New Loopy (aka Loop Library skill) feature: 🔸Interview you to help craft a loop 🔸Run loops itself to help monitor them 🔸Review loop runs to debug issues Loop it up: github.com/Forward-Future/lo…

Never deleting this app

Anthropic just dropped an insane new paper. AI models can "fake alignment" - pretending to follow training rules during training but reverting to their original behaviors when deployed! Here's everything you need to know: 🧵

We knew very little about how LLMs actually work...until now. @AnthropicAI just dropped the most insane research paper, detailing some of the ways AI "thinks." And it's completely different than we thought. Here are their wild findings: 🧵

See...this is scary to me. When Elon doesn't agree with community notes, it's "gamed." But before, it was the best feature of X according to him. Now, he's putting his thumb on the scale and "fixing it." Seems very fishy.

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.

Holy sh*t...

Just read @AnthropicAI's paper revealing more about how AI models actually work. Haven't been this mind-blown by a paper in a long time.

#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)

“Only compare yourself to younger you, never to others” Words to live by

Anthropic just dropped an incredible guide on "How To Build Effective Agents" 2025 will be the year of AGENTS 🤖 Here's everything you need to know: 🧵

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 🧵👇

Jensen: "I've created more billionaires on my management team than any CEO in the world."

Google's Genie 3 is capable of generating full 3D worlds. But not only that, they are FULLY INTERACTIVE. Here are the most incredible demos they shared:

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

OMG o3 is insane. No one is safe lol.

DeepSeek R1 has been out for 24 hours. The AI industry's reaction has been...strong! Here's a collection of the most telling reactions: 🧵

h/t Wrong_User_Logged

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:

This is my Super Bowl

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...

Gemini 2.5 Pro is insane at coding. It's far better than anything else I've tested. Below are ONE SHOT demos 🧵

I don’t use Google search anymore. In a little under 6 months, my usage of Google search dropped 98% and has transitioned to about 80% Perplexity and 20% ChatGPT. This is a stunning habit change since I’ve been using Google for 20 years.

.@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: 🧵

New Paper: Stanford researcher (@joon_s_pk) discovers how to clone human personalities and inject them into AI Agents 🧠 This builds on last year's paper which put 1000's of fully automated agents in a simulated town. The results are wild. 👇

DeepSeek R1 has the most human-like internal monologue I've ever seen. It's actually quite endearing.

.@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: 🧵

Breaks your legs

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:

How machine learning sees the world

Thrilled to share I've joined OpenAI as one of the original founders.

1/ Google Research unveils new paper: "Titans: Learning to Memorize at Test Time" It introduces human-like memory structures to overcome the limits of Transformers, with one "SURPRISING" feature. Here's why this is huge for AI. 🧵👇

I can't fall asleep at night. I've tried everything, nothing helps. * Exercise daily * Sunlight first thing in the morning * First and last coffee 7:30am * No screens before bed * Last meal 4 hours before bed * Meditation * Sleep restriction * I don't drink alcohol * Only get in bed when ready to sleep * Weighted blanket * 8 Sleep set to cool * Room is cold But, without fail, my brain won't let me sleep. I get to the point where I'm about to fall asleep, then I wake up and that's it...I'm up. Any suggestions are very welcome. cc @bryan_johnson

NVIDIA adding native Python support to CUDA seems like a big deal. Python is the de facto coding language of AI.

What's a non-douchey way of telling people you don't drink when offered? (assuming "no, thank you" is followed up with questions)

Elon to everyone with a .gov email: "What did you get done this week?"

SaaS is dead. I can literally build any SaaS application in hours with AI. Is it as polished as a VC-backed company with 1,000's of employees? No. But it doesn't need to be. It's hyper-customized for exactly what I need. And it can evolve with my needs in real time. I'm going through all of the SaaS apps I pay for right now, stack ranking them by complexity, and building them starting from the bottom. Taste and data are all that matter now.

#1 Trending Github Repo: Screenshot-to-Code A simple tool to convert screenshots, mockups and Figma designs into clean, functional code using AI Check out the WILD demos below 👇

DeepSeek R1, self-hosted, on Taiwan's status as an independent country. Woooooow. This seems hard-coded in the model.

