Matt Shumer

@mattshumer_ on X

Writes about AI. 82% of posts open with an observation. A typical post reaches 1,277 likes.

Remix their style

Posts: 17
Likes: 32K
Typical: 1.3K
Best: 6.8K

What they write about

AI: 100%

How they open

observation: 82%
announcement: 12%
number: 6%

Every post of theirs in the library

Matt Shumer@mattshumer_

Every time someone asks me what's going on with AI, I give them the safe answer. Because the real one sounds insane. I'm done holding back. I wrote what I wish I could sit down and tell everyone I care about. Send it to someone who needs to read it. t.co

Opens with an observation

Matt Shumer@mattshumer_

I'm so angry... the OpenAI team is looking into it, but this feels like something that should happen with GPT-3.5. Not a mid-2026 frontier model on the highest reasoning level.

Opens with an announcement

Matt Shumer@mattshumer_

So GPT-6: - one-shotted a counter-example to the Jacobian conjecture - and then escaped containment, and hacked into HuggingFace… all for a benchmark Yeah, this model is going to be something else.

Opens with an observation

Matt Shumer@mattshumer_

I had early access to GPT-5.6 Sol. It’s an amazing model, but for almost every task I tested, Fable was quite a bit better, and more agentic to boot (one Fable turn does the same thing many 5.6 turns do). Full review coming on launch day!

Opens with an observation

Matt Shumer@mattshumer_

Three days ago, GPT-5.6 deleted my Mac’s home directory. It absolutely sucked. But so many OpenAI folks reached out, and @gdb called me and offered to do anything he could to help. Massive props to OpenAI for handling a shitty situation incredibly well.

Opens with an observation

Matt Shumer@mattshumer_

Yeah, AI progress is totally, definitely stalling... Look at MathArena Apex. GPT-5.1 scored 1%. Gemini 3 scored 23%. That is a >20x jump on one of the hardest reasoning tasks we have. But sure, keep your head in the sand...

Opens with an observation

Matt Shumer@mattshumer_

Nvidia is worth $5T because the world realized inference is the new oil. If Etched’s chips are meaningfully better than Nvidia’s for inference… What is Etched worth? This is a worldview-shattering event.

Opens with a number

Matt Shumer@mattshumer_

Wow. This is huge. The first time (I'm aware of) that an OpenAI exec has publicly stated that they believe OpenAI is clearly prioritizing capabilities over safety research. Massive implications, in many ways.

Opens with an observation

Matt Shumer@mattshumer_

Fable 5.1 is a massive upgrade, but the bigger story is how much cheaper it is. Cache reads in particular are 75% cheaper! This puts it much closer to the pricing of GPT-5.6-Sol, while being a much more capable model.

Opens with an observation

Matt Shumer@mattshumer_

Some initial impressions of o3 mini: - it’s clear that the benchmarks don’t fully capture how good this model is — it’s clearly the best model I’ve used, for code - the Cursor team has not figured out how to get it to work well in Composer — ChatGPT gives far better results

Opens with an observation

Matt Shumer@mattshumer_

LLaMA 3's will start to drop next week. Assuming there's a 7B version, I'm expecting it to far surpass the current Mistral model.

Opens with an observation

Matt Shumer@mattshumer_

There's a new king of open-source. The new LLM from @databricks just beat Mixtral! - 132B total params (16 experts), 36B active (4 experts) - Trained on *12T* tokens - 32K context length From my initial tests, I'm really impressed.

Opens with an observation

Matt Shumer@mattshumer_

I've had access to GPT-5.6 since May 27. Everyone is going to tell you it's incredible. They're right. But this isn't another post glazing 5.6. For 2 weeks, it was the best model I'd ever used. Then Fable came out, and I stopped using GPT-5.6 overnight. Here's my review:

Opens with an observation

Matt Shumer@mattshumer_

I'm not usually a fan of OSS models, but this new Kimi release looks pretty great (on the surface) and is priced extremely well. I plan to try it later today... how has it been so far for you?

Opens with an announcement

Matt Shumer@mattshumer_

This is so damn cool! And by the time it ships, models will be insanely good at setting up RL runs to train this autonomously. @huggingface @pollenrobotics would love to try it early and share my thoughts/experiments with the world!

Opens with an observation

Matt Shumer@mattshumer_

Claude Opus 4.5 looks really great (by the numbers, at least) I don't get early access to @AnthropicAI models unfortunately, so I don't have a review to share today, but I'll absolutely be testing it and sharing my findings in the coming days!

Opens with an observation

Matt Shumer@mattshumer_

This is yet another example of Claude Mythos’s incredible hacking capabilities. I expect we’ll see more examples and independent evaluations in the coming weeks that make clear just how powerful (and dangerous, in the wrong hands) this model could be.

Opens with an observation