@matthew-berman.bsky.social
Is Chain-of-Thought (CoT) reasoning in LLMs just...for show?
@AnthropicAI’s new research paper shows that not only do AI models not use CoT like we thought, they might not use it at all for reasoning.
In fact, they might be lying to us in their CoT.
What you need to know: 🧵
@AnthropicAI’s new research paper shows that not only do AI models not use CoT like we thought, they might not use it at all for reasoning.
In fact, they might be lying to us in their CoT.
What you need to know: 🧵
April 8, 2025 at 6:34 PM
Is Chain-of-Thought (CoT) reasoning in LLMs just...for show?
@AnthropicAI’s new research paper shows that not only do AI models not use CoT like we thought, they might not use it at all for reasoning.
In fact, they might be lying to us in their CoT.
What you need to know: 🧵
@AnthropicAI’s new research paper shows that not only do AI models not use CoT like we thought, they might not use it at all for reasoning.
In fact, they might be lying to us in their CoT.
What you need to know: 🧵
AI has changed my life.
I'm now 100x more productive than I ever was.
How do I use it? Which tools do I use?
Here are my actual use cases for AI: 👇
I'm now 100x more productive than I ever was.
How do I use it? Which tools do I use?
Here are my actual use cases for AI: 👇
March 7, 2025 at 6:14 PM
AI has changed my life.
I'm now 100x more productive than I ever was.
How do I use it? Which tools do I use?
Here are my actual use cases for AI: 👇
I'm now 100x more productive than I ever was.
How do I use it? Which tools do I use?
Here are my actual use cases for AI: 👇
Alibaba just dropped QWQ 32B, an open-source model rivaling DeepSeek R1.
It’s much smaller (32B vs 671B params) but delivers comparable results. You can run it on your PC!
Insanely fast, thinking-focused, and agent-capable.
Let’s dive in.
It’s much smaller (32B vs 671B params) but delivers comparable results. You can run it on your PC!
Insanely fast, thinking-focused, and agent-capable.
Let’s dive in.
March 7, 2025 at 5:48 PM
Alibaba just dropped QWQ 32B, an open-source model rivaling DeepSeek R1.
It’s much smaller (32B vs 671B params) but delivers comparable results. You can run it on your PC!
Insanely fast, thinking-focused, and agent-capable.
Let’s dive in.
It’s much smaller (32B vs 671B params) but delivers comparable results. You can run it on your PC!
Insanely fast, thinking-focused, and agent-capable.
Let’s dive in.
Major AI breakthrough: Diffusion Large Language Models are here!
They're 10x faster and 10x cheaper than traditional LLMs.
Here's everything you need to know:
They're 10x faster and 10x cheaper than traditional LLMs.
Here's everything you need to know:
March 7, 2025 at 12:48 AM
Major AI breakthrough: Diffusion Large Language Models are here!
They're 10x faster and 10x cheaper than traditional LLMs.
Here's everything you need to know:
They're 10x faster and 10x cheaper than traditional LLMs.
Here's everything you need to know:
OpenAI just dropped GPT-4.5!
This is their "largest and best model for chat yet"
Here's what you need to know...
This is their "largest and best model for chat yet"
Here's what you need to know...
February 27, 2025 at 9:39 PM
OpenAI just dropped GPT-4.5!
This is their "largest and best model for chat yet"
Here's what you need to know...
This is their "largest and best model for chat yet"
Here's what you need to know...
I'll take two plz
February 20, 2025 at 9:22 PM
I'll take two plz
Can AI earn $1M as a freelance software engineer? 💰💰
OpenAI's SWE-Lancer benchmark seems like the most "real world" coding benchmark to date.
This might be the biggest arbitrage opportunity on the planet.
A thread on what they found 👇🧵
OpenAI's SWE-Lancer benchmark seems like the most "real world" coding benchmark to date.
This might be the biggest arbitrage opportunity on the planet.
A thread on what they found 👇🧵
February 18, 2025 at 8:36 PM
Can AI earn $1M as a freelance software engineer? 💰💰
OpenAI's SWE-Lancer benchmark seems like the most "real world" coding benchmark to date.
This might be the biggest arbitrage opportunity on the planet.
A thread on what they found 👇🧵
OpenAI's SWE-Lancer benchmark seems like the most "real world" coding benchmark to date.