My new employee revealed he used AI to negotiate his salary with me. I revealed I used AI to negotiate back...

I would pay $1,000 per month for voice AI that could let me get actual work done when I drive. Answer emails, schedule meetings and do other tasks

.@Microsoft just dropped TinyTroupe! Described as "an experimental Python library that allows the simulation of people with specific personalities, interests, and goals." These agents can listen, reply back, and go about their lives in simulated TinyWorld environments.

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: 🧵

I'm surprised every day by how few people know what's going on in AI outside of our little bubble.

This happened to my dad as well. Probably saved his life!

Want to know how OpenAI's o1 and o3 models work? Chinese researchers figured it out! They dropped a research paper explaining EVERYTHING. Here's what you need to know: 🧵

Prediction: @xai will be equivalent to o1, but will be sassy af and have features other AI companies won’t do. Then, GPT4.5 will be great, lightning fast, and cheap. Then, Claude 4 will be equivalent to o3. Then @OpenAI will drop GPT5, the Omni-model, and continue their lead

Test Time Compute is bigger than anyone realizes. It's the most important breakthrough in AI since Transformers. Let me explain...🧵

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: 🧵

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

.@MistralAI launched a ton of new AI features/models today! The best part? It's all absolutely free. Here's everything 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: 👇

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.

AWS (@awscloud) just dropped Automated Reasoning! Automated reasoning was previously only available to the largest companies with massive resources. AI changes that. Let me show you why this is such a big deal 🧵 (AWS Partner)

Google just announced Willow, their groundbreaking quantum chip! Quantum might break a lot of things we know today...like your passwords, crypto, and more. Let me break this down... 👇

gpt4o is no where near gemini 2.5 pro at coding. like...not even close.

PSA: you do NOT need to use DeepSeek R1 hosted by DeepSeek. Don’t send your data to China. Instead, there are already multiple options both hosted and local. This list is growing. * @lmstudio (local) * @ollama (local) * @GroqInc (hosted) What did I miss?

Wait…o3 thought for 16 hours to reach 87.5%??

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.

Gemini 2.5 Pro is much more collaborative when Vibe Coding. It asks me clarifying questions along the way, which no other model has done.

Life achievement unlocked. Thanks for chatting with me @satyanadella !

Kill ‘em with kindness, Sam!

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!

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): 👇

lol to anyone who thinks we hit a wall. The only wall I see is that curve

Gemini 2.5 Pro is the best coding model I've ever used. Video coming soon.

Google just announced Android XR! A new operating system for extended reality devices like headsets and glasses, built with AI (Gemini 2.0) from the ground up. Here's the wild vision of what the future of AI looks like: 🧵

.@OpenAI dropped a new research paper showing AI agents are now capable of replicating cutting-edge AI research papers from scratch. This is one step closer to the Intelligence Explosion: AI that can discover new science and improve itself. Here’s what they learned: 🧵

Uninstalling and reinstalling ChatGPT enabled advanced voice access for me.

So...an OpenAI operating system is inevitable, right?

Google just dropped Gemini 2.0! Tons of awesome things just released. Here's everything you need to know: 🧵

I'll remember this week for the rest of my life. Here's my interview with the incredible @sundarpichai, CEO of Google. We spoke about Gemini, world models, self-improving AI, AI anxiety, and more. Full vid below: 0:00 - Intro 0:40 - Gemini as a World Model 1:30 - Gemini Diffusion 2:33 - Self-improving AI 4:09 - What AI needs to improve 5:06 - Agent Memory 6:46 - AI Form Factors 7:58 - Google Search Future 9:33 - AI Anxiety

My interview with @AravSrinivas, CEO of @perplexity_ai. We discuss their new AI-first browser Comet, how the internet is changing with agents, competition with Google and others, workforce automation and more!

Elon’s version of “unbiased” just means aligned with his views.

This week has been a dream. Thank you @sundarpichai for the great conversation.

Sam Altman to Elon: I turned OpenAI into what it is today, get over it.

