momojo

How did you eventually get funding after those initial rejections? What changed?

I was wondering the same thing. Maybe with the way gemma4 and intelligence density is going, they don't predict the need for NPU's?

I bet it's more about what he didn't do. Like how a stable marriage seems boring but is the accumulation of many many right (by necessarily genius) decisions.

I'm all aboard the "Apple is simply waiting for the models to get dense enough to run on their hardware" hype train.

They're poised to consume the market for the "I want AI, but I don't want to sell my soul" demographic that is ever growing. Sure, the AI gluttony continues, and the vibes tell me people are only more and more willing to shovel their lives into the maw, but my thesis is people only value fire insurance after they've bought the house.

Put my down as bullish. Apple hardware is currently the worst it'll ever be, and gemma4 and qwen3.6 are the least intelligence-dense they'll ever be. Buy up taalas or spin up your own hardware. I'm confident Ive only scratched the surface of Ternus' 5-year plan.

A quote from their nextjs writeup but I like:

> Most abstractions in software exist because humans need help...It's not clear yet which abstractions are truly foundational and which ones were just crutches for human cognition... We took an API contract, a build tool, and an AI model, and the AI wrote everything in between.

I've seen many a JPEG explainer, but this one wins for most aesthetic. The interactive visuals were also nice. My only criticism is the abrupt ending; should have concluded with the "now lets put it all together" slider.

I'm surprised the point/comment ratio is this skewed. There's so much meat in the post to chew on. I like your writing. This was one of those blogs where I can tell you spent a massive amount of time on the technical, but simplified it to layman's terms. I hope you keep putting out stuff :).

I have a couple questions:

1. I think this quote should be raising *many more* eyebrows.

> The astounding thing about Goliath wasn’t that is was a huge leap in performance, it was that the damn thing functioned at all. To this day, I still don’t understand why this didn’t raise more eyebrows.

You put a cat's brain into a dog's head and its still breathing! It didn't flatline immediately! Is yesterday's news? This seems like the biggest take away. Why isn't every <MODEL_PROVIDER> attempting LLM-surgery at this moment? Have you noticed any increasede discourse in this area?

2. You mentioned you spent the beginning of your career looking at brains in biotech. How did you end up in a basement of GPU's, working not in biotech, but still kind of looking at brains?

Again, great post!

This doesn't match my own experience. I dream of the day the stuff I don't find interesting can get automated but again and again I find myself having to do things by hand.

I wonder if this is similar to Chess and Go getting 'solved'. Hard problem spaces that only the biggest brains could tackle. Maybe it turns out creating highly performant, distributed systems with a plethora of unittests is a cakewalk for LLMs, while trying to make a 'simple web app' for a niche microscopy application is like trying to drive around San Francisco.

Great work!

> In practice, that means more logic fits in context, and sessions stretch longer before hitting limits. The AI maintains a broader view of your codebase throughout.

This is one of those 'intuitions' that I've also had. However, I haven't found any convincing evidence for or against it so far.

In a similar vein, this is why `reflex`[0] intrigues me. IMO their value prop is "LLM's love Python, so let's write entire apps in python". But again, I haven't seen any hard numbers.

Anyone seen any hard numbers to back this?

[0] https://github.com/reflex-dev/reflex

Reminds me of the Clearview AI controversy[0].

I'm not diminishing the ethics debate, but it's crazy to me how easy it was for two non-technical rich dudes in a garage to build Clearview AI (And before vibe-coding!):

  1. scrape billions of faces from the internet
  2. `git clone` any off the shelf facial-recognition repo

It was just a matter of when.

[0] https://en.wikipedia.org/wiki/Clearview_AI#History

I'm surprised at the lukewarm reception. Admittedly I don't follow the image-to-3D space as much, but last time I checked in, the gloopy fuzzy outputs did not impress me.

I want to highlight what I believe is the coolest innovation: their novel O-Voxel data structure. I'm still trying to wrap my head around how they figured out the conversion from voxel-space to mesh-space. Those two worlds don't work well together.

A 2D analogy is that they figured out an efficient, bidirectional, one-shot method of converting PNG's into SVG's, without iteration. Crazy.

