Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: Photorealistic Times Square at dusk with dramatic purple-blue lighting, every billboard showing massive illuminated scales of justice weighing human figures against circuit boards and AI symbols, pedestrians walking below bathed in the glow of ethical questions, cinematic composition with strong foreground-background depth, moody atmospheric perspective.
The jump from “”agents are nowhere close to working”” to “”okay, narrow agents for research and coding work pretty well”” to (very recently) “”general purpose agents are actually useful for a range of tasks”” has been quick enough (less than a year) so that most people have missed it.”” / X https://x.com/emollick/status/1972141975458796020
https://t.co/321zjlMTmp paper doesn’t mention “”AGI”” but if you consider that we used to define AGI as “”outperform humans at most economically valuable work” then surely GDPVal is the most direct AGI benchmark we have ever had and we are between 77-95% of the way there and should https://x.com/swyx/status/1971427791770882463
I’m not sure that we should take the GDPval benchmark that literally I don’t believe that crossing the 50% win rate would actually translate to real world performance the trend is important”” / X https://x.com/scaling01/status/1971432462820802834
We aim to build the most intelligent and useful AI. But “useful” is a fuzzy word. GDPval (consisting of tasks spanning 44 occupations across the top 9 sectors contributing to US GDP) makes “”usefulness”” more concrete. Would love to saturate this one!”” / X https://x.com/markchen90/status/1971449404734439831
The gdpval dataset from @OpenAI is number one trending on @huggingface this week! https://x.com/ClementDelangue/status/1972640079559749632
Tejal Patwardhan on X: “Understanding the capabilities of AI models is important to me. To forecast how AI models might affect labor, we need methods to measure their real-world work abilities. That’s why we created GDPval. https://t.co/YsQvmdGK94″ / X
https://x.com/tejalpatwardhan/status/1971249532588741058
Amazon is training humanoids to move boxes. Makes sense! OmniRetarget is a data generation engine that enables complex loco-manipulation for humanoids. It uses offline retargeting from human MoCap datasets and augments data from single demos to produce 8 hours of trajectories https://x.com/TheHumanoidHub/status/1973489480813388240
We’re at an inflection point in AI’s impact on cybersecurity. Claude now outperforms human teams in some cybersecurity competitions, and helps teams discover and fix code vulnerabilities. At the same time, attackers are using AI to expand their operations. https://x.com/AnthropicAI/status/1974199155657748868
maybe the most impressive part from Sonnet 4.5 alignment information. Not only can it push back, but it has a sophisticated theory of user’s mind. Other models also can speculate about the user’s play (DS does that a lot) but aren’t trained to treat it as actionable info. https://x.com/teortaxesTex/status/1973264029599842380
AI for Cyber Defenders \ red.anthropic.com https://red.anthropic.com/2025/ai-for-cyber-defenders/
OpenAI’s Stargate project to consume up to 40% of global DRAM output — inks deal with Samsung and SK hynix to the tune of up to 900,000 wafers per month | Tom’s Hardware https://www.tomshardware.com/pc-components/dram/openais-stargate-project-to-consume-up-to-40-percent-of-global-dram-output-inks-deal-with-samsung-and-sk-hynix-to-the-tune-of-up-to-900-000-wafers-per-month
OpenAI staff grapples with the company’s social media push | TechCrunch https://techcrunch.com/2025/10/01/openai-staff-grapples-with-the-companys-social-media-push/
OpenAI is building a social network | The Verge https://www.theverge.com/openai/648130/openai-social-network-x-competitor
Very excited to see the Tinker release! @pcmoritz and I had a chance to experiment with the API. It does a nice job of providing flexibility while abstracting away GPU handling. Here’s a simple example showing how to generate synthetic data and fine tune a text to SQL model.”” / X https://x.com/robertnishihara/status/1973455582603649430
Tinker provides an abstraction layer that is the right one for post-training R&D — it’s the infrastructure I’ve always wanted. I’m excited to see what people build with it. “”Civilization advances by extending the number of important operations which we can perform without”” / X https://x.com/johnschulman2/status/1973450054238347314
