kveykva

ah okay, not just me

Did OpenAI obliterate the ability to use Codex from mobile the way you can use Claude Code on iOS?

Is there a specific explanation about how this is better or different than vega itself? https://vega.github.io/vega/docs/specification/

My understanding is that Vega was already an expressive DSL for visualizations and its probably already well spread through LLM training data.

Would you be open to sharing what you built for running the automation tests? I could really use this right now.

The example prompt "intense anime battle between a boy with a sword made of blue fire and an evil demon demon" is super clearly just replicating Blue Exorcist https://en.m.wikipedia.org/wiki/Blue_Exorcist

Even design wise this looks virtually identical to https://terragonlabs.com/

The scroll hijacking on this site itself erodes my trust in this being a browser I want to use because it demonstrates a lack of understanding of the web.

Max AI | SWE Fullstack, Backend, AI/ML Engineer | Hybrid In-Person (SF) | Full-time | $145k — $230k | https://www.maxcare.ai

Max AI is building the financial operating system for healthcare—turning the broken, error-prone billing process into a lightning-fast, autonomous experience, much like swiping a credit card. Our ambient AI agents handle medical coding, claim submission, and fund collection the moment a doctor finishes charting—no humans in the loop.

We’re addressing a $350 billion market where billing errors cause delayed treatments, higher healthcare costs, and massive medical debt. By eliminating billing mistakes, we enable faster access to care, reduce administrative burden on doctors, and protect patients from surprise bills.

Team:

- PhD/Postdoc AI scientists from MIT and Caltech

- Ex-VP of Engineering at Curative (led engineering for over 25M medical claims billed)

- Founders who’ve scaled products to 40M+ downloads and 150M+ monthly users

- A team of experienced engineers from Discord, Google, Facebook, Airbnb, Brex, and Figma

If you want to help build the AI infrastructure that will transform healthcare finance apply at: https://app.dover.com/apply/Maxcare/3684f186-f343-439a-9500-...

Max AI | SWE Fullstack, Backend, AI/ML Engineer | Hybrid In-Person/Remote (SF) | Full-time | $145k — $230k | https://www.maxcare.ai

Max AI is building the financial operating system for healthcare—turning the broken, error-prone billing process into a lightning-fast, autonomous experience, much like swiping a credit card. Our ambient AI agents handle medical coding, claim submission, and fund collection the moment a doctor finishes charting—no humans in the loop.

We’re addressing a $350 billion market where billing errors cause delayed treatments, higher healthcare costs, and massive medical debt. By eliminating billing mistakes, we enable faster access to care, reduce administrative burden on doctors, and protect patients from surprise bills.

We're a team of repeat founders and world-class builders:

- PhD/Postdoc AI scientists from MIT and Caltech

- Ex-VP of Engineering at Curative (led engineering for over 25M medical claims billed)

- Founders who’ve scaled products to 40M+ downloads and 150M+ monthly users

- A team of experienced engineers from Discord, Google, Facebook, Airbnb, Brex, and Figma

Traction:

- Launched in 2025

- Projecting $5M ARR in the next 2 months

- $2M in claims processed in the first 30 days (we earn 4–12% per claim)

We're 6 months in and just getting started. If you want to help build the AI infrastructure that will transform healthcare finance, reach out to: colton[@]maxcare[.]ai or apply at https://app.dover.com/apply/Maxcare/3684f186-f343-439a-9500-...

This pitches a lot but only seems to support a specific inpainting operation?

Their code base is also weirdly convoluted, with many crates and files just proxying over to others. I've read and written a lot rust, and haven't seen this style really. Could be nothing but its odd.

Have there ever been studies that demonstrate that those individuals don't simulate possible state transitions they'll go through in a different modality? I'd be curious if they visualize actions they'll take still, just not verbally.

This is truly depressing because the aspirations of tinygrad are so appealing in terms of being concise, effective and maintainable. Then, instead, they throw comprehensibility entirely out of the window.

Vector editing has a lot of CPU bound tasks. Evan has some writing regarding the transition to WASM https://medium.com/figma-design/webassembly-cut-figmas-load-...

There’s a 2023 article by the same author here it looks like https://timdettmers.com/2023/01/30/which-gpu-for-deep-learni...

The miniaturization of what are in some senses fired clay tablets for information storage would be a fascinating development.

Spline and Vectary are both in this space

You're not even close to baseline.

We're working on similar things! If you'd like to chat sometime I'd be interested. My HN username is also my twitter handle.

That is correct

The temptation to call something related to the replay functionality Sharingan is pretty strong now that I've used this name haha.

In what I think of as engineering terms: an agent is a long running service that invokes LLMs during its execution. In contrast to an LLM driven application where the primary function is synchronous with the end user (like chatGPT). There's a blurry line there but that's how I think about it.

AutoGPT and BabyAGI are probably the two most well known examples so far.

A significant struggle when building these types of applications is understanding and debugging behavior many execution steps deep. This tries to assist with that by giving a framework for structuring the way your agent runs.

Maybe a similar concept is breaking out a web application into services, or individual route handlers, rather than implementing everything as one massive loop that responds to events.

Currently it only supports openAI. I'm looking into patterns for supporting local LLMs! I think the best approach for that might be to allow you to update the api endpoint to your own and assume you're using something that mirrors the structure of their API.

The main escape hatch for everything right now are custom nodes though. But then you'll need to bring your own templating pattern.

I like the idea of the agent-protocol they define, I do feel that long term there might be a common interface for agent <-> agent or user <-> agent communication.

You could build a system that never invoked prompt nodes, and I think that would be equivalent! Re: timely dataflow, there are only hints of that in there right now - but I intend to expand on it.

The discord server linked from the github would work! I'd love to find out what the specific questions are. Alternatively twitter DM would also work @kveykva.

I hope that people will be able to build agents in a way that makes deliberate methodical progress towards them successfully executing given tasks. Part of this could be tracing alone, which is something that I explored using opentelemetry on my own agents. But I found that I wanted to be able to understand the specific structure of the agent as I iterated on it as well. So Chidori attempts to allow you to do that, while having some properties that I feel assist with managing complexity!

Agents that work 80-90% of the time rather than 30-40% of the time would be awesome. And I think part of that is being able to iterate on specific pieces of the agent's behavior in a deliberate way.

It's definitely a reference to it! It also broadly does mean thousand birds.

Hello! I'm the author. If anyone has questions or feedback I'd really appreciate it!

Ironic that the video of the vise is recorded handheld it seems