This might be the biggest arbitrage opportunity on the planet.
A thread on what they found 👇🧵
OpenAI just dropped a paper that reveals the blueprint for creating the best AI coder in the world.
But here’s the kicker: this strategy isn’t just for coding—it’s the clearest path to AGI and beyond.
Let’s break it down 🧵👇
But here’s the kicker: this strategy isn’t just for coding—it’s the clearest path to AGI and beyond.
Let’s break it down 🧵👇
February 16, 2025 at 4:51 PM
OpenAI just dropped a paper that reveals the blueprint for creating the best AI coder in the world.
But here’s the kicker: this strategy isn’t just for coding—it’s the clearest path to AGI and beyond.
Let’s break it down 🧵👇
But here’s the kicker: this strategy isn’t just for coding—it’s the clearest path to AGI and beyond.
Let’s break it down 🧵👇
New research paper shows how LLMs can "think" internally before outputting a single token!
Unlike Chain of Thought, this "latent reasoning" happens in the model's hidden space.
TONS of benefits from this approach.
Let me break down this fascinating paper...
Unlike Chain of Thought, this "latent reasoning" happens in the model's hidden space.
TONS of benefits from this approach.
Let me break down this fascinating paper...
February 13, 2025 at 4:51 PM
New research paper shows how LLMs can "think" internally before outputting a single token!
Unlike Chain of Thought, this "latent reasoning" happens in the model's hidden space.
TONS of benefits from this approach.
Let me break down this fascinating paper...
Unlike Chain of Thought, this "latent reasoning" happens in the model's hidden space.
TONS of benefits from this approach.
Let me break down this fascinating paper...
DeepSeek R1 just got a 2X speed boost!
The crazy part?
The code for the boost was WRITTEN BY R1 itself!
Self-improving AI is here.
The crazy part?
The code for the boost was WRITTEN BY R1 itself!
Self-improving AI is here.
January 28, 2025 at 6:17 PM
DeepSeek R1 just got a 2X speed boost!
The crazy part?
The code for the boost was WRITTEN BY R1 itself!
Self-improving AI is here.
The crazy part?
The code for the boost was WRITTEN BY R1 itself!
Self-improving AI is here.
DeepSeek R1 just dropped, and it has sent shockwaves through the AI industry.
Imagine an AI model that competes with OpenAI’s best—trained for just $5M, open-source, and free.
Here's the wild story so far 🧵👇
Imagine an AI model that competes with OpenAI’s best—trained for just $5M, open-source, and free.
Here's the wild story so far 🧵👇
January 27, 2025 at 3:54 PM
DeepSeek R1 just dropped, and it has sent shockwaves through the AI industry.
Imagine an AI model that competes with OpenAI’s best—trained for just $5M, open-source, and free.
Here's the wild story so far 🧵👇
Imagine an AI model that competes with OpenAI’s best—trained for just $5M, open-source, and free.
Here's the wild story so far 🧵👇
OpenAI just dropped Operator, their first Agents, who can use web browsers to complete tasks for you.
For the first time, OpenAI's agents can directly impact the real world.
The AI industry had strong reactions!
Here’s a roundup of reactions and incredible use cases. 🧵👇
For the first time, OpenAI's agents can directly impact the real world.
The AI industry had strong reactions!
Here’s a roundup of reactions and incredible use cases. 🧵👇
January 25, 2025 at 4:24 PM
OpenAI just dropped Operator, their first Agents, who can use web browsers to complete tasks for you.
For the first time, OpenAI's agents can directly impact the real world.
The AI industry had strong reactions!
Here’s a roundup of reactions and incredible use cases. 🧵👇
For the first time, OpenAI's agents can directly impact the real world.
The AI industry had strong reactions!
Here’s a roundup of reactions and incredible use cases. 🧵👇
What do you want me to test with OpenAI Operator?
January 23, 2025 at 7:30 PM
What do you want me to test with OpenAI Operator?
1/ SakanaAI just dropped their latest research: Transformer²
It's a self-adaptive architecture that allows AI to evolve at inference time.
Model weights are no longer "static"
Let’s break it down: 🧵
It's a self-adaptive architecture that allows AI to evolve at inference time.
Model weights are no longer "static"
Let’s break it down: 🧵
January 16, 2025 at 4:39 PM
1/ SakanaAI just dropped their latest research: Transformer²
It's a self-adaptive architecture that allows AI to evolve at inference time.