A flexible API for fine-tuning LMs – Tinker by @thinkymachines Write a simple CPU-only script, and it runs your exact training loop on distributed GPUs. You can fine-tune open models like Llama and Qwen, up to large MoE (Qwen3-235B-A22B), switching them by changing only one https://x.com/TheTuringPost/status/1973827605448306883
Really excited and proud to see Qwen models are in the first batch of supported models for the tinker service! 🤩 we will continue to release great models to grow research in the community 😎 https://x.com/wzhao_nlp/status/1973603599616974970
I’ve been using Tinker at Redwood Research to RL-train long-context models like Qwen3-32B on difficult AI control tasks – specifically teaching models to write unsuspicious backdoors in code similar to the AI control paper. Early stages but seeing some interesting backdoors 👀”” / X https://x.com/ejcgan/status/1973449963259699284
It turns out that the AI jagged frontier worked as a reverse salient, a term from the history of science for a technology or process that holds back the whole system & thus a focus of development. Math & planning were reverse salients, so they have seen the most improvement. https://x.com/emollick/status/1973148208894451908
I had the chance to try @thinkymachines’ Tinker API for the past couple weeks. Some early impressions: Very hackable & lifts a lot of the LLM training burden, a great fit for researchers who want to focus on algs + data, not infra. My research is in RL, and many RL fine-tuning”” / X https://x.com/tyler_griggs_/status/1973450947218252224
Tinker is cool. If you’re a researcher/developer, tinker dramatically simplifies LLM post-training. You retain 90% of algorithmic creative control (usually related to data, loss function, the algorithm) while tinker handles the hard parts that you usually want to touch much less”” / X https://x.com/karpathy/status/1973468610917179630
🚀With early access to Tinker, we matched full-parameter SFT performance as in Goedel-Prover V2 (32B) (on the same 20% data) using LoRA + 20% of the data. 📊MiniF2F Pass@32 ≈ 81 (20% SFT). Next: full-scale training + RL. This is something that previously took a lot more effort”” / X https://x.com/chijinML/status/1973451597393883451
thinking-machines-lab/tinker-cookbook: Post-training with Tinker https://github.com/thinking-machines-lab/tinker-cookbook
[1 Oct 2025] Thinking Machines’ Tinker: LoRA based LLM fine-tuning API https://x.com/Smol_AI/status/1973622595124863044
Announcing Tinker – Thinking Machines Lab https://thinkingmachines.ai/blog/announcing-tinker/
One interesting “”fundamental”” reason for Tinker today is the rise of MoE. Whereas hackers used to deploy llama3-70B efficiently on one node, modern deployments of MoE models require large multinode deployments for efficiency. The underlying reason? Arithmetic intensity. (1/5) https://x.com/cHHillee/status/1973469947889422539
Very excited to see the Tinker release by @thinkymachines! @robertnishihara and I had a chance to experiment with the API, see https://x.com/pcmoritz/status/1973456462346424641
Tinker – Thinking Machines Lab https://thinkingmachines.ai/tinker/
OpenAI’s social creation & consumption experiment is here. Unique handling of identity – if you opt-in others can use your likeness in their creations. You’re notified even if it’s used in a draft post & “liveness checks” are done to prevent impersonation. https://x.com/bilawalsidhu/status/1973103500511871277
Had fun being in Germany to launch a sovereign cloud offering with SAP and Microsoft; important to us to help governments use our frontier models.”” / X https://x.com/sama/status/1971433413086499044
An AI Index for all our customers https://blog.cloudflare.com/an-ai-index-for-all-our-customers/
The most important thing here: the models completed these tasks 100x faster and cheaper than the industry experts”” / X https://x.com/scaling01/status/1971431825433374866
We’re beginning to roll out parental controls in ChatGPT, including the first-of-its-kind safety notification system to alert parents if their teen may be at risk of self-harm. Read more here: https://x.com/fidjissimo/status/1972602249907146967
Introducing parental controls | OpenAI https://openai.com/index/introducing-parental-controls/
My favorite thing about the new college grads that I’ve hired is that they don’t ask you how to do stuff They just put it in ChatGPT and fucking try even if wrong So much better than new grads asking you how to do shit without trying Chat is breeding agency into kids”” / X https://x.com/dylan522p/status/1971425552902082941
