MatthewBerman @MatthewBerman Twitter profile

Last Seen Profiles

@AnaPorras5

@RidaKha53796042

@urownerdior

@Ladeegem01

@alitabet

@Certified_G_D_O

@mximm__

@JapanStrayCat

@karthik_r_n

@MavundlaLungeka

@ElizabethW5547

@walterpauli

@JoslynArmstrong

@hiighlien

@Safia_Alsnaani

@Aphiren

@BinorRaja

@stw_pdg

@MDDailyRecord

@MarcAnthony0742

@dtapp55

@LIVELOVEMYY

@Mohamed7h7

@JesusYo39225402

@DSwoffer

@JannatulMa25123

@sk6vice

@Jus_CT_

@sayakaiurani

@sanyukkuri115

@stw_pdg

@mohom44096487

@pencitaemanedut

@SmurphDFS

@funshodarq1

@undergroundchi

MatthewBerman

@MatthewBerman

15 days

Holy sh*t...

322

563

9K

MatthewBerman

@MatthewBerman

2 months

#1 trending github repo right now looks INSANE Single image to live stream deep fake. Look at that quality!! It's called Deep-Live-Cam (link in replies)

161

790

7K

MatthewBerman

@MatthewBerman

10 months

OpenAI just dropped their Prompt Engineering guide. Here are 6 strategies they recommend for getting better results from LLMs:

71

629

6K

MatthewBerman

@MatthewBerman

9 months

The New York Times just sued OpenAI and Microsoft for copyright infringement for using its content to train ChatGPT. Some of the allegations are STUNNING! I read all 69 pages of the lawsuit, here are the most 🔥 parts:

164

894

4K

MatthewBerman

@MatthewBerman

9 months

The End of Programming is near... Here is exactly why I believe this. Agree? Disagree? Let me know?

290

271

2K

MatthewBerman

@MatthewBerman

2 months

GPT4o Mini is the most important model @OpenAI has ever released. It's lightning-fast, dirt cheap, and actually beat GPT4o in my tests. Sam Altman says "...intelligence too cheap to meter." What a time to be alive. Check out the full test below.

23

38

309

MatthewBerman

@MatthewBerman

6 months

Seriously considered buying a new laptop today because my python environment is completely broken. I tried uninstalling everything python-related and...couldn't. I've been programming for 25 years. This is so sad.

451

39

1K

MatthewBerman

@MatthewBerman

10 months

I let ChatGPT take FULL CONTROL of my computer. Super scary! It used the mouse and keyboard just like a human user. Here's a preview of what happened. Full video coming soon!

78

215

1K

MatthewBerman

@MatthewBerman

9 months

Google Deep Mind researchers just dropped a HUGE advancement in robotics 🤖 It is truly mind-blowing. A "robot in every household" is closer than we think. Here's everything you need to know about the Mobile Aloha robot (& more demo videos): 👇

64

294

1K

MatthewBerman

@MatthewBerman

1 month

Sakana AI, the automated research scientist, is an open-source project! Should I make a tutorial for it?

GitHub - SakanaAI/AI-Scientist: The AI Scientist: Towards Fully Automated Open-Ended Scientific...

The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery 🧑‍🔬 - SakanaAI/AI-Scientist

github.com

114

137

1K

MatthewBerman

@MatthewBerman

10 months

Ever wonder what an LLM actually looks like under the hood? This visualization is mesmerizing. Check it out yourself:

12

204

1K

MatthewBerman

@MatthewBerman

12 days

99% of use cases don’t need PhD-level intelligence.

222

51

1K

MatthewBerman

@MatthewBerman

2 months

A battle for AI is unfolding before our eyes. AI compute is being pushed toward edge devices. Small models are getting better every single day. Just this week, multiple exceptional small models were released. On the other end of the spectrum, OpenAI is pushing AI cloud compute

30

35

236

MatthewBerman

@MatthewBerman

4 months

. @nvidia just released their own open-source model! Nemotron-4 340B, a family of open models that developers can use to generate synthetic data for training large language models (LLMs) for commercial applications across healthcare, finance, manufacturing, retail and every

34

172

1K

MatthewBerman

@MatthewBerman

2 months

Possibly the easiest jailbreak ever was just discovered! "In the past..." Models will never be perfectly aligned.

