simple MS Paint-like editor to Mac, Linux, and Windows. It never really got much traction over the years and there have been several faithful other projects that tried to achieve similar goals and ported to the web.

For 2.0 I'm trying to stay faithful to what made the original Paint but with a few quality of life improvements, like multiple tabs with image previews, cross platform support, and generally cleaning up a very old codebase that relies heavly on wxWidgets.

I've been using AI quite a bit to help maintain and modernize this 2.0 version but I want OpenPaint to remain a simple image editor that most people are familar with.

Still a work in progress, but it's been fun bringing a 17-year-old project back to life and keep the idea going.

https://github.com/murdockq/OpenPaint/tree/dev

https://murdockq.github.io/OpenPaint/

How would you get different behavior when you have completely decompiled the original binary and recompiled it to wasm? Shouldn't the result have identical behavior?

I am working on developing a distributed operating system for managing machine learning resources (the entire memory hierarchy, networking and compute hierarchy). I worked in O/S development early in my career, now retired and ML is a hobby I've invested in over the past few years.

I'm curious about:

- how to "upcycle" end-of-life NVIDIA GPUs, like K80, M40, P100, etc.

- how to offload MoE expert calculations to MCUs with NPUs, possibly FPGAs

- how to manage a 10G LAN-attached network of older Xeon servers with the above GPUs and USB-attached MCUs as a machine learning "fabric"

- do ideas from IBM's ACP/TPF O/S (now Z/TPF) carry over?

- do ideas from IBM's Workload Manager (business priorities managed resources) carry over?

- can I bypass end-of-life CUDA driver support for older NVIDIA hardware by using Rust/Vulcan?

- can I leverage the architecture of Freetoken (MoE caching/routing, Engram, KV-cache compression) and other approaches like Baby Dragon Hatchlings, Hierarchical Reasoning Models, Tiny Recursive Models, Recurisve Language Models, Multi-token predicion, etc?

- is anyone else intersted in actually building something like this?

I've explored many of these parts individually, now combining them...

SO-101 ARM robotics handling. Options for getting into robotics more expensive and less varied than I expected!