Unitree CEO Wang Xingxing at a Trade Fair in Hangzhou on Saturday: ⦿ Unitree R1 will become the world’s best-selling humanoid robot next year. ⦿ In the first half of this year, the domestic robot industry grew an average rate of 50% to 100% for Chinese intelligent https://x.com/TheHumanoidHub/status/1973158573317501243
Unitree CEO Wang Xingxing expects R1 to be the world’s best-selling humanoid robot next year. Won’t shock anyone if it happens. The company announced the starting price of $5,900 but even at $12k this will sell like hot cakes https://x.com/TheHumanoidHub/status/1973452915366044096
Bring me the healthiest snack.”” The robot goes to the kitchen and gets the snack, fully autonomously. This is the first public demo of NVIDIA’s Isaac GR00T N1.6 foundation model presented by Yuke Zhu at CoRL 2025. The previous versions focused only on bimanual stationary https://x.com/TheHumanoidHub/status/1972698708975440349
Go check out @yukez’s talk at CoRL! Project GR00T is cooking 🍳”” / X https://x.com/DrJimFan/status/1971370444474417658
Governor Newsom signs SB 53, advancing California’s world-leading artificial intelligence industry | Governor of California https://www.gov.ca.gov/2025/09/29/governor-newsom-signs-sb-53-advancing-californias-world-leading-artificial-intelligence-industry/
Sam Altman’s playbook is clear: When a model becomes usable, OpenAI rushes to turn it into a blockbuster app: ChatGPT – chat app Codex – coding app Sora – video app Raw model power isn’t a strong moat in a crowded field (Google, xAI, Anthropic). But once an app embeds into”” / X https://x.com/Yuchenj_UW/status/1973435314195800392
Excited to launch Sora 2! Video models have come a long way; this is a tremendous research achievement. Sora is also the most fun I’ve had with a new product in a long time. The iOS app is available in the App Store in the US and Canada; we will expand quickly.”” / X https://x.com/sama/status/1973072672683684239
Obsessed with Rick and Morty explaining 3D Gaussian Splatting. Sora 2 nails it – and yes they really are training on everything by default. https://x.com/bilawalsidhu/status/1973451863442989155
The voice cloning quality in the Sora 2 app is REALLY impressive. Wonder if this is the same tech behind “”Voice Engine”” which OpenAI never released because they were worried about just how good it was. https://x.com/bilawalsidhu/status/1973229885742051465
I tried this test with Sora 2. It fails and the output looks a lot more fake than Veo 3. But interestingly, the audio output that narrates the scene gets the explanation right. https://x.com/fofrAI/status/1973745038195830891
I have been warning about this for a couple years (the post below is from February 2023), but you really cannot trust any image or video you see online. It isn’t just Sora 2, it is a host of tools (many open source) that make cloning voice & images easy. https://x.com/emollick/status/1973461311649718302
This is legitimately mind-blowing… How the FUCK does Sora 2 have such a perfect memory of this Cyberpunk side mission that it knows the map location, biome/terrain, vehicle design, voices, and even the name of the gang you’re fighting for, all without being prompted for any of https://x.com/elder_plinius/status/1973124528680345871
Everyone should throw down for the Sora 2 gpu fund. The nano banana reference was elite. Turns out Sam can jump over a chair, and no amount of popped collars is enough. https://x.com/bilawalsidhu/status/1973113511887905021
Bruh scrolling sora just is videos of people saying “theres a secret feature of sora app double tap and an emoji appears” because it gives them a like, bleh”” / X https://x.com/Teknium1/status/1973347911980454274
My test of any new AI video model is whether it make an otter using wifi on an airplane Here is Sora 2 doing a nature documentary… 80s music video… a thriller… 50s low budget SciFi film… a safety video.. film noir… anime… 90s video game cutscene… French arthouse https://x.com/emollick/status/1973220923810652523
No wonder Snap’s stock dropped 7% yesterday. OpenAI did something with Sora 2 nobody else could by nailing three things at once: 1. Consumer first interface (like Meta Vibes attempted) 2. Production grade output (like Google Veo 3 has) 3. Social collaboration built in (made to”” / X https://x.com/bilawalsidhu/status/1973406327058661815
The labs learned from the Studio Ghibli thing that images & video could produce viral moments that turn into user gain. The Sora 2 launch is the ultimate implementation of this: gated invites, an app that selects for virality, reasons to share with friends, provocative content”” / X https://x.com/emollick/status/1973424720549929054