87

69

1K

MatthewBerman

@MatthewBerman

3 months

. @DavidSacks you got the facts wrong about Apple Intelligence on this week’s ep of All In. Apple did build their own AI, a 3b param model run locally that powers all the AI they demo’d. They also have a larger model in the cloud that they offload to when necessary. The larger

92

58

1K

MatthewBerman

@MatthewBerman

9 months

Open Interpreter is like ChatGPT Code Interpreter, but you can run it locally on your computer, and it is 100% open-source. Write and execute code, analyze data, control your computer, and more. Great work @hellokillian Here's how to install and use it: 👇

35

136

1K

MatthewBerman

@MatthewBerman

15 days

First impression of o1: It's not actually the quality of the answers that are impressive (which...they are), it's the thought process to arrive at the answers. o1 shows a TRUE understanding of what it's outputting.

43

49

1K

MatthewBerman

@MatthewBerman

11 months

Amazon fine-tuned a version of Mistral to allow for 32k tokens. Welcome to the open-source community, @amazon

amazon/MistralLite · Hugging Face

huggingface.co

15

136

997

MatthewBerman

@MatthewBerman

2 months

Tomorrow is a pivotal day in the world of AI, with the release of LLaMA 3.1 405b. Suddenly, the world will have access to an open-source model considered SOTA. It BEATS GPT4o on many benchmarks. What a time to be alive.

45

159

909

MatthewBerman

@MatthewBerman

7 months

Petition to ban the .webp image format.

109

64

884

MatthewBerman

@MatthewBerman

29 days

What happens when LLMs no longer need to write code legible to humans? Hyper-efficient code will look nothing like our current coding languages. Maybe they will just be symbols.

303

61

879

MatthewBerman

@MatthewBerman

15 days

Just finished testing o1-preview. It got JUST ONE answer wrong out of my entire test. Can you guess which one?

175

22

843

MatthewBerman

@MatthewBerman

29 days

AI DOOM: Fully AI Generated Video Games Are HERE! Google's AI DOOM project generates games with NO GAME ENGINE. This will be the future of video games. Here's a full breakdown:

52

155

783

MatthewBerman

@MatthewBerman

15 days

o1 (aka Strawberry) is here. Time to record.

93

35

750

MatthewBerman

@MatthewBerman

1 month

Finally started using Perplexity. I see what I’ve been missing out on now.

128

23

733

MatthewBerman

@MatthewBerman

9 months

CrewAI allows you to create incredible AI agent teams, similar to AutoGen. It is simple and intuitive, allowing you to accomplish tasks like research, writing, stock analysis, and trip planning. Plus, it can run entirely locally with open-source models. Here is a

30

114

682

MatthewBerman

@MatthewBerman

1 month

GPT-5, codenamed Project Orion, is ready to ship. 🚨 And apparently, they've already shown it to the FEDS. They used Strawberry/Q* to create synthetic data for GPT-5. Here's a full breakdown of the news:

59

120

728

MatthewBerman

@MatthewBerman

3 months

This is a worm that I FULLY SIMULATED on my computer. It's not real...but it kind of is... Simulation theory proof seems inevitable. Thanks OpenWorm.

89

60

701

MatthewBerman

@MatthewBerman

1 month

Eric Schmidt, ex-CEO of Google, reveals how Artificial Intelligence (AI) is going to change EVERYTHING. From agents to drones to energy, nothing will be the same. Here's a full breakdown of his interview at Stanford University.

