Brandon Hudgens

Owner/CEO of Agentic Solutions | ML/AI Engineer |
I like to build things |
2006 Time Magazine's Person of the Year

[Russillo @ryenarussillo]

This is what we mean when we say the league has never been as deep as it is right now.

[Brandon Hudgens @agentengineer]

Indeed. Always a scam.

[Mark Gurman @markgurman]

These plans are a pure gold mine for Apple. The vast majority of people will pay $240 a year for the next several years and never use it. That’s why insurance is a great business!

[Brandon Hudgens @agentengineer]

Very excited for this

[Mark Gurman @markgurman]

Power On: Apple’s first foldable iPhone is arriving next year and it will be its most un-Apple like launch yet. Here’s why — https://t.co/iFYiSb162n

[Mark Gurman @markgurman]

Yet still no Instagram for iPad 🤣

[Theo - t3.gg @theo]

Okay wtf is going on

[Greg Kamradt @GregKamradt]

We got a call from @xai 24 hours ago

“We want to test Grok 4 on ARC-AGI”

We heard the rumors. We knew it would be good. We didn’t know it would become the #1 public model on ARC-AGI

Here’s the testing story and what the results mean:

Yesterday, we chatted with Jimmy from the xAI team, who wanted us to validate their Grok 4 score. They did their own testing on the ARC-AGI-1 & 2 public evaluation set

To validate their score (and measure possible overfitting), we self-tested the new model on our semi-private evaluation set

We walked them through our testing policy:
* No data retention
* Model checkpoint must be intended for public use
* Temporary increase in rate limits for burst testing

They were on board, so we got started

Initially, we ran into timeout errors with normal requests, so we switched to streaming. That resolved the issue

So, what do these results mean?

First, the facts: Grok 4 is now the top-performing publicly available model on ARC-AGI. This even outperforms purpose-built solutions submitted on Kaggle.

Second, ARC-AGI-2 is hard for current AI models. To score well, models have to learn a mini-skill from a series of training examples, then demonstrate that skill at test time.

The previous top score was ~8% (by Opus 4). Below 10% is noisy

Getting 15.9% breaks through that noise barrier, Grok 4 is showing non-zero levels of fluid intelligence

But the mission isn’t over. We need new ideas to solve ARC-AGI-2. Scale alone won’t get us there

Come work on ARC-AGI with us

[Brandon Hudgens @agentengineer]

Where's the API?! You said it was live! @elonmusk

[Brandon Hudgens @agentengineer]

I knew it. The worst.

[Theo - t3.gg @theo]

Surprise! Grok 4 is not dropping on the API today. I'm sure it will happen in a few months...

[Theo - t3.gg @theo]

Surprise! Grok 4 is not dropping on the API today. I'm sure it will happen in a few months...

[Brandon Hudgens @agentengineer]

For anyone interested in AI, I can't recommend @natebjones YT channel enough. A refreshing voice in a forest of ill-informed channels

[Brandon Hudgens @agentengineer]

@markgurman How does @theo feel about this

[Mark Gurman @markgurman]

It is incredible that Apple design decisions developed over multiple years can be influenced by a week of Twitter and YouTube commentary.

[Russillo @ryenarussillo]

Had anyone ever done this before? Take two weeks off, then come back to work?

[Theo - t3.gg @theo]

I've used Pocket for managing my "read later" list for over 15 years. It shuts down today.

RIP to a real one.

[allen institute @AllenInstitute]

The fireworks in your mind. 🧠✨ This sparkling video shows the neurotransmitter glutamate being released into synapses, made possible by an indicator developed by @abhi_aggarwal1, @PodgorskiLab, and team.

#HappyNewYear #NYE

[Simon Willison @simonw]

Quitting programming as a career right now because of LLMs would be like quitting carpentry as a career thanks to the invention of the table saw.

[Chris @ChrissGPT]

Now the only question is did they achieve these with cons@n or no..

[Ash @AshsVerse]

Grok-4 benchmark leak just dropped.
-HLE: 35 → 45 w/ reasoning
-GPQA: 87 → 88
- AIME’25: 95
- SWEBench (Code model): 72 → 75

If validated, Grok-4 is flirting with Claude Opus territory. Release looks imminent. xAI is officially in the frontier model race.

[nil @dhvanil]

would you rather fight 1000 4o-mini sized agents, or 1 o3-pro sized agent?

[Theo - t3.gg @theo]

Talk is cheap, ship research and models