3 quick thoughts on the Sora app 1. OpenAI was smart to feature their team in the Sora launch video memes. High-fidelity output + showing that you buy what you sell + having a sense of humor deflects some slop-dealer hate. 2. “Cameo” feature does what the Ghibli template did:”” / X https://x.com/anuatluru/status/1973125101047451830
Sora 2 can solve questions from LLM benchmarks, despite being a video model. We tested Sora 2 on a small subset of GPQA questions, and it scored 55%, compared to GPT-5’s score of 72%. https://x.com/EpochAIResearch/status/1974172794012459296
Sora 2 might be cooked. Veo 3 seems better. Look at the artifacts and inconsistency in the background 💀 https://x.com/scaling01/status/1973076175342756152
we’re scaling-up invites to Sora tomorrow! as a heads-up, when more people come onto the platform, gens/day will need to decrease to support the increased volume. we’ll keep everyone posted as things change. in the meantime, enjoy the expanded limits! https://x.com/billpeeb/status/1973260687553208344
Sora 2 – Sam Altman https://blog.samaltman.com/sora-2
Sora 2 is here. https://x.com/OpenAI/status/1973075422058623274
We are launching a new app called Sora. This is a combination of a new model called Sora 2, and a new product that makes it easy to create, share, and view videos. This feels to many of us like the “ChatGPT for creativity” moment, and it feels fun and new. There is something”” / X https://x.com/sama/status/1973073987023352250
Sora 2 is cooming “ OPENAI IS PREPARING to launch a stand-alone app for its video generation AI model Sora 2, WIRED has learned. The app, which features a vertical video feed with swipe-to-scroll navigation, appears to closely resemble TikTok except it’s AI-generated. “ https://x.com/apples_jimmy/status/1972756684297978256
Sora 2 Pro is now rolling out so here’s one more vid. 15 seconds ( max length ) at high quality. Very nice. https://x.com/apples_jimmy/status/1973979773354586379
Sora https://sora.chatgpt.com/explore
Sora 2 is here | OpenAI https://openai.com/index/sora-2/
I found this to be a pretty solid overview of what Sora 2 can do and where it falls short. Super helpful! My take from this: Sora 2 samples look super impressive — but it seems just like with Sora 1, things were polished for marketing.”” / X https://x.com/Tim_Dettmers/status/1973728079395856396
So, generating Sora 2 videos and scrolling the feed is my new favorite thing. I love it. The ultimate AI slop machine. https://x.com/ostrisai/status/1973210333532811419
This kind of re-prompting layer is used in some other video generation models, such as HunyuanVideo, to improve the quality and diversity of outputs. We don’t know for sure whether this is happening in Sora, but it wouldn’t be surprising.”” / X https://x.com/EpochAIResearch/status/1974172901567004762
i have the most liked video on sora 2 right now, i will be enjoying this short moment while it lasts cctv footage of sam stealing gpus at target for sora inference https://x.com/GabrielPeterss4/status/1973120058907041902
Sora 2 is out! I’m incredibly impressed that despite Sora 1 setting a ridiculous bar — as the largest jump in video capability basically ever — Sora 2 somehow matches up here. For people that played around a lot with Sora 1, it didn’t feel clear that Sora really ‘understood’”” / X https://x.com/willdepue/status/1973089331284681110
TODAY WE LAUNCH SORA 2, THE WORLDS BEST VIDEO GENERATION MODEL feature you and your friends with raw real world physics, putting an end to the uncanny ai vibes let me show you how insane our model is, featuring me & sam altman: https://x.com/GabrielPeterss4/status/1973071380842229781
This is the Sora app, powered by Sora 2. Inside the app, you can create, remix, and bring yourself or your friends into the scene through cameos—all within a customizable feed designed just for Sora videos. See inside the Sora app👇 https://x.com/OpenAI/status/1973087446469406732
Sora 2 deeply understands Indian uncles 😭 https://x.com/bilawalsidhu/status/1973273224147509437
Physics with Sora 2 …and some anime. https://x.com/OpenAI/status/1973143639200243959
My favorite trend in the Sora app is these body cam footage videos This clip with Spongebob hit 1M+ TikTok views! 🤯 I built a workflow to remove the Sora 2 watermarks👇 https://x.com/angrypenguinPNG/status/1974144279955325191
How can Sora solve these questions, despite being a video model? One explanation: Sora users’ prompts might be rewritten by an LLM before video generation. In that case, the LLM layer might first solve the problem, then include the solution explicitly in the rewritten prompt.”” / X https://x.com/EpochAIResearch/status/1974172889676177682