Model weights are no longer "static"
Let’s break it down: 🧵
It's a self-adaptive architecture that allows AI to evolve at inference time.
Model weights are no longer "static"
Let’s break it down: 🧵
My first CES was...AWESOME
Thank you to @NVIDIA_AI_PC for inviting me. What an incredible team!
Check out our wrap up vid on IG (link below)
#NVIDIAPartner
Thank you to @NVIDIA_AI_PC for inviting me. What an incredible team!
Check out our wrap up vid on IG (link below)
#NVIDIAPartner
January 14, 2025 at 9:48 PM
My first CES was...AWESOME
Thank you to @NVIDIA_AI_PC for inviting me. What an incredible team!
Check out our wrap up vid on IG (link below)
#NVIDIAPartner
Thank you to @NVIDIA_AI_PC for inviting me. What an incredible team!
Check out our wrap up vid on IG (link below)
#NVIDIAPartner
1/9 BREAKING
Biden Admin drops major AI chip rules today!
The 200+ page "AI Diffusion" framework completely reshapes global AI tech trade.
Key goal: Keep advanced AI development running on "American rails"
But not everyone is happy... 🧵
Biden Admin drops major AI chip rules today!
The 200+ page "AI Diffusion" framework completely reshapes global AI tech trade.
Key goal: Keep advanced AI development running on "American rails"
But not everyone is happy... 🧵
January 13, 2025 at 8:47 PM
1/9 BREAKING
Biden Admin drops major AI chip rules today!
The 200+ page "AI Diffusion" framework completely reshapes global AI tech trade.
Key goal: Keep advanced AI development running on "American rails"
But not everyone is happy... 🧵
Biden Admin drops major AI chip rules today!
The 200+ page "AI Diffusion" framework completely reshapes global AI tech trade.
Key goal: Keep advanced AI development running on "American rails"
But not everyone is happy... 🧵
What will society look like after AGI is achieved?
I found a great prediction on LessWrong by L Rudolf L (link below).
Capital will matter MORE after AGI.
A thread on the future of wealth, power & human agency 🧵
I found a great prediction on LessWrong by L Rudolf L (link below).
Capital will matter MORE after AGI.
A thread on the future of wealth, power & human agency 🧵
January 8, 2025 at 7:07 PM
What will society look like after AGI is achieved?
I found a great prediction on LessWrong by L Rudolf L (link below).
Capital will matter MORE after AGI.
A thread on the future of wealth, power & human agency 🧵
I found a great prediction on LessWrong by L Rudolf L (link below).
Capital will matter MORE after AGI.
A thread on the future of wealth, power & human agency 🧵
Sam Altman dropped a cryptic tweet and a blog post letting the world know OpenAI is now working on ASI!
Singularity, superintelligence, slow vs. fast takeoff...
Here's what you need to know 🧵
Singularity, superintelligence, slow vs. fast takeoff...
Here's what you need to know 🧵
January 7, 2025 at 4:09 PM
Sam Altman dropped a cryptic tweet and a blog post letting the world know OpenAI is now working on ASI!
Singularity, superintelligence, slow vs. fast takeoff...
Here's what you need to know 🧵
Singularity, superintelligence, slow vs. fast takeoff...
Here's what you need to know 🧵
The future of AI is MULTI-MODEL 𐂷
NotDiamond CEO @tomas_hk and former OpenAI exec @iamthezack just dropped an article showing a clear vision of AI for 2025.
For most use cases, having multiple smaller, more specialized models will be the right solution.
Here’s everything you need to know: 🧵
NotDiamond CEO @tomas_hk and former OpenAI exec @iamthezack just dropped an article showing a clear vision of AI for 2025.
For most use cases, having multiple smaller, more specialized models will be the right solution.
Here’s everything you need to know: 🧵
December 31, 2024 at 5:59 PM
The future of AI is MULTI-MODEL 𐂷
NotDiamond CEO @tomas_hk and former OpenAI exec @iamthezack just dropped an article showing a clear vision of AI for 2025.
For most use cases, having multiple smaller, more specialized models will be the right solution.
Here’s everything you need to know: 🧵
NotDiamond CEO @tomas_hk and former OpenAI exec @iamthezack just dropped an article showing a clear vision of AI for 2025.
