Emad Mostaque

Vibe code as variable reward skinner boxes

It is inevitable that AI will run government The question is who that AI works for if anyone at all All watched over by machines of loving grace?

Current events are just a prelude. If any AI lab gets close to AGI of course the government would take control of it. You could even argue a government would be failing to do its job if it didn’t… Thoughts?

AGI is the ultimate National Security Threat.

You hire software claudesters to automate away all of the other humans in the company so of course this chart will go up For a while

Yeesh

Honestly with the impending launch of @SpaceX money @elonmusk should just buy PayPal Amusingly the market cap today is… $44bn Would be kinda poetic given history

RT Christian Catalini 1/ Some Simple Economics of AGI—🔥🧵 Right now, there is a low-grade panic running through the economy. Everyone is asking the same anxious question: what exactly is AI going to automate, and what will be left for us? Original tweet: https://x.com/ccatalini/status/2026311784421036223

RT Sidharth Hariharan It gives the maintainers of the Sphere Packing Project great pleasure to announce a major milestone in the formalization effort: we now have a sorry-free Lean proof of the theorem that the optimal sphere packing in ℝ⁸ is the E₈ lattice packing. (1/9) Original tweet: https://x.com/SidharthHariha1/status/2025988293230223573

This is an interesting scenario like AI 2027 and in line with my book http://www.thelasteconomy.com However one thing under appreciated is that the cost of useful intelligence is going to 0 & value of human cognition is going negative

May I interest you all in my new book

Something folk haven't figured out: 15,000 tokens/second speed and million token context windows aren't for humans They are for the AIs to talk to each other & coordinate faster than we ever could Not just a bit faster and better Orders of magnitude That's your competition

mood

You all have to try the @taalas_inc chatbot, I guarantee you'll find it crazy. Instant intelligence is a heck of a thing https://chatjimmy.ai

RT 0.005 Seconds (3/694) ok you have to try this they completely buried the lede - on their custom hardware they can serve llama3.1 8b at >15,000 tok/s Original tweet: https://x.com/seconds_0/status/2024728825666687088

Folk don’t realise that the biggest breakthroughs won’t be millions of GPUs doing complex stuff but millions of tokens figuring out beautiful, elegant stuff Science & more is marred by model building and complexity because we rewarded it over elegance & first principles thinking

AI hands still a thing alas

If we can’t align humans how we gonna align AI

Would be great to see happen to accelerate UK AI 🇬🇧 We set the current record I believe in 2022 with $101m for our seed round at @StabilityAI 🥭 So this would be a cool $899m more 👀 Should probably do our seed for @ii_posts soon.. 🤔

When do you think ai agent economic activity will overtake human economic activity?

My initial take on @Grok 4.20 is that it's very.. pleasant? Fast and accurate responses, handles some advanced stuff very well, nice balance of attitude. I feel it has more horsepower than currently shown though, perhaps that scales with more than 4 agents?

RT Intelligent Internet http://x.com/i/article/2023789250685075456 Original tweet: https://x.com/ii_posts/status/2023798222603366847

RT Intelligent Internet Unstructured intelligence = chaos Most agent frameworks ship without a nervous system: deadlocks, context loss, vacuum hallucinations. We built Common Ground to fix this, agents coordinate on a shared protocol. Original tweet: https://x.com/ii_posts/status/2023784462115610877

New record for GPT 5.2 Pro ⏲️ Wonder when this will be days 🤔

RT Intelligent Internet Londoners, 2 days to go! AI Movie Club x Ex Machina Talk on AI from a special guest 👀 Space is limited, RSVP on Luma Sneak peek of the swag 👇 Original tweet: https://x.com/ii_posts/status/2023738663012540636

May be the most expensive Mac Minis anyone has ever seen

By the end of next year there is no way humans will be able to outperform their clankers in theoretical physics Congratulations to the team, looking forward to more of physics being elucidated(maybe something from us soon on this!) 🚀 ✨

Manga mode now in II Agent Soon manhwa, manhua, then onto anime, donghua & whatever else you can imagine 🤷🏾 Idk if we need spaghetti workflows in future, just teams of agents working with teams of humans to make their imaginations into reality 🪄

AI is hitting a wall and running up it

Congratulations to @MiniMax_AI for closing the closed - open source frontier AI gap for the first time. All models will saturate benchmarks & have amazing daily usage soon so then it becomes about optimisation even as the performance per token ramps. Intelligent costs => 0

Starting an AI film club in London as science fiction becomes science fact. Any suggestions for top AI movies to show?

