Ask HN: What Are You Working On? (March 2026)

I made a SmallTalk inspired collaborative development environment for writing web components - https://playground.now.

A "world" is a locally stored CRDT that contains the source code and data you store.

The home page, login and world editing experience is all built from within itself.

I was talking about part of this problem this morning, I keep having it with our large OpenAPI spec and the JSON parsing. But it occured to me that recently sometimes Claude will work around the context blowing problem by using a sub-agent to do the parsing. Have you seen that?

It's still a big time saver to have something like this, and stops it even taking the risk of doing it in the first place.

I’m building a blackjack simulator/research tool called Blackjack Wonk. It started as a hobby project and turned into a deeper engineering rabbit hole than I expected. The goal is to make the simulations reproducible and trustworthy enough to test the kinds of blackjack questions that usually get answered with rules of thumb, anecdotes, or forum arguments. So a lot of the work is around deterministic runs, validation, configuration, and chasing down small discrepancies until I understand them. Very niche, but a fun mix of math, software, and curiosity.

Is identity preservation technically possible for LLMs, or only information transfer?

1. flopper.io - a resource to find AI infrastructure benchmarks and soon - compare providers. Reach out if you're a provider!

2. boulderinglist.com - a catalog of climbing gyms across the world

3. livedsupport.com - AI driven support for colorectal issues

4. radiusing.uk - find where people are willing to do something to improve the social fabric of where they live

5. llmstxt.studio - AI-SEO via sitemaps, llms.txt, and AI search

6. probe.bike - tell stories with your cycling data

need to find a way to get more sleep