Matt Shumer
@mattshumer_
Followers
101K
Following
10K
Media
838
Statuses
10K
CEO @HyperWriteAI, @OthersideAI, creator of https://t.co/PSUlubx5bb (Github for prompts), investor in @GroqInc @Etched @Rork_App @OpenRouterAI + many more
MODEL ALPHA 👉
Joined November 2019
Today, we’re unveiling Personal Assistant - @HyperWriteAI's groundbreaking AI agent that can use a web browser like a human. One agent to rule them all. It’s time to reimagine the way we interact with the internet.
250
414
2K
If you use Superhuman for email, a great prompt to use: "What lists should I unsubscribe from? Go through my inbox and list, in priority (by marketing email frequency) order, the worst offenders, and give me the unsubscribe links for each."
9
0
26
These guys are fucking crazy. They’re going to absolutely dominate this space.
The infrastructure layer is being rebuilt for AI, agents, and a new kind of cloud. So we’re hosting our inaugural @Daytonaio COMPUTE Conference on the Warriors’ home court, bringing 1,500 builders together. March 9, 2026 Apply for tickets below
1
0
16
So proud of Koby and the team. They continue to help so many people while crushing it on the business side as well. They’ve just raised fresh funding, consider working with them if you have the opportunity!
Sunflower has now raised a total of $6.5M, with our most recent valuation being a $60M post 🥳 We stumbled upon magic, AI has the power to convince us to stop doing drugs & not kill ourselves We're going to create 1 trillion days sober, 1 day at a time 🌻
6
0
12
This looks like an amazing release. Visual fidelity is insane and physics look spot-on. Only downside is that the audio doesn't seem to be as polished as Sora 2.
Introducing Whisper Thunder aka Gen-4.5. Today, we are excited to share our new frontier model. Gen-4.5 was built by a team that fits onto two school buses and decided to take on the largest companies in the world. We are David and we’ve brought one hell of a slingshot.
3
1
51
GPT-5.1 prompt to save $$ on Black Friday: “It’s Black Friday! Can you see if any products we’ve discussed recently are available at a nice discount?”
23
4
91
I'll be sharing a more detailed blog post on this soon. If you want to be the first to see it, go here:
tally.so
Made with Tally, the simplest way to create forms.
1
0
13
My Model Stack After the Wildest 8 Days in AI: Opus 4.5: Most daily code tasks where I know how I want the model to do what I want. Fast, clean, reliable, but often starts writing code before it grabs all the context it needs (measures once, cuts twice). Codex‑Max: Larger, more
37
16
342
With model releases as frequent as they've been lately, we’re slicing the “wow” into tiny monthly increments, so each one feels underwhelming. Zoom out, though, and the step-change is obvious.
0
1
28
LLMs aren't improving... sure, yeah, ok buddy. Just go to https://t.co/iaCePEAp8a, select a set of models across major version jumps (e.g., GPT-3.5, GPT-4, and GPT-5), and give them all the same prompt. It's impossible to unsee.
40
29
481
Tech is commoditized even at the highest levels. Great for consumers!
I've now been testing Opus 4.5 against GPT-5.1-Codex-Max on backend tasks for the last 24 hours, and honestly, I can't decide on a clear winner. Usually when comparing models, the winner is pretty clear very quickly. Not this time. The testing continues.
5
1
23
I've now been testing Opus 4.5 against GPT-5.1-Codex-Max on backend tasks for the last 24 hours, and honestly, I can't decide on a clear winner. Usually when comparing models, the winner is pretty clear very quickly. Not this time. The testing continues.
69
20
635
4
1
5
AI Researcher now supports Claude Opus 4.5. Try it now at:
Introducing: AI Researcher 🧪 A Gemini 3-powered multi-agent AI system that autonomously runs ML experiments Just give it a research question, and it will: - Design experiments - Spin up specialist agents with their own GPUs to run them - Write a paper And it's open-source!
5
4
115
We need a new way to express AI costs… $/token doesn’t make much sense anymore. Maybe a benchmark that tries to give a sense of the cost to run an average workload?
This is notable: Opus 4.5 is ~60% more expensive than Sonnet ($25/million output compared to $15/million) but if it can use 76% fewer output reasoning tokens for the same complex task it may end up cheaper!
25
2
122
Watching certain AI skeptics reply to AI bots in their comments without realizing it is a beautiful kind of irony
12
1
100
Opus 4.5 is impressing me on backend tasks, but so far, it still feels inferior to 5.1 Pro (and ~on par w/, maybe slightly worse in context-gathering, than gpt-5.1-codex-max). Still very impressive and a huge jump for Anthropic! Good prompting might put it over the top though.
25
10
337
AI Researcher now supports Claude Opus 4.5 as the driver model! Open-source, try it:
github.com
Contribute to mshumer/autonomous-researcher development by creating an account on GitHub.
Introducing: AI Researcher 🧪 A Gemini 3-powered multi-agent AI system that autonomously runs ML experiments Just give it a research question, and it will: - Design experiments - Spin up specialist agents with their own GPUs to run them - Write a paper And it's open-source!
4
10
106
Welp, I guess it's time to add Claude Opus 4.5 support!
Introducing: AI Researcher 🧪 A Gemini 3-powered multi-agent AI system that autonomously runs ML experiments Just give it a research question, and it will: - Design experiments - Spin up specialist agents with their own GPUs to run them - Write a paper And it's open-source!
4
0
56
Okay wow, I'm kind of blown away. In one shot, Opus 4.5 made the UI actually functional, with Python running in the browser... Within the constraints of Claude Artifacts!
First test of Claude Opus 4.5, and I'm already impressed. I asked it for a Colab competitor UI, and it quickly pulled together this screen. Definitely better than my similar test with GPT-5.1 and (shockingly) Gemini 3! More testing to go, but this is a good start.
16
19
456