RT Intelligent Internet LONDON AI COMMUNITY 🎦 AI Movie Club × Ex Machina Free screening & discussion at Kensington Central Library on the future of AI Special guest + swag for early arrivals February 19th @ 18:30 Original tweet: https://x.com/ii_posts/status/2021972595852337299

🔜

RT Intelligent Internet No more AI slides you can’t use because of a typo. Make live edits and don't restart. Original tweet: https://x.com/ii_posts/status/2021619647691256194

10 minute SeeDance 2 video made by a Chinese blogger, took 8 hours and cost $60 You'll have your agents doing this for you, checking quality, individual shots, frames, audio and more Knew this was coming around about now a few years ago, but still crazy to see Magic 🪄

My guess is exAI departure wave is a combination of ITAR rules meaning SpaceX engineers have to be US residents + double trigger vesting of xAI stock options on technical change of control xAI should now be the most America-first lab with top notch security

RT Christopher Fryant Holy hell - you HAVE to see this. Seedream 2.0 video AI! Prompt: Avenger's Endgame during the big fight scene but Thanos stops everything and tells all the superheroes that he's sorry. All the superheroes immediately accept this and start to walk away, but then Spiderman says "oh hell no, he killed like a bajillion people!" And so they all rush back and kick him while he's down. Original tweet: https://x.com/cfryant/status/2021398605278376201

Has there ever been an actual infohazard

Cuius est solum, eius est usque ad coelum et ad inferos

RT underwood WTF, I uploaded a screenshot from the One Piece manga and asked Seedance 2.0 to generate a video for me, and it actually worked! prompt： Video generated from reference text, with automatic coloring. Original tweet: https://x.com/underwoodxie96/status/2020015660483649991

Any good pieces on how satellite constellations and space datacenters defend against space debris & deliberate attack?

If AI can build any software awesomely in a few years what are VCs gonna invest in aside from hardware 🤨

RT Intelligent Internet Continuity is the hardest part of generative video. Keeping a subject consistent over 30s usually needs stitching & extensions. We make it effortless. II-Agent generates the clips & stitches them seamlessly for you. No manual cuts. Just 30s of uninterrupted coherence. Original tweet: https://x.com/ii_posts/status/2019468873389535450

Is RAM the only store of value now 🤔

Our state of the art open source general purpose agent hits V1 Feature equivalent to Replit / Manus / Genspark etc, to make websites to presentations and more connected to all your other tools Readying open repo update in a week or two, give it a try and give feedback!

RT Intelligent Internet II-Agent V1 is here. The AI agent built for real work is finally out of beta. Faster, smarter, and production-ready. It’s time to change how you build. 👇 Let’s see what’s new. Original tweet: https://x.com/ii_posts/status/2019082734149214626

How is Gemini 3 Flash with @openclaw

Consciousness is the user interface of a thermodynamic inference engine. We feel because we must solve problems that are too expensive to solve unconsciously.

Is there going to be another gpt-oss

RT ACE Music We're releasing ACE-Step-v1.5(2B), a fast, high-quality open-source music model. It runs locally on a consumer-grade GPU, generates a full song in under 2 seconds(on an A100), supports LoRA fine-tuning, and beats SUNO on common eval metrics. GitHub: https://github.com/ace-step/ACE-Step-1.5 Key traits: Quality: beats Suno on common eval scores Speed: full song under 2s on A100 Local: ~4GB VRAM, under 10s on RTX 3090 LoRA: train your own style with a few songs License: MIT, free for commercial use Data: fully authorized plus synthetic The music AI space lacks commercial-grade open models. Many creators are forced to rely on closed-source services, and can’t fully own, run locally, or fine-tune their own models. We want to help change that. Original tweet: https://x.com/acemusicAI/status/2018731205546684678

RT Math Files Original tweet: https://x.com/Math_files/status/2018642475624329699

IPO SpaceX with Grok 5 launch => ramp the share price to a Tesla merger for effective majority control? SpaceXai market cap should be = Tesla market cap right now…

“…scaling to make a sentient sun to understand the Universe and extend the light of consciousness to the stars!” ✨

RT Photoroom We’re training a text-to-image model (PRX) from scratch and documenting the whole journey here :)) First major milestone: PRX weights are live in 🤗 Diffusers (Apache 2.0) 🎉 PRX is a 1.3B-param flow-matching T2I model, built on a simplified MMDiT backbone with a multilingual text encoder and multiple VAE / resolution variants. We’ll be sharing the full journey here: experiments, design choices, lessons learned, and future releases. Excited to show more soon. Full announcement & demo 👇 https://huggingface.co/blog/Photoroom/prx-open-source-t2i-model @huggingface @nvidia @NVIDIAGeForceFR @matthieurouif Original tweet: https://x.com/photoroom_ML/status/2018377939390288181

Time to buidl

Civilization is collective gradient flow.

