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

Thinking of new ways to visualize market data in 3D like this helical candlestick chart:

https://cybernetic.dev/helix

I’m building AgentOS [1], trying to experiment where agent substrates/sandboxes will head next. It's a deterministic, event-sourced runtime where an “agent world” is replayable from its log, heavy logic runs in sandboxed WASM modules, and every real-world side effect (HTTP, LLM calls, code compilations, etc.) is explicitly capability-gated and recorded as signed receipts. It ensures that upgrades and automations are auditable, reversible, and composable. The fun bit is a small typed control-plane intermediate representation (AIR) that lets the system treat its own schemas/modules/plans/policies as data and evolve via a governed loop (propose > shadow-run > approve > apply), kind of “Lisp machine vibes” but aimed at agents that need reliable self-modification rather than ambient scripts.

[1] https://github.com/smartcomputer-ai/agent-os

I am working on a very simple browser extension that allows you to take notes right on youtube. It is open source, offline and everything is locally stored.

Here are the links.

Chrome store: https://chromewebstore.google.com/detail/video-notes/phgnkid...

Firefox store: https://addons.mozilla.org/en-US/firefox/addon/video-notes-f...

I actually would love some feedbacks and suggestions from someone here :)

Using Claude Opus 4.5 to query time-series data.

I have GBs of time-series data in a TimescaleDB database. It’s more complicated than this, but the gist is: I use natural language to ask questions about relationships in the data, Claude Opus 4.5 generates queries, and it finds patterns.

For example, I classify tens of thousands of news articles using different classification models. Then I ask Claude to write a query that tests for statistically significant changes in the time-series data at specific intervals after a given classification of article—and it finds patterns.

It passes train / test split validation. It will train on 2 years of data (2023 and 2024) and being able to effectively predict movements on the time series data using the classified news articles on the last year of data (2025).

I’m building a simple CSS redesign for Hacker News on Safari iOS as an extension: https://apps.apple.com/us/app/y-redesign-for-hacker-news/id6...

I have a few other small side projects that generally improve my day-to-day life, including a better calendar widget for shift workers and a video speed controller that floats on websites where I frequently watch videos for easy access.