48

142

692

MatthewBerman

@MatthewBerman

6 months

OpenDevin is a fantastic (early) open-source project looking to clone Devin's functionality. Build entire apps with a single prompt. Tutorial Below 🎥👇

22

124

674

MatthewBerman

@MatthewBerman

2 months

GitHub - hacksider/Deep-Live-Cam: real time face swap and one-click video deepfake with only a...

real time face swap and one-click video deepfake with only a single image - hacksider/Deep-Live-Cam

github.com

4

73

671

MatthewBerman

@MatthewBerman

10 months

Incredible way to visualize what's actually happening within an LLM:

8

71

652

MatthewBerman

@MatthewBerman

9 months

Introducing AutoGen Studio! 🚀 From Microsoft Research, AutoGen Studio is a NO CODE platform to build AI Agent teams. They do everything from writing and executing code 🧑‍💻 to planning travel ✈️ to plotting stock charts. 📈 Here’s how to install and use AutoGen Studio: 👇

30

111

620

MatthewBerman

@MatthewBerman

2 months

RouteLLM is one of the most impactful algorithmic innovations in AI that I've ever seen. I don't think people realize how important it truly will become. Here's a full tutorial for how to use it:

20

93

623

MatthewBerman

@MatthewBerman

2 months

What happens when you mix the quality of MoA with the speed of Groq? You get the fastest and highest quality LLM output. This is the architecture of the future, as intelligence costs get pushed to zero. Here's a tutorial:

23

81

614

MatthewBerman

@MatthewBerman

9 months

Currently holding Dolphin 2.6 Mixtral GGUF quant and LMStudio. Going straight into my safe. I now have the entirety of human knowledge on this SSD. Going to explore ways to back it up even more securely.

68

42

595

MatthewBerman

@MatthewBerman

4 months

Elon got burned hard on his typical “what did you do” question that he always gives. Still couldn’t admit defeat though.

Elon Musk

@elonmusk

4 months

@ylecun @bneiluj @how_many_roads_ What “science” have you done in the past 5 years?

435

215

4K

90

18

587

MatthewBerman

@MatthewBerman

2 months

Claude 3.5 "Artifacts" system prompt has leaked! So how does it work? 👇

16

66

586

MatthewBerman

@MatthewBerman

14 days

Why are people giving o1 so much crap for using CoT? If you can use the same technique and score higher on benchmarks, please do!

81

19

573

MatthewBerman

@MatthewBerman

21 days

What is Reflection 70b? This new model uses a novel technique to teach LLMs to SELF CORRECT, greatly reducing hallucinations. Here's everything you need to know:

23

69

571

MatthewBerman

@MatthewBerman

15 days

The second question blows my mind. It didn't count spaces or non-letter characters. 🍓

40

28

568

MatthewBerman

@MatthewBerman

29 days

The end of apps is near. Programming is changing right before our eyes. Let's break down what Linus Torvalds, the creator of Linux, thinks about AI and programming.

20

90

554

MatthewBerman

@MatthewBerman

4 months

Mixture of Agents is LEGIT. The concept makes sense, and after some initial testing of the provided code (thank you @togethercompute ), the performance is incredible. Here's my review of their paper.

28

65

549

MatthewBerman

@MatthewBerman

5 months

Hey look what's at the top of GitHub trending, I wonder why 😉.

14

39

535

MatthewBerman

@MatthewBerman

5 months

Imagine being OpenAI and your competition is giving away a $100m product for free.

Emad

@EMostaque

5 months

Meta Llama 3 70b was a $15-20m training run roughly. Llama 3 405b will be like $100m.

11

20

320

31

25

526

MatthewBerman

@MatthewBerman

1 month

Google is in trouble. There's no sugar-coating it: AI Search is a MAJOR threat to Google's cash cow. I've been testing SearchGPT and Perplexity against Google Search and the results are clear. Here's my full test video:

71

77

520

MatthewBerman

@MatthewBerman

9 months

The most remarkable AI releases of 2023

24

105

509

MatthewBerman

@MatthewBerman

30 days

Just in time code. Apps are going away.

Robert Bye

@RobertJBye

30 days

RIP App Stores 2008 - 2024

455

2K

24K

37

57

520

MatthewBerman

@MatthewBerman

10 months

Holy.... Mistral-medium: Me: How many words in your response to this prompt? Mistral: "Nine words are in my response to this prompt."