OpenAI Is Preparing to Launch a Social App for AI-Generated Videos | WIRED https://www.wired.com/story/openai-launches-sora-2-tiktok-like-app/
Sora 2 is here”” / X https://x.com/gdb/status/1973075357940195665
Sora-2 counting the R’s https://x.com/scaling01/status/1973414141370179876
so.. apparently sora 2 is also a browser it’s wild to see what types of capabilities emerge in the model this is sora 2 rendering pasted html (actual browser-rendered html on the right) https://x.com/jesperengelen/status/1973147038499086523
“sora is number 1 in the app store! it’s been epic to see what the collective creativity of humanity is capable of so far. team is iterating fast and listening to feedback. feel free to drop us feature requests! (we’re sending more invite codes soon, i promise!) https://x.com/billpeeb/status/1974035563482116571
Sora2 from @openai released just 2 days ago, and people are creating and remixing incredible things. The top video with 2.2K likes is a selfie video of Jesus during the last supper, but it only starts there. Sora allows folks to remix any video by adding their own prompt. Here https://x.com/altryne/status/1973568567489798144
Sora: Creating video from text | Hacker News https://news.ycombinator.com/item?id=39386156
“The crazy thing is you can recursively jailbreak Sora” 😅 https://x.com/fabianstelzer/status/1973300086949253410
Sora 2 delivers. It nails accents, aesthetics, and actually has comedic timing. My initial tests making AI videos across a variety of styles 👇 https://x.com/bilawalsidhu/status/1973151157137842416
Pretty much my entire feed on @OpenAI Sora 2 is now just videos saying “”Double tap on Sora videos to see an emoji of the vibe of the video. Leave a comment if you didn’t know””. And there are thousands of these engagement farming videos polluting the feed.”” / X https://x.com/ostrisai/status/1973476918285709398
Introducing Vibes: A New Way to Discover and Create AI Videos https://about.fb.com/news/2025/09/introducing-vibes-ai-videos/
Anyone who sees this video can instantly grasp the (at least) potential for malicious use. And yet nobody with any power (either in the public or at the corporate level) has anything to say (let alone do) to address it, or even acknowledge it.”” / X https://x.com/TheStalwart/status/1973372434133950665
yep that’s it we’ve crossed the chasm ai video is now indistinguishable from real video https://x.com/mattshumer_/status/1973077933481677245
A Research Agenda for the Economics of Transformative AI | NBER https://www.nber.org/papers/w34256
I get data sovereignty in some cases, but there is just no way for new countries to join the frontier model race as long as scaling (in any sense) matters. There is no sovereign model. You will be dependent on the production of Chinese (or US or French) open models as a base.”” / X https://x.com/emollick/status/1972018517919826099
Do Humans Really Have World Models? | Daniel Miessler https://danielmiessler.com/blog/humans-dont-have-world-model
This job posting suggests that Meta is developing an egocentric AI system that will form the foundation for AI-enabled humanoid robots and AR devices. https://x.com/TheHumanoidHub/status/1972544881303417338
The sudden viability of economically valuable AI agents has been surprising. This partially as this has happened very quickly & very recently. But also because we had bad assumptions about the limits of agents, which are more self-correcting than expected https://x.com/emollick/status/1973160975290544469
I accidentally ran Codex in a totally unrelated repository again. It found the correct repository from the error message alone (no GitHub link was given), found the source code, cloned it, and started fixing the issue. The commit was 15.3 thousand lines of code”” / X https://x.com/Sauers_/status/1970727099162861788
A new paradigm of proactive, steerable AI – Fidji Simo https://fidjisimo.substack.com/p/a-new-paradigm-of-proactive-steerable
We get transfered SO MANY STUFF out of the box from our ancestors, its ridiculous. My favorite example is that empirically, we seem to have “”snake detection module”” hard-wired to our brain (that detects snake faster than other things) We are, in so many ways, literally https://x.com/cloneofsimo/status/1973655922506605046
How can we enable finetuning of humanoid manipulation policies, directly in the real world? In our new paper, Residual Off-Policy RL for Finetuning BC Policies, we demonstrate real-world RL on a bimanual humanoid with 5-fingered hands (29 DoF) and improve pre-trained policies https://x.com/larsankile/status/1973191635904373243