@simonw's successful port of JustHTML from python to javascript proved that an agent iteration + an exhaustive test suite is a powerful combo [0].

I don't know if TLA+ is going to suddenly appear as 'the next language I want to learn' in Stackoverflow's 2026 Developer Survey, but I bet we're going to see a rise in testing frameworks/languages. Anything to make it easier for an agent to spit out tokens or write smaller tests for itself.

Not a perfect piece of evidence, but I'm really interested to see how successful Reflex[1] is in this upcoming space.

[0] https://simonwillison.net/2025/Dec/15/porting-justhtml/ [1] https://github.com/reflex-dev/reflex

I wonder if humans are any different. We don't have LIDAR in our eyes but we approximate depth "enough" with only our 2D input

> But that just describes basically everyone, none of us have no agency, but all of us are also caught up in larger systems we can't opt out of.

But isn't the drama between the billionaire heiress and her starving-artist lover more interesting than the lawyer girlfriend deciding whether she wants to marry her below-average-salary boyfriend?

Or maybe I don't understand your complaint.

Anyone have any thoughts? ARC-AGI (and 2) is pretty much the only benchmark of interest to me anymore, due to its abstract nature.

You mentioned "step change" twice. Maybe a once over next time? My favorite Mark Twain quote is (very paraphrased) "My apologies, had I more time, I would have written a shorter letter".

Does anyone else feel like they buried the lead?

> Omnilingual ASR was designed as a community-driven framework. People around the world can extend Omnilingual ASR to new languages by using just a few of their own samples.

The world just got smaller

> CUDA based ray tracer with HTML UI

I'm curious if you mean they're running a raytracer on the back end, and you interact with an HTML UI, or if it runs browserside, maybe via WASM. AFAIK CUDA isn't directly compilable to WASM (yet?)

One of my favorite modern dithering methods is blue-noise dithering (see Figure 3b): https://developer.nvidia.com/blog/rendering-in-real-time-wit...

The only catch is that generating blue noise is a roughly O(n^2) algorithm. Its not feasible to be generated on the fly, so in practice you just pregenerate a bunch of blue-noise textures and tile them.

If you google 'pregenerated blue noise' you find plenty of them: https://momentsingraphics.de/BlueNoise.html

> Now a lot more attention is paid to the quality of the training data.

I wonder if Google's got some tricks up their sleeves after their decades of having to tease signal from the cacophony of noise that the internet has become.

So the technology developed for gaming unintentionally built the foundation for an unrelated industry.

doing the lords work

I apologize for the slop but it came from a genuine moment of confusion over the author's seeming misquote.

What was the concept?

How did we get here?

I'm guessing first draft was AI. I had to re-read that part a couple times because the flow was off. That second paragraph was completely unnecessary too since the previous paragraph already got the point across that "context window small in 2022".

On the whole though, I still learned a lot.

I know this is polemical article is in good fun but chatGPT gives me the impression Descarte should not be counted with the others:

https://chatgpt.com/share/68dd758f-2bc0-8008-955d-a7dbd89399...

" Given:

- The blog offers no primary evidence for Descartes’s having a proof.

- Scholarly histories, based on critical assessment of surviving letters, treat the solution of the area problem as due to Roberval (and independently Torricelli) rather than to Descartes.

- The more carefully vetted sources place Descartes in the position of reacting to, or endorsing, Roberval’s result but not of originating it.

Therefore, the weight of evidence supports that the historical consensus is correct — Descartes did not solve the area under a cycloid; the blog's claim is likely an overstatement or misinterpretation."

sure

This excites me. Everyone wants to learn a new language, no one wants to slog. But I'd pay money to play Grim Fandango, and get the chance to ask "What does Morte mean?" and have the LLM, in real-time, tutor me. And if I get bored, I can game on, business as usual.

But just give the AI a list of words I currently know, and it can slowly slip in more and more vocabulary, and maybe full sentences.

I'm not a fan of mysticism. I'm also with you that these are simply statistical machines. But I don't understand what happened when understood transformers at a high-level.

If you're saying the magic disappeared after looking at a single transformer, did the magic of human intelligence disappear after you understood human neurons at a high level?