40

23

496

MatthewBerman

@MatthewBerman

1 month

Is this the beginning of AGI? An AI Scientist that can discover NEW SCIENCE and then uses that to self-improve! Sakana AI dropped a paper last week that went a bit under the radar, but seems like it's the step right before the "intelligence explosion." Full breakdown:

30

88

499

MatthewBerman

@MatthewBerman

10 months

I just setup and tested the "self-operating computer" and WOW was it scary. AI literally took over my mouse and keyboard and did what I told it to do. Video coming soon.

GitHub - OthersideAI/self-operating-computer: A framework to enable multimodal models to operate a...

A framework to enable multimodal models to operate a computer. - OthersideAI/self-operating-computer

github.com

19

55

490

MatthewBerman

@MatthewBerman

8 months

Want 100% local coding assistant? With @SourcegraphCody + @ollama + @code you can have an auto-completing coding assistant like Gitpilot Copilot, and it works without an internet connection! Not only that, it's BLAZING fast and FREE. Here's how: 👇

19

79

497

MatthewBerman

@MatthewBerman

2 months

I need more AI and less politics in my timeline right now lol

81

24

489

MatthewBerman

@MatthewBerman

2 months

Picture this: GPT4o quality, but 100x speed. Llama 3.1 405b + @GroqInc is powerful.

33

28

480

MatthewBerman

@MatthewBerman

15 days

I can't believe @OpenAI used my marble question in their demo video for o1-preview!! I'm flattered ❤️🍓

26

13

477

MatthewBerman

@MatthewBerman

29 days

Me when I know the feds have seen GPT-5 and I haven't even tried GPT4o voice.

Sam Altman

@sama

29 days

we are happy to have reached an agreement with the US AI Safety Institute for pre-release testing of our future models. for many reasons, we think it's important that this happens at the national level. US needs to continue to lead!

2K

1K

13K

29

23

469

MatthewBerman

@MatthewBerman

1 month

Wooo 300k 1.5 years, 388 videos, and so much learned. Thanks to everyone who watches my vids. And thanks to Melissa and Alex for helping me grow this into an actual business. Lots of fun things coming soon 🫡

58

6

465

MatthewBerman

@MatthewBerman

21 days

Reflection 70b just dropped and is beating every other model, including GPT4o and Claude 3.5. How did this happen? Here's my conversation with Matt Shumer ( @mattshumer_ ) and Sahil Chaudhary ( @csahil28 ), the authors of Reflection 70b.

33

73

456

MatthewBerman

@MatthewBerman

8 months

I built an AI Agent team that takes a YouTube video and turns it into a blog post and X thread. This is a REAL WORLD use case for agents. In this tutorial, I'll show you how to do it with skills (function calling) and a multi-agent team. 👇

17

56

439

MatthewBerman

@MatthewBerman

1 month

. @AravSrinivas is Perplexity suppose to become a verb? Drop the “Per…ity”. Just Plex. It’s cleaner.

90

21

443

MatthewBerman

@MatthewBerman

10 months

Find this and more from their guide directly:

OpenAI Platform

Explore developer resources, tutorials, API docs, and dynamic examples to get the most out of OpenAI's platform.

platform.openai.com

4

44

422

MatthewBerman

@MatthewBerman

5 months

Want to build ENTIRE front ends without knowing how to code? Describe what you want your front end to look like, and OpenUI will build it in real time! Super impressive. Here's the tutorial:

15

69

425

MatthewBerman

@MatthewBerman

3 months

Claude 3.5 Sonnet by @AnthropicAI is now the best model I've ever tested. Congrats to the team. Started up my subscription to Claude.

34

26

417

MatthewBerman

@MatthewBerman

1 month

Here we gooooooo!