We’ve focused on improving Claude’s skills in defensive cybersecurity. The results of this are visible in Claude Sonnet 4.5, which is comparable or superior to Opus 4.1 in cybersecurity tasks—yet both faster and cheaper. Read more: https://x.com/AnthropicAI/status/1974199158929305738
I’m feeling like sonnet 4.5 is bad its really really fucking up in ways sonnet 4 and opus 4.1 did not unfortunately”” / X https://x.com/Teknium1/status/1973476714924876218
Sonnet 4.5 with “”significant improvements in sycophancy”” https://x.com/scaling01/status/1972713224727412804
Anthropic to triple international workforce in global AI push https://www.cnbc.com/2025/09/26/anthropic-global-ai-hiring-spree.html
like surely the human graders were influenced by style similar to lmarena”” / X https://x.com/scaling01/status/1971432758817050970
I think the idea of “workslop” is not that helpful, as it places the burden of appropriate AI use on workers who are given AI tools & told to increase productivity, without efforts by managers to figure out which processes to change or define what good AI productivity looks like”” / X https://x.com/emollick/status/1971029270786129928
AI-Generated “Workslop” Is Destroying Productivity https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity
“I think one of the misunderstandings is that AI is this magic wand or it can solve all problems, and that’s not true today. But there is a ton of value when you get it right.” Our CEO @jdroege shared his AI success framework with CNN’s @claresduffy. https://x.com/scale_AI/status/1973478940640715240
Meet Dex (an AI recruiter) https://meetdex.ai/landing-swe-linkedin
chatgpt-usage-and-adoption-patterns-at-work.pdf https://cdn.openai.com/pdf/3c7f7e1b-36c4-446b-916c-11183e4266b7/chatgpt-usage-and-adoption-patterns-at-work.pdf
A critical part of the path to AGI is feeling the warmth of humanity along the way”” / X https://x.com/OfficialLoganK/status/1973194298981339420
The sheer audacity of AI researchers in the 60s https://x.com/bilawalsidhu/status/1972658669507457351
LLMs are a distillation of human knowledge and creativity, and yet they’ve become something greater than the mere sum of its parts.”” / X https://x.com/bilawalsidhu/status/1972716031173910944
Every artificial sweetener was found by chemists who stupidly licked their hands! Like sucralose: “”a chemist in London was asked by his boss in 1976 to “test” a new chemical, but misheard it as “taste”—a potentially lethal error for a chemist, but one that worked out well.”” https://x.com/emollick/status/1971460497636831580
OK, let me see. How GPT 5 Pro shows its thinking traces is interesting. I am working through how useful they are and what patterns they follow. นี่เป็นภาษาไทยเพราะเหตุผลบางอย่าง. I am considering whether anyone is going to get the joke.”” / X https://x.com/emollick/status/1971972668481306980
Slack is giving AI unprecedented access to your workplace conversations | VentureBeat https://venturebeat.com/ai/slack-is-giving-ai-unprecedented-access-to-your-workplace-conversations
SB 53, the landmark AI transparency bill, is now law in California | The Verge https://www.theverge.com/ai-artificial-intelligence/787918/sb-53-the-landmark-ai-transparency-bill-is-now-law-in-california
Introducing parental controls in ChatGPT. Now parents and teens can link accounts to automatically get stronger safeguards for teens. Parents also gain tools to adjust features & set limits that work for their family. Rolling out to all ChatGPT users today on web, mobile soon. https://x.com/OpenAI/status/1972604360204210600
CharacterAI removes Disney characters after receiving cease-and-desist letter | TechCrunch https://techcrunch.com/2025/10/01/character-ai-removes-disney-characters-after-receiving-cease-and-desist-letter/
Daiwa Securities is hiring startup Sakana AI to build an AI tool analyzing investor profiles, joining other firms adopting the technology (Bloomberg: https://x.com/SakanaAILabs/status/1974109165623853365
We are pleased to announce our partnership with Daiwa Securities, a major financial services firm in Japan. https://x.com/SakanaAILabs/status/1973935631354245286
The Secret Power of Global Load Balancing in MoE Models New research reveals how a simple change to load balancing in MoE architecture delivers remarkable performance gains 📈 Traditional frameworks calculate Load Balancing Loss (LBL) within each micro-batch, forcing even https://x.com/daddyofadoggy/status/1973759113554174251