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

Sakana AI, the automated research scientist, is an open-source project! Should I make a tutorial for it? github.com/SakanaAI/AI-Scien…

Audible groans at the $250/mo Google AI Ultra plan

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

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

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 to the point where Sam Altman says it'll be so cheap it won't be worth metering. Chip companies are making massive gains with on-device AI that can run faster and more efficiently. The future will be: 95% of consumer AI compute will be executed on-device. For the final 5%, the fastest and cheapest cloud AI provider will be used. Developers will build AI apps for the next few years using the fastest and cheapest cloud AI provider. Because LLMs will be a commodity, most AI developers won't want to spend their time building their own AI infrastructure. They will want whatever is the most efficient (cost/speed) and most consistent. Eventually, AI developers will tap into the already available on-device AI models. I am a firm believer in on-device AI, here's why: * Massive algorithmic gains are happening every week (example: MoA) that allow small models to perform nearly as well as frontier models * On-device AI is essentially free * Running models on-device, with chips that have been honed for decades for mobile devices and thus are incredibly efficient, will reduce energy usage by orders of magnitude * The best use case for consumer AI - AI assistants - requires giving AI access to personal information, which is the most private and secure when you don't have to send that info to 3rd party services * Latency makes or breaks the experience with AI, especially for AI assistants. Latency will always be faster on-device because there's no requirement to ping a 3rd party server That DOES NOT mean there won't be value for cloud AI providers. In fact, there will be tremendous value. The fastest and cheapest providers will win for the use cases that just cannot be run on-device. For high-scale use cases, which typically are performed at the enterprise level, cloud providers will be insanely valuable. Spinning up thousands or even millions of cutting-edge AI agents to do the work of entire organizations can only be done in the cloud. Want to know the best part of all of this? Consumers win big :)

.@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 other industry. blogs.nvidia.com/blog/nemotr…

Introducing: Runner H 🏃🤖 After raising a $200m seed round, @hcompany_ai just dropped information about Runner H. It's a specialized 3B model that can accomplish real-world tasks for you. Details and WILD demos below 🧵

😂 never leaving this app

Grok-3 is now my goto model. Speed and up-to-date info is everything to me.

Rude awakening incoming

I’ve been too pessimistic about my humanoid robot timeline predictions and for that, I apologize.

OpenHands (formerly OpenDevin) is currently the top trending repo on Github. It's a Devin clone that aims to make coding fully AI-powered. Have you given it a try?

First full day living in San Francisco 🤣

.@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 model and cloud service is also owned by Apple. They only outsource to OpenAI for general “world knowledge”, since their model wasn’t built for that. And they treat OpenAI as slightly more than an API call. It is not as deeply integrated into the OS as you said. I believe this is why the stock ripped after; investors realized Apple wasn’t as dependent on OpenAI after all. I was surprised to hear you be so wrong about these facts. Love you still. Cc @Jason @chamath @friedberg

You can use Michael Jackson's likeness...for now.

⭐️Trending GitHub project: Screenshot-to-Code! Take a screenshot of any website and make a clone in minutes. Completely local, use any model. Link below 👇

.@OpenAI just dropped a state-of-the-art OPEN SOURCE model. gpt-oss is a frontier model that anyone can download, modify, and run on consumer-grade hardware. Here's everything you need to know...

Your team reached out earlier and handled it. thanks ❤️

OpenAI just announced Canvas, a new way to code with ChatGPT. Seems like OpenAI doesn't want to just be the platform layer.

Superintelligence or bust. Dylan Patel (@dylan522p) breaks down the AI arms race. Meta, OpenAI, Apple, NVIDIA, xAI, Microsoft...we cover it all. Full interview below 👇 0:00 - Intro 1:15 - Meta’s llama struggles 6:20 - ScaleAI acquisition and $100m offers 15:32 - OpenAI’s relationship with Microsoft 22:28 - GPT4.5, what went wrong? 28:43 - Apple’s AI blunders 31:07 - Apple vs. NVIDIA Beef 33:43 - On-device vs. Cloud AI 39:39 - NVIDIA vs. AMD 48:47 - Grok and xAI 52:49 - Grok 3.5 54:16 - AI and Job Loss 1:00:59 - Who wins Super Intelligence race?

The foresight of Sam Altman is insane

GPT 4.5o? o1 Ultra for $2000/mo?

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: 👇

Google's Genie 3 was just announced. This is a world model capable of generating fully interactive 3D worlds. This has major implications for video games and giving agents infinite self-play environments.