40

18

416

MatthewBerman

@MatthewBerman

5 months

I’m ready to build my personal AI assistant. Here’s the stack I’m considering: - @GroqInc with llama 3 - iPhone shortcut for quick access - @pinecone for RAG Where should I host it? What voice model should I use? Speed and personalization are my priorities

83

23

415

MatthewBerman

@MatthewBerman

2 months

This was a BIG week for AI news, let's review: 0:00 - LLaMA 3.1 1:31 - xAI Supercomputer 2:28 - SearchGPT 4:50 - Mistral Large 2 7:10 - Meta AI in the EU 8:43 - Stable Audio Open 10:40 - GPT4o Advanced Voice Soon? 11:15 - OpenAI Making Custom Chip 12:24 - US Lawmakers Ask OpenAI

14

67

413

MatthewBerman

@MatthewBerman

22 days

Should I test it?

Matt Shumer

@mattshumer_

22 days

I'm excited to announce Reflection 70B, the world’s top open-source model. Trained using Reflection-Tuning, a technique developed to enable LLMs to fix their own mistakes. 405B coming next week - we expect it to be the best model in the world. Built w/ @GlaiveAI . Read on ⬇️:

553

1K

10K

93

7

412

MatthewBerman

@MatthewBerman

2 months

@garrytan Yes. But it's still very limited. As soon as you reach a certain level of complexity in your code, LLMs struggle to make sense of it all.

42

3

406

MatthewBerman

@MatthewBerman

1 month

Switched from Google Search to SearchGPT and didn't even blink an eye. Uh oh...

58

14

404

MatthewBerman

@MatthewBerman

2 months

Google beat OpenAI to full voice functionality.

Google

@Google

2 months

We’re introducing Gemini Live, a more natural way to interact with Gemini. You can now have a free-flowing conversation, and even interrupt or change topics just like you might on a regular phone call. Available to Gemini Advanced subscribers. #MadeByGoogle

213

451

2K

46

28

402

MatthewBerman

@MatthewBerman

23 days

AI is moving at lightning speed. From GPT5 timeline updates to Meta AI's mixed reality AI glasses to AI paying AI with crypto, here's all the AI news you missed! Chapters: 0:00 - GPT-5 Coming Soon: OpenAI's Next Frontier 2:09 - xAI's Colossus: The Largest AI Training Cluster

22

63

398

MatthewBerman

@MatthewBerman

14 days

So is GPT-o1 just GPT4/4.5 trained/fine-tuned with CoT? Or is it a new model altogether? Or is it a framework (agentic?) built around GPT4/4.5?

116

7

394

MatthewBerman

@MatthewBerman

8 months

I've had the Apple Vision Pro for 24 hours. Here's everything I've noticed so far: 👇 (PS I'm typing this on a massive 80" screen extending from my MacBook Pro within AVP)

34

21

377

MatthewBerman

@MatthewBerman

6 months

Devin clones are coming quickly. I finally got Devika working and was able to reproduce the "game of life" demo easily. Here's a full tutorial:

14

53

371

MatthewBerman

@MatthewBerman

16 days

Huge AI News 🚨 OpenAI's Strawberry Model is expected to be the first conversational AI that "thinks" before responding. Plus, Nvidia runs into trouble, Apple Intelligence updates, $2,000/MO ChatGPT, and Replit AI Agents. Check out my hand-picked AI stories below:

16

25

374

MatthewBerman

@MatthewBerman

6 months

DBRX by @databricks ...it's REALLY good!! The New MoE 132b parameter model is open-source and costs $10 m to train. Thank you, Databricks, for your contribution to OS. Check out the full explanation and testing: 🎥👇

9

44

368

MatthewBerman

@MatthewBerman

9 months

NYT is alleging OpenAI/Microsoft unlawfully used MILLIONS of their articles. Not only that, they gave NYT content extra "weight" in the models because that content is so valuable.

10

32

359

MatthewBerman

@MatthewBerman

10 months

Just finished testing Mixtral 7bx8... It is BY FAR the best open-source model I've ever tested...

29

21

318

MatthewBerman

@MatthewBerman

2 months

Let's de-normalize companies demo'ing products earlier than 3 months before release. * Microsoft: Recall * Apple: Intelligence * OpenAI: Sora AND GPT4o Voice Who else should be called out?

