Ask HN: What Are You Working On? (June 2025)

[dead]

working on a cross web, self organizing digital footprint that serves as a human memory and identity for LLMs

https://github.com/saxenauts/persona

90 percent of the AI companion use cases today can work well with just a vector DB to retrieve facts, and chunks of memory, but a connected digital footprint would need a graph+vector hybrid.

memory in the coming future will not just be about fact retrieval but need backlinks of memetics, new streams of data, holistic analysis, infinite schema-less key value store, causal reasoning and other things that define "who and why" of a human and imitate neuroscience's understanding of how our identities work today. this then needs to be translated as language chunks to LLMs

benchmarking this against popular tools, on longmemeval. getting good results so far. i would love to learn from you guys, what's your take on identity and human representation for LLMs in the coming future

Compiler back-end for WASM and more, but with the core at a slightly lower abstraction level than WASM and with a somewhat novel ABI.

To abstract around register file differences in different ISAs, I'm using SSA-form with spilling to a separate "safe stack". Enforces code-pointer integrity for security's sake (not unlike WASM) but extended also to virtual method tables.

"Partial-ISA migration" allows a program to run on multiple cores with slightly different ISA extensions. "Build-migration" is migration to another build of the same program in the same address space: Instead of trying to debug an optimised program, you would migrate it to a "debug-build" to attach a debugger. Or you could run a profiling build, compile a new build using the result and then migrate the running program to the optimised build: something that previously only JIT-compilers have done AFAIK.

I'm out of the research stage and at the stage of writing the first iteration of the main passes of the compiler, but now and then I've had to back-track and reread a paper on a compiler algorithm or refine the spec. It has taken a few years, and I expect it to take a few years more.

I moved my blog over to Jekyll hosted via Github: https://blog.andrewrondeau.com/

The site itself isn't anything "special." I've had a personal website for about 25 years; the past few years I finally moved from making HTML by hand to using various CMSes. I tried a "no database" CMS that my hosting page had, then I wrote my own CMS, https://github.com/GWBasic/z3, to learn node.js, but then I had to go back because Heroku dropped the free tier.

Jekyll is interesting. As a Mac user, I'm surprised there isn't a push-button app, like MAMP, to just run it. Instead, I got exposed to some weirdness with Ruby versioning that, because I don't have any Ruby experience, was frustrating.

The default Jekyll template has warnings, but when I tried to fix them, I ended up jumping into a rabbit hole of sass versioning.

I also ended up jumping into a rabbit hole with setting up redirects from old urls on my blog to their new locations. I don't touch Apache / cpanel that often, so there was a bit of a learning curve for me.

One funny thing was that I set up two redirects, in cpanel, from the same url to two different urls. (It was a mistake!) I couldn't delete them, so I had to submit a service request with my host.

Two interesting things that I do not have time to do:

- Set up Github actions to deploy on my original host (andrewrondeau.com) - Set up redirects from blog.andrewrondeau.com -> andrewrondeau.com

Something similar regarding American history by mail was pitched as a successful business on shark tank this season:

https://www.youtube.com/watch?v=aA4h3WBeohc