A senior government official of the UAE, Abdulla M. Alhamed, met Optimus and Elon at Tesla HQ in California. https://x.com/TheHumanoidHub/status/1972093872177401983
We applaud @CAgovernor for signing @Scott_Wiener’s SB 53, establishing transparency requirements for frontier AI companies that will help us all have better data about these systems and the companies building them. Anthropic is proud to have supported this bill.”” / X https://x.com/jackclarkSF/status/1972773280877826232
We’re updating GPT-5 Instant to better recognize and support people in moments of distress. Sensitive parts of conversations will now route to GPT-5 Instant to quickly provide even more helpful responses. ChatGPT will continue to tell users what model is active when asked.”” / X https://x.com/OpenAI/status/1974234951928459450
Partnering with AARP to help keep older adults safe online | OpenAI https://openai.com/index/aarp-partnership-older-adults-online-safety/
i get the vibe here, but… we do mostly need the capital for build AI that can do science, and for sure we are focused on AGI with almost all of our research effort. it is also nice to show people cool new tech/products along the way, make them smile, and hopefully make some”” / X https://x.com/sama/status/1973381552621887706
Ok. This is art. The art of slop. Divine slop. OpenAI, I concede. You have cooked. https://x.com/teortaxesTex/status/1973369392944812069
Taste is your moat — with Dylan Field, Figma – Latent.Space https://www.latent.space/p/figma
Founder Of AI Indie Particle6 Launches New AI Talent Studio Xicoia https://deadline.com/2025/09/eline-van-der-velden-particle6-ai-talent-studio-xicoia-1236555680/
The personalized web is here Ads with you in them Cool or creepy? https://x.com/bilawalsidhu/status/1971154612355989971
NEO humanoid robot by 1X uses a vacuum cleaner. Truly general-purpose robots have to be able to work with tools built for people. https://x.com/TheHumanoidHub/status/1972077093363327135
Security researchers have discovered that Unitree robots, like the humanoids G1 and H1 and quadrupeds Go2 and B2, suffer from a severe security hole in their Bluetooth Low Energy (BLE) setup for the Wi-Fi configuration interface. This flaw lets attackers perform command https://x.com/TheHumanoidHub/status/1971314995779748126
Endpoint Security for AI eBook https://www.delltechnologies.com/asset/en-us/solutions/business-solutions/briefs-summaries/endpoint-security-for-ai-ebook.pdf
Own AI Securely with SANS | SANS Institute https://www.sans.org/mlp/ai-security-blueprint
AI Security Starts Here | SANS Institute https://www.sans.org/mlp/artificial-intelligence
DevSecCon: Securing the Shift to AI Native | Register for Free | Oct ’25 | Snyk https://snyk.io/events/devseccon/
Last week we found an issue with SWE-Bench, allowing agents to cheat by looking at future commits. Instead of celebrating the SWE-Bench Devs for quickly fixing the issue and being transparent, the HN crowd is dunking on them and drawing wildly inaccurate conclusions about”” / X https://x.com/TacoCohen/status/1966421688846778561
Its kind of funny that AI can definitely do most common CAPTCHAs better than humans and the reason that CAPTCHAs still work is because the big LLMs often refuse to do them. https://x.com/emollick/status/1972177860086612283
The narrative around LLMs is that they got better purely by scaling up pretraining *compute*. In reality, they got better by scaling up pretraining *data*, while compute is only a means to the end of cramming more data into the model. Data is the fundamental bottleneck. You can’t”” / X https://x.com/fchollet/status/1972477946700190081
When should an LLM learn to reason? 🤔 Early in pretraining or late in fine-tuning? Our new work, “”Front-Loading Reasoning””, challenges the “”save it for later”” approach. We show that injecting reasoning data into pretraining is critical for building models that reach the https://x.com/__SyedaAkter/status/1973841632249172096
Modular Manifolds – When we train large neural networks, we need to keep them healthy. We do not want the tensors in the network—either the weights, activations or gradients—to grow too large or too small. Thinking Machines Lab https://thinkingmachines.ai/blog/modular-manifolds/
We are building Grokipedia @xAI. Will be a massive improvement over Wikipedia. Frankly, it is a necessary step towards the xAI goal of understanding the Universe.”” / X https://x.com/elonmusk/status/1972992095859433671
Man, imagine being Mark Zuckerberg, spending billions to build a slop machine, only for another slop machine to out-slop you just days later.”” / X https://x.com/skirano/status/1973248497853018168