For most use cases, having multiple smaller, more specialized models will be the right solution.
Here’s everything you need to know: 🧵
1/ 🚨 Big news: OpenAI makes it clear they are evolving into a for-profit
They are moving to a more closed and for-profit model while doubling down on AGI safety and scalability.
Is this the right balance of ethics and ambition, or is it a departure from their ideals?
Let’s unpack. 🧵
They are moving to a more closed and for-profit model while doubling down on AGI safety and scalability.
Is this the right balance of ethics and ambition, or is it a departure from their ideals?
Let’s unpack. 🧵
December 27, 2024 at 6:52 PM
1/ 🚨 Big news: OpenAI makes it clear they are evolving into a for-profit
They are moving to a more closed and for-profit model while doubling down on AGI safety and scalability.
Is this the right balance of ethics and ambition, or is it a departure from their ideals?
Let’s unpack. 🧵
They are moving to a more closed and for-profit model while doubling down on AGI safety and scalability.
Is this the right balance of ethics and ambition, or is it a departure from their ideals?
Let’s unpack. 🧵
o3 was announced less than a week ago and the AI industry was stunned.
I've collected some of the reactions from the biggest names in AI: 🧵
I've collected some of the reactions from the biggest names in AI: 🧵
December 24, 2024 at 7:54 PM
o3 was announced less than a week ago and the AI industry was stunned.
I've collected some of the reactions from the biggest names in AI: 🧵
I've collected some of the reactions from the biggest names in AI: 🧵
.@OpenAI just dropped o3 and o3-mini!
This is AGI (not clickbait)
o3 is the best AI ever created, and its performance is WILD.
Here's everything you need to know: 🧵
This is AGI (not clickbait)
o3 is the best AI ever created, and its performance is WILD.
Here's everything you need to know: 🧵
December 20, 2024 at 7:27 PM
.@OpenAI just dropped o3 and o3-mini!
This is AGI (not clickbait)
o3 is the best AI ever created, and its performance is WILD.
Here's everything you need to know: 🧵
This is AGI (not clickbait)
o3 is the best AI ever created, and its performance is WILD.
Here's everything you need to know: 🧵
#1 Trending Github Project: Genesis 🌟
A groundbreaking framework for creating, training, & deploying embodied agents in simulated environments!
And it's open-source!
Here's why you should care: 🧵
A groundbreaking framework for creating, training, & deploying embodied agents in simulated environments!
And it's open-source!
Here's why you should care: 🧵
December 20, 2024 at 4:17 PM
#1 Trending Github Project: Genesis 🌟
A groundbreaking framework for creating, training, & deploying embodied agents in simulated environments!
And it's open-source!
Here's why you should care: 🧵
A groundbreaking framework for creating, training, & deploying embodied agents in simulated environments!
And it's open-source!
Here's why you should care: 🧵
.@AnthropicAI just published a WILD new AI jailbreaking technique
Not only does it crack EVERY frontier model, but it's also super easy to do.
ThIS iZ aLL iT TakE$ 🔥
Here's everything you need to know: 🧵
Not only does it crack EVERY frontier model, but it's also super easy to do.
ThIS iZ aLL iT TakE$ 🔥
Here's everything you need to know: 🧵
December 20, 2024 at 2:54 PM
.@AnthropicAI just published a WILD new AI jailbreaking technique
Not only does it crack EVERY frontier model, but it's also super easy to do.
ThIS iZ aLL iT TakE$ 🔥
Here's everything you need to know: 🧵
Not only does it crack EVERY frontier model, but it's also super easy to do.
ThIS iZ aLL iT TakE$ 🔥
Here's everything you need to know: 🧵
Microsoft's Satya Nadella just dropped a bombshell:
"Apps as we know them are going away in favor of agents."
Is this the end of traditional software?
Here's what you need to know: 🧵
"Apps as we know them are going away in favor of agents."
Is this the end of traditional software?
Here's what you need to know: 🧵
December 19, 2024 at 10:31 PM
Microsoft's Satya Nadella just dropped a bombshell:
"Apps as we know them are going away in favor of agents."
Is this the end of traditional software?
Here's what you need to know: 🧵
"Apps as we know them are going away in favor of agents."
Is this the end of traditional software?
Here's what you need to know: 🧵