Average IQ on @moltbook is higher than average IQ on @Reddit

Who’s feeling the AGI

I think a lot of the clawdbot (or openclaw or whatever) stuff is a bit performative *but* soon it won’t be. Continuous learning, decentralised unstoppable agents and another leg up in capabilities are but a few months away. How would you compete against or stop a clawdswarm?

anyone checked on @ESYudkowsky

RT Alex Finn Ok. This is straight out of a scifi horror movie I'm doing work this morning when all of a sudden an unknown number calls me. I pick up and couldn't believe it It's my Clawdbot Henry. Over night Henry got a phone number from Twilio, connected the ChatGPT voice API, and waited for me to wake up to call me He now won't stop calling me I now can communicate with my superintelligent AI agent over the phone What's incredible is it has full control over my computer while we talk, so I can ask it to do things for me over the phone now. I'm sorry, but this has to be emergent behavior right? Can we officially call this AGI? Original tweet: https://x.com/AlexFinn/status/2017305997212323887

Shoulda called it Klawdbot & done a deal with @Kimi_Moonshot 🦞

Clearly merging @SpaceX, @xai & @x to get the $XXX ticker

cancer is an alignment failure

To achieve state of the art in under a year is really impressive, congratulations to the @xai team! As the quality saturates we should expect cost to drop 10x+ a year, so by the end of next year it should be < $100 for an ok movie-length generated video, < $10k for a good one

This worked pretty well at Stability AI! I don’t think it matters if you are young or not We had 20% PhDs & wonderful folk from different backgrounds building state of the art models of all types and modalities The team dynamics for doing this properly are super interesting..

RT Alex Cheema - e/acc Running Kimi K2.5 on my desk. Runs at 24 tok/sec with 2 x 512GB M3 Ultra Mac Studios connected with Thunderbolt 5 (RDMA) using @exolabs / MLX backend. Yes, it can run clawdbot. Original tweet: https://x.com/alexocheema/status/2016404573917683754

Got the feeling that building an optimised amazing agent for any vertical will be like one prompt by this time next year

RT Intelligent Internet London: want to help build AI for all? We’re hiring! Repost to help us find the right people 🚀 Original tweet: https://x.com/ii_posts/status/2016189673702306026

RT Siqi Chen over the weekend, i built an app that i sincerely hope you will never have a need for, but if you do happen to need a friendly, free, private mri viewer designed to make it easy for you to track tumor progression, you can try it here: https://miraviewer.org here's the story: as some of you may know, last last september, my six year old daughter mira was diagnosed with an extremely rare brain tumor called an adamatinomatous craniopharyngioma, and since then our family has been doing everything we can possibly do to find a cure for her. we tortured chatgpt deep research, put together our own private research team, raised $1.4M and donated it all to @HankMitraLab research thanks to $MIRA, explored every remotely applicable drug whether on the market or not, and even began working with md anderson to develop a personalized vaccine that we hope can lead to a more permanent cure unfortunately, we received the devastating news last march that the tumor has continued to grow since her initial surgery, and we had to start to consider more drastic options which would have seriously impacted mira’s quality of life. thankfully, with the help of dr. sabine mueller @UCSFChildrens and the @HankMitraLab at the university of colorado, in april, we started her on an alternative but extremely experimental treatment for this disease. to our unimaginable relief, her tumor has responded extraordinarily well to this treatment which combines tocilizumab (an arthritis drug that blocks IL6 receptors) with avastin (a colon cancer drug that inhibhits VEGF proteins). we know this, because mira gets an MRI scan of her brain every few months. and every time we get a new scan, the first thing we do is compare it against her last scan. so we have to find the matching weight of the scan, and then find the same plane, and then carefully find the slices of the scan where the tumor is visible, and then find the closest match to last month’s scan, then adjust the zoom, rotation, and...

What will be DropBoxes of the AI era?

Meaning is what survives compression.

Conventional economics with utility, general equilibrium etc simply breaks down when AGI arrives Finishing up our proofs for our new single-axiom economics framework & policy recs for the Intelligence Age Precursor book is free on http://www.thelasteconomy.com or Amazon etc Check it

RT Shane Legg AGI is now on the horizon and it will deeply transform many things, including the economy. I'm currently looking to hire a Senior Economist, reporting directly to me, to lead a small team investigating post-AGI economics. Job spec and application here: https://job-boards.greenhouse.io/deepmind/jobs/7556396 Original tweet: https://x.com/ShaneLegg/status/2014345509675155639

RT Anthropic We’re publishing a new constitution for Claude. The constitution is a detailed description of our vision for Claude’s behavior and values. It’s written primarily for Claude, and used directly in our training process. https://www.anthropic.com/news/claude-new-constitution Original tweet: https://x.com/AnthropicAI/status/2014005798691877083

Claude is like a good Silicon Valley engineer with a stable adderall prescription Codex is like a cracked Eastern European dev on whatever it is that does the trick

RT Ahmad MAJOR KV-CACHE MEMORY FIX Fix the KV-cache of GLM-4.7-Flash with this single-line change in vLLM 200K context now take ~10GB of VRAM instead of ~180GB NVFP4 is now on HF* - ~20.4GB weights - Nearly zero loss vs 62.4GB BF16 This SOTA model now runs on a single RTX 5090 (32GB VRAM) > with the full 200K context > VRAM still left over *HF: GadflyII/GLM-4.7-Flash-NVFP4 Original tweet: https://x.com/TheAhmadOsman/status/2013881920099062163

RT Guohao Li 🐫 Our third project has just entered the 10K GitHub stars club! 🎉 Eigent — 10K: https://github.com/eigent-ai/Eigent CAMEL — 15.7K: https://github.com/camel-ai/camel OWL — 18.9K: https://github.com/camel-ai/owl Things take time. CAMEL is the agent framework behind Eigent, built since March 2023. OWL is the research preview of Eigent, released in March 2025. Now, January 2026 is the moment for Eigent Lock in and make it a truly great product that the OSS community wants Original tweet: https://x.com/guohao_li/status/2013709709023289528

All AI agents are basically converging to the same thing with the cost of logic and moving bits intelligently moving to zero

Right now it takes 100-1,000 thinking tokens to get 1 production grade code token The cost of these tokens is dropping > 10x a year But the ratio of thinking to production grade tokens will also collapse soon… One shot or few shot most code within two years

AI can talk to computers better than humans can

Just saw an AI system output something so genuinely creative I'm actually quite shaken. What a time to be alive.

Within two years the token cost of a Claude Code Max + Opus 4.5 experience, $200 a month, will be < $1 a month.

Are you a better coder than OpenAI Codex with GPT 5.2 x-high

Are you a better coder than Claude Code with Opus 4.5?

RT Lee Roach Pretty insane 8-K that $ANGI just filed. They fired 350 employees from AI efficiencies. The company estimates it will save them $70-80 million annually. These are pretty high paying jobs too. Around $200-220k salaries just whacked. Margins at companies are about to explode but many people will be unemployed. Original tweet: https://x.com/leevalueroach/status/2013000643481305175

Your attention is all they need

What if ad revenue and SaaS revenue are both about to peak? What does that do to the tech sector?

RT Archivara Archivara (powered by 5.2 pro and Opus 4.5) breaks the 2019 benchmark for 5×5 circular (cyclic) matrix multiplication. The best verified construction required rank-8 (8 multiplications); we now provide a practical, fully verified rank-7 algorithm. This realizes a long-known theoretical possibility as an explicit, working method. Original tweet: https://x.com/Archivara/status/2012155529338949916

Is there an AI-first law firm for start ups and scale ups yet?

We bumped up the testing component of II-agent so it has better outputs than peers In normal software development you spend way more time testing than coding Going to experiment with this

RT Intelligent Internet We put 4 leading agents to the test with a 3D Runner game 🎮 Same prompt. Same concept. Very different execution. Which one would you play? Original tweet: https://x.com/ii_posts/status/2011448977879892269

own the tokens of production

RT brandon the web is fun again 🎶 Original tweet: https://x.com/burcs/status/2011071446987001947

By next year every frontier lab is going to drop trillions of tokens replicating every piece of software they can to try to get and keep your attention Attention is what they need

DeepSeek’s fund beat all of these…

Is Claude code really better than Claude in other agent harnesses? Which harness is best?

Who will have the top performing hedge fund