Haider.✓
@haider1
open intelligence for a decentralized future
4.3K Following72.6K Followers35.7K Posts
- i think Sergey Brin getting directly involved and Google going code red is finally paying off, much like openai's response to opus 4.5 and gemini 3 brin recently said google is getting onto the "RSI treadmill" with AI models helping build the next generation of models argon will be an early sign of that acceleration113665
- funny how just a month ago everyone was mocking what google was even doing well gemini 4 carbon is your answer. yesterday they launched the genie 4 world model and 15 days ago they announced gemini 4 argon you have to understand that google does far more than just LLMs and pushing progress across multiple fields101592.7K
- google has made huge progress on one of gemini's models biggest weaknesses: hallucinations and overconfident wrong answers gemini 4 argon is still an impressive improvement in hallucination reduction even better than Carbon
Haider. @haider1gemini 4 carbon is now the leading model on the updated AA Index and finally dethroning opus 5.5 after more than a month at the top i smell fable 5.5 very soon55574K - gemini 4 carbon is now the leading model on the updated AA Index and finally dethroning opus 5.5 after more than a month at the top i smell fable 5.5 very soon
Haider. @haider1gemini 4 carbon looks like a big jump in cybersecurity it can handle the entire workflow from finding vulnerabilities to testing exploits and patching the code and it achieved over 2x the secure-code generation success rate of gemini 4 argon in internal tests359511.6K - gemini 4 carbon looks like a big jump in cybersecurity it can handle the entire workflow from finding vulnerabilities to testing exploits and patching the code and it achieved over 2x the secure-code generation success rate of gemini 4 argon in internal tests
Haider. @haider1holy shit. Google launched two SOTA models in just 15 days Gemini 4 Argon Gemini 4 Carbon77676.6K - holy shit. Google launched two SOTA models in just 15 days Gemini 4 Argon Gemini 4 Carbon
Logan Kilpatrick @OfficialLoganKIntroducing Gemini 4 Carbon, it's a big jump over Argon which we released just 2 weeks ago. Gemini 4 Carbon is SOTA on cyber, long horizon tasks, ranks #1 on Artificial Analysis V5, significantly improves on usability in domains like coding, and is more aligned.1151146.1K 
Introducing Gemini 4 Carbon, it's a big jump over Argon which we released just 2 weeks ago. Gemini 4 Carbon is SOTA on cyber, long horizon tasks, ranks #1 on Artificial Analysis V5, significantly improves on usability in domains like coding, and is more aligned.
1K91812.8K739K- another terrible example of how clueless most people are about exponential progress 2 years ago, people in the math subreddit were questioning whether AI can solve unsolved math problems in a few decades. now AI has proposed a solution to one Millennium Prize Problem with more within reach
262650116.1K - there's no doubt anthropic has stronger models internally (fable 6 or maybe even 6.5) but sees no reason to release them while openai is struggling right now they might also be saving major improvements for their IPO and this is why we need strong competition from all sides not just open-source models131836.1K
- i can't believe how quickly we're normalizing AI progress since the navier-stokes breakthrough and even openai's announcements of hundreds of new mathematical proofs are starting to feel expected rather than surprising P vs NP should now be the next benchmark for AI progress77834.4K
- it's crazy that in 2026 you can run a 1.7B model on your phone that beats a 9B model on an agent benchmark LFM3 is a family of 6 open-weight models around 2B parameters spanning text, vision, audio and decision-making -- their 1.7B model even beats Qwen3.5 9B on webArena-verified in @liquidai own tests
Liquid AI @liquidaiIntroducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding > 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages. > 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language. > 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒 Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧵57323.5K 
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding > 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages. > 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language. > 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒 Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧵4711.1K8.3K1.7M- demis is heavily world-model-pilled and Google's Genie 4 announcement today makes that even more obvious he's always believed that scaling LLMs alone might not be enough for AGI and that world models will play a huge role in helping actually understand the world seems like Sergey Brin has his foot fully on the accelerator10423513.2K
- what a time to be alive all seven Millennium Prize problems have now seen breakthroughs or related advances of some kind: P vs NP: ETH disproved; UGC proof announced Riemann Hypothesis: quasi-RH proved and formalized; Landau–Siegel zeros excluded Yang–Mills: existence proof announced and formalized Navier–Stokes: counterexample announced and formalized Hodge Conjecture: partial proof for CM-type abelian varieties Birch–Swinnerton-Dyer: full formula proved for selmer coranks 0 and 1 Poincaré Conjecture: solved by Perelman in 2003
134349336.4K - the fact that sam sees internal AGI as possible by year-end is incredibly exciting oai said Astra marks the start of the AGI era while highlighting the performance of an unreleased internal model and the recent breakthroughs in math, physics and science suggest there might be something much bigger behind that claim861095.7K
- Anthropic CEO Dario Amodei: "gpt-2 was the first real test of the scaling laws hypothesis. i looked at it and said, My God, you scale this thing up another factor of a million, and it'll be better than humans at everything" This is the defining economic and scientific moment of the modern age. We don’t know it yet132947439.2K
- math singularity is well and truly underway after openai's latest announcements of breakthroughs in math, CS and physics, Axiom just announced that its AI agents formalized the longest proof in mathematical history in just 24 days 43.1 million lines of Lean code
Axiom @axiommathaiThe largest math proof in human history is now computer-verified. Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry." About 15,000 pages, built by over 100 mathematicians across generations.591114.8K 
The largest math proof in human history is now computer-verified. Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry." About 15,000 pages, built by over 100 mathematicians across generations.4101.3K6.9K3.5M- openai have achieved decades of mathematical progress in two months: navier–stokes: claimed complete solution riemann hypothesis: claimed quasi-riemann proof yang–mills: claimed existence proof, mass gap still open p vs np: claimed ETH disproof local smoothing: claimed proof in all dimensions kakeya & stein's restriction: claimed solutions quantum chromodynamics: claimed existence proof 100+ longstanding math problems solved 700+ mathematical results published on github16142429.5K
- another big detail from openai's Yang–Mills millennium prize claim: their internal models produced a navier–stokes proof on sept 5 and a yang–mills existence proof just 4 days later on sept 9 which means they had two millennium prize-related breakthroughs in the same week
Dan Roberts @danintheoryWe are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.8101126.2K 
We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.3081.1K7.5K2.4M- openai's internal model (astra's successor) is an absolute beast at scientific discovery here's what they've already achieved: sep 8 — navier–stokes and euler equations breakthroughs sep 21 — 100+ longstanding math problems solved oct 6 — hundreds of mathematical results published oct 21 — new biological discoveries oct 26 — local smoothing conjecture, exponential time hypothesis, yang–mills theory and quantum chromodynamicsGreg Brockman @gdbAnnouncing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory). AI starting to help us improve our understanding of the universe!9192908.8K
Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory). AI starting to help us improve our understanding of the universe!6712K16.1K3.1M- now that anthropic has officially filed for its IPO, this week is the perfect time to drop fable 5.5 given how long they're taking i feel like they're working on a significant jump in capabilities and with recent releases like opus 5.5 and haiku 5.5, things are looking really good for them right now83654.4K
- Show more posts