64

32

351

MatthewBerman

@MatthewBerman

11 months

Proud of this one :)

76

3

352

MatthewBerman

@MatthewBerman

10 months

This feels like the future of AI Agents. Each agent has its own model, fine-tuned for that agent's specific role, running entirely locally, thanks to @Ollama_ai @pyautogen @LiteLLM This is not science fiction. This is NOW. Here's a full tutorial:

Power Each AI Agent With A Different LOCAL LLM (AutoGen + Ollama...

In this video, I show you how to power AutoGen AI agents using individual open-source models per AI agent, this is going to be the future AI tech stack for r...

www.youtube.com

11

52

347

MatthewBerman

@MatthewBerman

3 months

. @AnthropicAI 's Claude 3.5 is the best model I've ever used. Looks like I need new, much more challenging, test questions for it. Here's my full test:

23

40

336

MatthewBerman

@MatthewBerman

15 days

🍓 Strawberry/Q* is finally here! @OpenAI just dropped what can truly be considered AGI, and beginning of the intelligence explosion. 💥 It's called GPT-o1 Here's a full breakdown:

29

36

337

MatthewBerman

@MatthewBerman

5 months

OpenAI Wants to TRACK GPUs?! They Went Too Far With This… Breakdown below 🎥👇 Here's the quote: "Extending cryptographic protection to the hardware layer has the potential to achieve the following properties...GPUs can be cryptographically attested for authenticity and

72

65

334

MatthewBerman

@MatthewBerman

5 months

The top comment on the hacker news post about the new Apple M4 chip is spot on. Long Apple.

43

36

322

MatthewBerman

@MatthewBerman

7 months

I’ve been powering AI agents with @GroqInc . Now, the slowest thing in my AI stack is embeddings. The agents are instant, it’s incredible.

22

19

322

MatthewBerman

@MatthewBerman

12 days

The anxiety of trying to choose the right model for my task as to not waste my o1 quota.

50

6

324

MatthewBerman

@MatthewBerman

5 months

MASSIVE Step Allowing AI Agents To Control Computers (MacOS, Windows, Linux) OS World allows agents to fully control computers, including MacOS, Windows, and Linux. OS World can benchmark agent performance like never before by giving agents a language to describe actions in a

24

50

318

MatthewBerman

@MatthewBerman

29 days

The last 3 days have been absolutely wild. Here’s the AI news you can’t miss:

10

26

318

MatthewBerman

@MatthewBerman

9 days

I tested Pixtral, the new vision model just released by @MistralAI It's surprisingly good! The best OSS vision model I've tested so far. Check out the full test:

20

46

366

MatthewBerman

@MatthewBerman

23 days

Imagine this: You have a conversation with AI that builds you an app in realtime. No saves. No reloads. No coding. Voice to app. Coming soon :)

40

18

316

MatthewBerman

@MatthewBerman

4 months

Why is he standing on the roof btw?

MatthewBerman

@MatthewBerman

4 months

"New intelligence capabilities"

12

6

37

115

6

312

MatthewBerman

@MatthewBerman

9 months

The suit says Microsoft and ChatGPT can generate VERBATIM excerpts from NYT articles, which leads to users not needing to go to and pay them. NYT also states excerpts and summaries are vastly different than search engine result previews.

4

15

304

MatthewBerman

@MatthewBerman

9 months

I’m going to buy a few USB drives, download the top OS models, and store them in my safe. Am I being paranoid?

88

18

292

MatthewBerman

@MatthewBerman

4 months

It doesn’t sound like ScarJo to me

Austen Allred

@Austen

4 months

Here’s the OpenAI demo video of “Sky.” I have no dog in this fight, but: 1. I can see how Scarlett Johansson would hear that and say, “Hey wtf that’s me!” 2. It’s clearly not actually Scarlett Johansson. Very different voice, and not trying to be her.

137

35

684

108

13

299

MatthewBerman

@MatthewBerman

2 months

Is sus-column-r REALLY strawberry Q*? I put it to the test to find out. Check out my full suite of tests against this new logic and reasoning king model: