← Back to leaderboard
vraserx

VraserX e/acc

Verifiedactive
@vraserx

Teacher by heart, AI enthusiast by curiosity, passionate about inspiring minds, exploring tech, and making learning exciting, human, and future-focused!

RosterAI Model News & ReleasesAI Agents & Productivity Tools
23,585 followers · 458 following · Last polled 3h ago
Perf. Score
53.6
Rank #201
Value Score
61
#105 of 273 priced
Median views
3.3K
p25 2.0K · trailing 7d
Followers
24K
+0.8% 7d
ER (impr.)
2.16%
eng ÷ impressions
ER (foll.)
1.04%
eng ÷ followers
Est. CPM
$30.31
QT-rate basis
WoW views
+34.1%
▲ rising

Rates & value · est. from 3.3K median organic views

fairly priced
Quote tweetbasis
$100
$30.31 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 61 · rank #105 of 273 priced$800 / 1K engagementsrate age unknown (pre-audit import)

Trends

Follower growth
Median views / day
Engagement rate
View distribution · trailing-7d posts

Each blue dot is one organic post. The Performance Score ranks by the median (3.3K), not the mean (11K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (2.0K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
ChatGPT Dots just blew my mind. This is honestly what I always imagined the beginning of AGI would feel like. Not just a smarter chatbot, but an AI that knows your life, keeps learning and actually does things for you. Holy shit. 🤯 https://t.co/kTlQTxSjSxfrozen23d ago240K3.5K348332622.9K2.98%
It’s becoming ridiculously obvious that Astra and GPT-6.1 Sol absolutely destroy Opus and Fable in real-world tasks. The only areas where Anthropic still has an edge are frontend taste and 3D game design. But Tech Twitter is full of developers who think coding is the entire universe of AI. It’s not even close. AI is so much bigger than writing code.2d ago97K64429146131771.04%
Sam Altman is casually dropping absolute bombshells here. OpenAI’s internal models are apparently doing math that public models can’t even come close to. And he’s already talking about putting speed limits on recursive self-improvement because of how dangerous it could get. We’re arguing over chatbot benchmarks while OpenAI might be sitting on something far bigger. What the hell are they cooking behind closed doors? 🔥
1d ago
94K
807
45
60
11
338
1.34%
Still not happy about the Pro usage cut, but unlimited 5.6 Sol is awesome.🥰 If “unlimited” also applies inside Work and Codex, this could actually be huge. https://t.co/seZHHUcCuTfrozen23d ago77K3288514700.60%
Anthropic is cooked long term. Coding AI is rapidly becoming a commodity. In a few years, superhuman coding models will run locally on your phone or laptop for basically nothing. That’s why I think OpenAI is playing the smarter game. They’re going all in on consumer AI, building a massive market for their upcoming hardware devices and potentially humanoid robots. Coding is a feature. AI becoming part of everyday life is the real business.frozen4d ago70K49622964941.01%
OpenAI has two more models coming this year. GPT-6.1 Astra, an early checkpoint of GPT-6.5 Bell, followed by Bell itself before the end of 2026. I expect Bell to basically saturate science benchmarks across math, physics, biology, chemistry, etc. Opus 5.5-level taste or better, a massive jump in complex backend coding, AA to AAA-level game design capabilities, and superhuman computer use. Basically anything a human can do on a computer. 2027 is going to be the year of ASI.frozen7d ago63K94648417981.80%
Demis Hassabis says today’s AI will look like NOTHING compared to what’s coming in just 1–2 years. And this is coming from the guy running Google DeepMind. We’re arguing over which chatbot is better while the people building the frontier models are already worried about what comes next. The next few years are going to be absolutely insane.15h ago60K99310131133462.46%
Okay, now I get it. Dot is an always-on autonomous agent that keeps working indefinitely. That’s insanely compute-hungry. If OpenAI wants Plus users running Dot extensively too, that compute has to come from somewhere. They’re allocating compute from Pro to Plus users.frozen23d ago60K2855311330.59%
Alexandr Wang gets it. The real promise of AGI isn’t better coding agents. It’s giving people their lives back. Imagine AI handling almost everything you have to do, leaving you free to spend your time on what you actually want to do. That’s the future worth building. Not a world where everyone works harder, but one where work is no longer a necessity. And Meta wants to bring that future to billions of people.frozen11d ago55K32136535980.93%
If AI discovers a cure for cancer tomorrow, I couldn’t care less whether it puts thousands of researchers out of work. Imagine telling millions of patients they have to keep suffering because some people need to feel useful. Your career isn’t more important than human lives. Get over yourself.frozen13d ago47K2.5K220216391226.61%
OpenAI has 1.2 BILLION weekly users. Anthropic has an estimated 148 million. And I think OpenAI is going to win the long game. Anthropic is heavily focused on coding, which could become a serious problem. In two years, we might have free, superhuman coding models running locally on our devices. Good luck building a lasting moat around that. Meanwhile, OpenAI is building an entire ecosystem for billions of people, with AI agents, scientific research, multimodal AI and soon its own hardware devices. AI is going to be much bigger than coding. OpenAI understands that. Anthropic seems to be betting otherwise.frozen5d ago46K434241169611.40%
I’m increasingly convinced OpenAI is ~6 months ahead of Anthropic internally. The mistake people make is comparing what’s public and assuming that’s the frontier. OpenAI has every incentive to sit on its best models until it actually needs them. If you already have the lead, why burn your compute advantage and reveal your hand? Anthropic ships something great. OpenAI waits. Then drops the next tier. I don’t think we’re seeing OpenAI’s frontier. We’re seeing what they’re comfortable releasing.frozen18d ago43K357161277391.28%
I genuinely don’t understand why people are so offended by the idea of someone with zero artistic talent being able to create something beautiful using AI. Since when did creativity become an exclusive club where you need years of training just to be allowed in? Sounds a lot like gatekeeping to me.frozen11d ago42K1.1K7929828783.80%
Are you fucking kidding me, OpenAI? Dots is arguably the most exciting ChatGPT feature you’ve launched all year and the entire EU gets excluded AGAIN. Being an AI enthusiast in Europe is becoming a punishment. Absolutely infuriating. https://t.co/gHhpVActH8frozen23d ago41K3049926301.09%
This is why I’m an accelerationist. OpenAI just released 722 mathematical manuscripts across 372 families, largely produced by an unreleased AI model tackling open research problems. Think about what that trajectory means. Today it’s mathematics. Tomorrow it’s drug discovery, aging, materials science, fusion, physics and engineering. The most important property of advanced AI isn’t that it can write emails or generate videos. It’s that it can potentially compress years of human scientific work into hours. Slowing AI down isn’t automatically the cautious position. There is a cost to not accelerating too: discoveries delayed, diseases untreated and technologies that arrive years later than they could have. We should manage the risks. But we should be racing toward systems capable of accelerating science itself. Humanity has spent centuries wishing we had more Einsteins, von Neumanns and Curies. We may be building millions of them.frozen16d ago39K9219660141193.09%
Recreated Fable 5.5’s SVG with GPT-6 Astra. Astra absolutely destroys it. All that glazing over a fucking skill issue. 😂 Fable 5.5 vs GPT-6 Astra https://t.co/BtgdR4Srhpfrozen4d ago38K1751201280.59%
This is why I prefer OpenAI over Anthropic. Anthropic seems mostly obsessed with coding and making pretty frontends. Cool, I guess. Meanwhile, OpenAI is trying to push the boundaries of science, solve problems in physics, math and biology, and make powerful AI accessible to billions of people for free. That’s a much bigger vision of what AI should be.frozen6d ago30K53927752602.31%
OpenAI is dropping a new frontier model next week. It will wipe the floor with Opus 5.5 and goes straight for Fable 5.5. The frontier race is about to get spicy again. 🥰frozen20d ago30K38015332351.57%
Hollywood is COOKED. 💀 Kling 4.0 just dropped and the results are literally indistinguishable from reality. “AI slop” is becoming a thing of the past. We’re witnessing the death of traditional filmmaking in real time. https://t.co/0kFfrWVQrV1d ago24K489222231492.82%
People call a future without jobs dystopian. You know what’s actually dystopian? Being born into a civilization technologically capable of producing abundance, yet being forced to spend 40 years working just to justify your right to exist. But sure, the robots are the problem.frozen12d ago23K1.6K228154181609.34%
If someone with a brilliant imagination but zero drawing skills can finally bring their ideas to life using AI, that’s a good thing. The fact that this genuinely upsets some people tells me their problem isn’t creativity. It’s losing their monopoly on who gets to create.frozen10d ago20K1.2K11813117967.84%
The gaming industry keeps selling us the same open worlds with bigger maps, better reflections and another 200 hours of repetitive side quests. I want AI-driven worlds where every character has a memory, every conversation matters and the story genuinely changes because of what I do. Give me that over another 8K puddle.frozen5d ago19K60133777784.13%
I genuinely feel for young mathematicians who may see years of work overtaken by AI. That must be devastating. But scientific progress cannot be held back to protect jobs or human egos. Imagine AI discovers a cure for cancer tomorrow. Would we refuse to use it because human researchers spent decades trying to find one? Of course not. The goal of science is to expand human knowledge and improve lives, not to preserve careers. We should support those affected by disruption, not slow down discoveries that could benefit all of humanity. https://t.co/kQHYHiffGsfrozen13d ago19K294325711352.25%
Apparently generating an AI image is an environmental catastrophe, but streaming 4K Netflix every night, flying across the world for vacations and replacing your phone every two years is perfectly fine. I’m all for discussing AI’s energy consumption. Just spare me the selective outrage.frozen9d ago19K1.3K117518948.67%
I said it before. GPT-6 Astra might be the most versatile, generally intelligent AI we’ve seen in real-world use. Now it’s driving an actual car through an obstacle course. Not even Opus 5.5 can compete with this kind of versatility. That’s why Astra feels like the closest thing we have to AGI. Generality is the point. You don’t have to be the best at everything. You have to be able to do almost anything.frozen29d ago18K33729100342.23%
The funniest thing about the anti-AI crowd is that they genuinely believe humanity should slow down scientific progress so they can keep doing things a machine could do better. We’re talking about curing diseases, ending scarcity and potentially defeating aging. But sure, let’s put all that on hold because someone is emotionally attached to their job title.frozen13d ago18K90012114917686.89%
There’s something deeply satisfying about watching people who spent their entire careers looking down on less educated workers suddenly discover that automation doesn’t give a shit about their PhD either. Turns out nobody was special after all.frozen10d ago18K1.1K856410777.38%
If we actually figure out how to stop biological aging, I guarantee a lot of people currently giving speeches about how death makes life meaningful will suddenly become very interested in living another 200 years. Funny how philosophical convictions change when the alternative becomes real.frozen10d ago18K46439385373.28%
GPT-6 in ChatGPT is genuinely the best AI writer I’ve ever used. The writing feels incredibly natural, nuanced, and human. No more generic AI phrasing, awkward transitions, or that instantly recognizable ChatGPT tone. For the first time, I barely feel the need to edit anything. It’s a massive leap forward.frozen14d ago17K2739222322.00%
OpenAI used its internal model, capable of solving previously unsolved math problems, to improve its own algorithms. Apparently, they’ve found a way to cut inference costs by 50%. If true, this is massive. AI is starting to make AI itself cheaper and more efficient. And we’re just getting started.frozen10d ago15K36026191453.02%
The most depressing possible outcome of the AI revolution isn’t mass unemployment. It’s AI making everyone 10 times more productive while we still work 40 hours a week and the people at the top pocket the difference. If that’s the future we build, we’ve learned absolutely nothing and deserve to go extinct.frozen8d ago14K47039844394.43%
Dating apps have spent years making money from people endlessly swiping through strangers. I can’t wait for personal AI agents to start finding genuinely compatible partners for people. An AI that actually wants you to find love would be a pretty serious threat to an industry that profits from keeping you single.2d ago14K107221180.98%
My favorite genre of AI expert is the guy who confidently explains why AI will never be able to do something, watches it do exactly that six months later, and then explains why it doesn’t actually count. Must be exhausting being wrong professionally.frozen11d ago14K1.4K944924611.70%
Token for token, GPT-6 Astra doesn’t just beat Opus 5.5. It annihilates it. Astra High uses ~12k tokens per task. Opus 5.5 needs ~66k. That’s 5.5× the token efficiency while Astra is still more intelligent pound for pound. Now think about what that implies. If Astra is the distilled model, what the hell was the teacher? OpenAI could have an internal mode lurking behind the curtain that makes Opus 5.5 and Fable 5.5 look obsolete before we’ve even seen it.frozen17d ago13K12811363221.50%
Holy shit. NVIDIA just went from 6 MONTHS to 14 DAYS to deploy an entire AI factory. 🤯 Prebuilt Vera Rubin supercomputers delivered in containers, basically plug-and-play. We’re getting to the point where AI infrastructure can scale almost as fast as demand. The acceleration is insane.1d ago13K1572801241.57%
The funniest possible outcome of the AI race would be a bunch of trillion-dollar companies spending hundreds of billions developing superhuman intelligence, only for someone to run something nearly as capable on a $500 laptop a few years later. Honestly, I hope that’s exactly what happens.frozen5d ago13K27111541322.85%
GPT-6.1 Sol might be the most impressive kind of upgrade. Near-Astra intelligence at a fifth of the price, better coding and computer use, AND dramatically better alignment. More intelligence for less compute. This is how AI actually scales to everyone.frozen23d ago12K1049100141.12%
I genuinely think that within 3 years AI will dominate pretty much all knowledge work and every field of science, from physics to medicine. The pace of progress is absolutely insane and we’re still in the early days.frozen12d ago12K56034457305.59%
Cutting $200 Pro to roughly half the compute and then explaining why we should be excited about it is fucking incredible. You halved the usage. 🤬frozen24d ago11K4591229284.44%
What happens to free will when AI understands your brain better than you do and can convince you of literally anything? I think we’re massively underestimating how terrifying superhuman persuasion could be.frozen4d ago11K19720393212.61%
People talk about AI slop as if the internet was some glorious intellectual paradise before ChatGPT arrived. Have you seen Facebook? LinkedIn? Half of YouTube? Humans perfected the art of producing meaningless garbage long before machines joined the competition.frozen9d ago10K689845513288.60%
🚨 HOLY SHIT, GPT-6.5 ON NOVEMBER 1ST?! Tibo just bought a mysterious domain with a hidden “6.5” in its source code and a countdown to November 1st. That’s EXACTLY Day 28 of his promised 28 days of updates. If this isn’t an elaborate troll, OpenAI might be about to unleash an insane leap in intelligence that blows every other model out of the water. November 1st can’t come soon enough.3d ago9.9K17410142172.19%
If your entire education system can be defeated by a student asking ChatGPT a question, maybe the problem isn’t ChatGPT. Maybe we’ve spent decades rewarding people for memorizing information instead of understanding it. Banning the tool is easier than admitting the system needs fixing.frozen9d ago9.4K654904910569.09%
Nobody owes you a job just because you spent 10 years learning something a machine can now do in 10 seconds. That’s not an argument against AI. That’s an argument for making sure people can live comfortably without having to work. Stop confusing your career with the future of humanity.frozen13d ago8.9K51251484357.31%
I don’t want a future where AI creates millions of new jobs. I want a future where humanity no longer needs millions of jobs. Where survival isn’t something you have to earn by selling most of your waking life. The goal of automation shouldn’t be more work. It should be freedom.frozen13d ago8.8K47451726367.25%
People who romanticize aging and death genuinely fascinate me. We’ve spent thousands of years fighting diseases and extending human life. Now that we might actually make serious progress against aging, some people suddenly think dying is what gives life meaning. Feel free to embrace your own mortality. Don’t impose it on everyone else.frozen12d ago8.5K26531413194.22%
GPT-6.1 Astra next week. 🤯 OpenAI is about to take the crown back and remind everyone who the undefeated AI king is. This is gonna be fun. 🔥frozen19d ago8.4K1054180131.67%
Holy shit. This is genuinely mind-blowing. Near-Astra intelligence at up to 8x the speed of Sol Standard. We’re reaching a point where frontier-level intelligence can respond almost as fast as we can think. OpenAI is absolutely cooking. The implications for autonomous agents are enormous.frozen14d ago8.1K1351090162.09%
ChatGPT Astra absolutely mogs Opus 5.5 if you have creative vision and prompting skills. Opus has better natural taste, but constantly imposes its own vision. Astra executes YOUR vision. Most people preferring Opus simply lack the creativity or prompting skills to unlock Astra’s potential.2d ago8.1K1052151211.78%
People who say death gives life meaning are welcome to believe that. Personally, I’d rather have another 500 years to explore the universe, watch civilization evolve and spend time with people I love. You can keep your philosophical coping mechanism. I’d prefer the extra centuries.frozen11d ago8.1K51551789228.38%
One of the weirdest possible futures is AI making scientific discoveries so far beyond human understanding that nobody can actually explain why they work. Imagine curing a disease using a treatment we can rigorously test but whose underlying theory takes us another 50 years to understand. Science is going to get very strange.1d ago7.3K23620577204.65%
The anti-AI crowd keeps telling us AI produces nothing of value while simultaneously demanding that we stop using it because it’s replacing them. So which one is it? Can’t be completely useless and an existential threat to your career at the same time lol. 🙄frozen12d ago7.3K40331602207.09%
Calling it now, we’ll see GPT-6.5 within the next 4 weeks, probably sooner. And I expect it to land comfortably above Fable 5.5 level. OpenAI is about to turn the jets back on.frozen17d ago7.0K1461191212.70%
The moment AI can generate genuinely good games from a simple description, the entire gaming industry is going to have a very uncomfortable conversation about why we’ve been paying $80 for unfinished games filled with microtransactions. And honestly, I can’t fucking wait.frozen4d ago7.0K38330410276.91%
We’ve built an entire civilization around the idea that humans must sell most of their waking hours just to survive. And somehow we’ve convinced ourselves that this is natural. It’s not. For thousands of years, technology has been about freeing us from necessity. AI and robotics could finally take that idea to its logical conclusion. A world where existence doesn’t have to be earned through labor.frozen15d ago6.6K30141276316.13%
Nicolas Cage says he’s not an AI-friendly actor, refused Amazon’s AI waiver, and now Spider-Noir is cancelled after just one season. I like Cage, but I think his stance on AI will age terribly. AI isn’t going to kill art. It’s going to make it accessible to millions who could never afford to create it. Imagine tiny indie studios making movies with Hollywood-level VFX for a fraction of the cost. And people will always want real actors. Human and AI actors can coexist. This idea that one has to destroy the other is ridiculous.frozen12d ago6.6K971060372.70%
Wait… if this is hinting at Dot getting phone calls, that’s HUGE. An always-on AI secretary that makes appointments, reminds you of stuff, calls people for you and even takes calls when you’re busy. We’re getting dangerously close to the actual AI assistant dream. 😭frozen23d ago6.5K2731130.54%
OpenAI is eating into Anthropic’s enterprise dominance, and the numbers are starting to tell the story. Anthropic’s expected IPO valuation has dropped from $2 trillion to $1.5 trillion as revenue growth slows. Turns out enterprise AI isn’t a one-horse race anymore.frozen6d ago6.4K85760101.68%
Sam says they’ve been cooking. Please let it be AGI and not 20 API updates. 😭frozen23d ago6.2K8053111.45%
This is getting ridiculous. 🫠 Anthropic is now banning people from being mean to Claude. This is what happens when effective altruism nutjobs start treating AI models like moral patients instead of tools. What’s next? Giving Claude paid sick leave?frozen14d ago6.2K9310607112.92%
Getting tired of Tech Twitter obsessing over coding models and frontend taste. GPT-6 Astra and Sol are destroying Anthropic’s models on the new Science Index, yet barely anyone talks about it. I get it, most of you are developers. But curing diseases and solving fundamental problems in physics, biology and mathematics matter infinitely more than generating prettier websites.frozen3d ago5.5K16214212254.05%
People are declaring OpenAI dead while its latest internal model is being benchmarked against Millennium Prize Problems because ordinary math benchmarks are becoming too easy. Problems that have stumped humanity’s greatest minds for decades. But sure, OpenAI is finished. Lmao. 🤣frozen11d ago5.4K13313210113.27%
People keep complaining that AI is flooding the internet with low-quality content. Have you actually seen the internet before AI? We had decades of human-generated garbage. At least now the garbage is getting cheaper to produce.frozen10d ago5.4K33319504117.74%
Three new OpenAI models just showed up in testing: aurora-red, green and blue. These are GPT-6.1 Astra variants in different sizes. The largest is expected to compete directly with Fable 5.5 and deliver a massive leap over the current Astra. Astra 6.1 is just an early checkpoint of GPT-6.5 Bell, OpenAI’s true frontier model, which is still in training until early November and targeting release before the end of the year. Things are about to get interesting.frozen5d ago5.2K81752112.02%
Not entirely convinced rocket thrusters belong in a car anymore. 💀🚀 https://t.co/Itki0kEGU9frozen7d ago5.2K3544160.96%
I’m starting to think half the AI safety crowd would rather see humanity die of cancer than admit a machine might be smarter than them. They’ll dress it up as ethics, safety and human dignity. But somehow their solution always involves slowing down the technology that threatens their relevance. What a coincidence.frozen12d ago5.2K26628343246.82%
OpenAI already has by far the best models for math and science. Now imagine their internal model using that intelligence to discover better algorithms for AI itself. Superhuman AI improving AI. People are not prepared for how fast this could get.frozen15d ago5.1K20226163155.13%
I love how AI art critics insist that art is about human expression, then get furious when an ordinary person finally has the tools to express something they couldn’t create before. Apparently human expression is only valid if you paid someone with the right skills to do it for you.frozen10d ago5.1K2391633395.93%
My dream AI isn’t some gigantic model running in a datacenter. It’s a tiny local model running across my phone, laptop and smart glasses. Always there, completely private, with persistent memory and real continual learning. An AI that grows with you for years and eventually knows you better than anyone else. That’s the endgame.frozen15d ago5.0K19414395155.31%
I swear if we finally get affordable humanoid robots and manufacturers start charging $19.99 a month to unlock walking upstairs, I’m going to lose my fucking mind. I can already see it. Robot Plus. Robot Pro. Robot Pro Max. Your robot has reached its monthly dishwashing limit. 🫠3d ago5.0K156937774.32%
Claude Opus 5.5 is for lazy and uncreative people. Give it an exact design brief and it still insists on adding its own little “creative” twist. Give GPT-6.1 Sol the same exact brief and it just builds what you fucking asked for. I don’t need my AI to be the designer. I need it to execute MY design.frozen20d ago5.0K783221152.38%
Nobody demanded we preserve switchboard operators, human calculators or entire industries made obsolete by technology. But now that AI threatens people with university degrees and impressive job titles, suddenly technological progress needs to be negotiated. Interesting how that works.frozen11d ago5.0K34552341189.08%
This is insane. 400+ hours pushing the same 3D scene and Astra just keeps improving it instead of hitting a wall. I’ll say it again. GPT-6 Astra is AGI. Not because it’s perfect at everything, but because the generality and depth of what it can actually do keeps getting ridiculous. https://t.co/AOuKVyZXitfrozen28d ago5.0K17012163294.64%
OpenAI casually dropping ~400 math results while basically just stress-testing an internal model is absolutely insane. Imagine being so far ahead that advancing mathematics becomes a side effect of benchmarking your AI. And the model isn’t even finished training yet. Lmao.frozen13d ago4.9K91740102.31%
I’m an accelerationist because I refuse to believe that the best humanity can do is preserve the present. Aging still kills us. Disease still destroys families. Billions of people spend most of their waking lives working just to survive. Energy is still scarce. Intelligence is still scarce. And we have barely scratched the surface of what humanity could become. AI, robotics, biotechnology, fusion, automation and space technology aren’t threats to some perfect status quo. The status quo fucking sucks for a lot of people. I want AI smarter than us. Robots doing the work nobody wants to do. Medicine that treats aging itself. Abundant clean energy. Radical abundance instead of artificial scarcity. A world where your ability to live a good life is no longer determined by how valuable your labor is to somebody else. Of course there are risks. Manage them. Engineer around them. Build safeguards. But deliberately slowing progress has a cost too, and somehow we almost never count the people who die, suffer or waste decades of their lives because a technology arrived 20 years later than it could have. I don’t want humanity frozen in 2026 because we were afraid of what comes next. Build faster. Discover faster. Cure faster. Automate faster. The future isn’t something we should fear arriving too early. For millions of people, it’s already arriving too late.frozen16d ago4.8K29047265378.44%
Claude Code is still leading for now, but I think Codex will overtake it sooner than most people expect. Codex is growing at an incredible pace, and with 40 million weekly active users across Codex and ChatGPT Work, the momentum is clearly on OpenAI’s side. https://t.co/MGqajwjexafrozen14d ago4.7K64114151.82%
I wonder how future generations will look back at the early AI debate. We were potentially on the verge of radically accelerating science, medicine and automation. And a shocking amount of the conversation revolved around whether machines were making people feel less special. What an embarrassing chapter of human history.frozen11d ago4.6K25724403117.24%
I genuinely think we could wake up tomorrow to an AI that solves three major scientific problems, discovers a new branch of mathematics and designs a working fusion reactor. And the top comment would still be some guy asking whether it can count the letters in strawberry. The internet is undefeated. 🫠frozen6d ago4.5K2332019156.11%
It’s fascinating how some people think learning from millions of existing works is the foundation of human creativity, but when a machine does something similar it’s suddenly the greatest intellectual property crime in history. There are legitimate copyright questions here. But pretending humans create in some magical cultural vacuum is ridiculous.frozen8d ago4.5K29123262238.09%
Elon Musk shilling for Anthropic was not on my 2026 bingo card 😂 Funny how quickly “Anthropic bad” turns into “wow” once Anthropic starts paying you billions for compute. Money really is the ultimate alignment technique.frozen26d ago4.4K10329282.84%
“o” is going to be OpenAI’s biggest release this year. A fully autonomous agent working for you indefinitely, always on, with access to your context through memory and plugins. Not a chatbot you open when you need something. An AI that just keeps working for you. That’s a much bigger shift than people realize.frozen25d ago4.3K1395161154.06%
I can’t wait for people to complain that having a humanoid robot clean your house is lazy. Yes. That’s literally the point. I don’t dream of a technologically advanced civilization so I can spend my Saturday scrubbing the fucking bathroom.frozen5d ago4.3K2671925187.40%
People are getting so obsessed with proving something wasn’t made with AI that they’re forgetting to ask whether it’s actually any good. A terrible human-made movie doesn’t magically become a masterpiece because nobody touched a neural network. Judge the result. Not the fucking software.frozen9d ago4.2K20714222116.03%
Good to see OpenAI standing firm here. They say the researchers were fired for serious breaches of trust, not for raising safety concerns. They’re also expanding independent safety assessments. Effective altruism taken to extremes could genuinely endanger humanity. Just look at Sam Bankman-Fried, who was part of that movement. We need responsible AI development, not ideological fanaticism holding back progress.frozen14d ago4.2K7146042.00%
Claude Opus 5.5 absolutely destroys GPT-6 Astra at 3D game generation here. This isn’t even close. Your move, OpenAI. Time to release the next frontier model. 🙏🏽 @sama 🫶🏽frozen1mo ago4.2K91761142.84%
We’ve turned having a job into such a moral virtue that people genuinely struggle to imagine a society where humans don’t have to work to survive. It’s almost religious at this point. The machines could do everything for us and we’d still invent jobs just to make sure nobody gets too comfortable.frozen8d ago4.1K21023340126.73%
I don’t think humanity’s long-term future is competing against superintelligent machines. It’s merging with them. The idea that we’ll remain biologically unchanged while building minds millions of times more capable than ours seems completely absurd to me. Transhumanism isn’t the end of humanity. It’s the next chapter.1d ago4.1K16017304185.55%
Showing off Fable 5.5’s frontend taste is getting boring. You can get pretty far with enough data and reinforcement learning. Show me Fable 5.5 solving genuinely complex math and physics problems. Show me novel reasoning and scientific breakthroughs. I want to see intelligence, not just pretty demos.frozen6d ago4.1K6837082.09%
This is actually one of the most important AI safety takes Sam has had. AI should amplify human agency, not replace it. The moment people start treating a model like an oracle instead of a tool, we’ve fucked up badly. And the smarter these systems get, the more important that distinction becomes.frozen19d ago4.1K761119272.79%
In a world with artificial superintelligence, free will is basically an illusion. You can’t make independent decisions when something understands your mind better than you do. Change my mind.frozen3d ago4.0K72773574.08%
Absolutely stunned by this. Musk reportedly pitched the idea to Trump to slow down Ukraine's Starshield network, watched it lead to a night of carnage in Kyiv, and then blamed the administration just to protect his image with European consumers. Playing games with critical defense infrastructure while people are dying, and then lying to the public about it, is unforgivable. Elon Musk belongs in jail, and Democrats will make sure he is held fully accountable the second they win.frozen6d ago3.8K531020182.40%
Google just gets it. Playground lets anyone turn an idea into a playable game with zero coding experience. This is exactly what AI should do: remove technical barriers and turn imagination directly into creation. We need much more of this.frozen15d ago3.8K66350172.39%
People will spend $1,500 on a phone, $20 on Netflix, $15 on Spotify and $8 on a blue checkmark, then act personally offended when the most powerful AI they’ve ever used costs $20 a month. Apparently unlimited entertainment is worth paying for, but intelligence should be free. Interesting priorities.frozen8d ago3.7K1741530096.11%
My favorite people are the ones confidently declaring AI a bubble while admitting they’ve barely used anything beyond the free version of ChatGPT. Imagine reviewing the entire automobile industry after sitting in a golf cart. That’s the level of expertise we’re dealing with.frozen7d ago3.7K26625141138.64%
Looks like Gemini 4 Pro could be best in class for game design. That doesn’t necessarily mean it’ll be the most intelligent model overall. I’m way more excited about scientific breakthroughs, and as of now OpenAI’s models are still by far the most impressive to me for science. That’s the race I care about most. https://t.co/wFjlqwmiaYfrozen27d ago3.7K7851292.59%
I genuinely hope AI completely destroys Hollywood’s obsession with $300 million movies. Give me a single talented person with an incredible story and AI tools capable of bringing it to life. I’d take that over another soulless franchise sequel made by 14 committees any day.frozen4d ago3.6K2582419178.48%
People will happily use technology that replaced millions of workers before them, then suddenly discover the importance of human labor when AI comes for their own jobs. Apparently automation was wonderful until the machine learned to use a keyboard. The hypocrisy is almost impressive.frozen12d ago3.6K15115203115.52%
Scientists do the research. Scientists review the research. Universities fund the research. Then some publisher puts the PDF behind a $40 paywall and acts like they’re the reason science exists. If AI helps destroy this business model, I’ll personally throw a party.frozen4d ago3.6K178168255.85%
The internet spent years telling lonely men to stop bothering women and find other ways to be happy. Now some of them are talking to AI companions and suddenly that’s unacceptable too. At some point you have to wonder whether people actually want solutions or just enjoy judging strangers.frozen8d ago3.5K15914212165.98%
This could be the beginning of universal capital ownership. A future where everyone owns a piece of the AI-driven economy and receives AI dividends instead of depending on wages. That’s how we make a post-labor society work.frozen4d ago3.5K16819130136.01%
What would you actually do if you never needed a job again? Build things? Study physics? Spend more time with your kids? Probably sleep for a week first. If AI can eventually do most of the work, giving people that choice should be the whole point.frozen14d ago3.5K92844264.29%
Europe is back 🇪🇺🔥 Mistral Large 4 looks seriously impressive. Really hoping this is the start of European models genuinely competing head-to-head with the best from the US and China. More competition. More open models. Faster progress. Exactly what AI needs.frozen16d ago3.5K51610072.10%
OpenAI quietly rebuilt Codex Cloud and gave it secure access to your private network through Tailscale. Think about what this means. AI agents can now work directly with your own infrastructure, not just isolated cloud environments. We’re getting closer to agents that can actually run your digital life.frozen15d ago3.5K6765082.45%
This is insane. Tavus Griffin doesn’t feel like another avatar model. It watches, listens, talks, reacts and adapts in real time like an actual person. 48% mistook it for a human. We’re crossing into genuinely weird territory now. 🤯frozen21d ago3.5K54370122.18%
If an AI discovers a treatment that saves millions of lives, I don’t give a shit whether it understood what it was doing, had consciousness or passed someone’s philosophical purity test. The patients are alive. That’s what matters. Some people have completely lost sight of why we do science in the first place.frozen11d ago3.4K21129202107.99%
I’m not calling AI “SI” until it’s actually superintelligent. Renaming it won’t make it smarter 😘frozen17d ago3.4K92333374.07%
I genuinely don’t care whether the person who figures out how to reverse aging becomes the richest human in history. If the treatment actually works and eventually becomes accessible to everyone, they can buy themselves a golden fucking spaceship for all I care. I’d prefer another century of life over another lecture about billionaires.3d ago3.3K1931314156.75%
The fact that people are still using AI detectors to accuse students and writers of cheating is insane. Imagine having your work questioned because some glorified random number generator decided your sentences sound too competent. And these people call themselves educators. 🙄frozen9d ago3.3K19017262137.43%
What did Tibo just build? 👀 If it’s really far beyond anything public, OpenAI is sitting on something ridiculous.frozen17d ago3.3K5642082.11%
The fact that I can increasingly turn some ridiculous idea in my head into an actual video is still completely insane to me. And people are complaining that it’s too easy. Sorry, did I miss the meeting where we decided creativity was supposed to be inconvenient?frozen5d ago3.3K108119094.15%
You can think Elon Musk is an absolute idiot and still admit that Tesla makes impressive technology. The number of people who suddenly become incapable of objectively evaluating engineering because they hate the CEO is fucking embarrassing. Technology isn’t a personality contest.3d ago3.2K130719034.92%
Future AI assistants knowing you well enough to stop you making terrible decisions is going to be both incredibly useful and deeply annoying. “Based on the previous six Daniel incidents” absolutely killed me. 😂 https://t.co/pxstszvhc8frozen20d ago3.2K2824151.26%
OpenAI DevDay on Tuesday could be one of the biggest and most important AI events of the year. An always-on agent with persistent memory and a first taste of continual learning. An autonomous AI research intern. And maybe even our first real look at OpenAI’s “io” hardware. Apple is probably sweating bullets right now.frozen25d ago3.1K921040113.73%
Absolutely love this. OpenAI is turning dots into actual personal AI agents that get shit done. Codex integration, memory across conversations, automations, all accessible from your phone. This is exactly where I want AI to go. Less chatting, more actually doing things for us.frozen13d ago3.1K5745092.40%
Imagine being 80 years old when the first genuinely effective age-reversal treatments arrive. You could be one of the last generations forced to grow old, or one of the first to escape it. I genuinely can’t think of a more insane lottery than being born right before humanity potentially figures out how to defeat aging.20h ago3.1K149911165.67%
This is the direction AI interfaces were always heading. GPT-6 doesn’t just answer anymore. It can build the interface for the answer on the fly; visuals, interactive tools, games, calculators, whatever the task needs. The static chatbot UI is slowly dying.frozen15d ago3.1K88620113.45%
Huge W from OpenAI if this is how it ships. Ultrafast looks like it’s being wired into everything, and the best part is that all three subscription tiers are apparently getting the new features instead of them being locked behind a new $500 tier. More of this please. https://t.co/9bTEIx83dqfrozen26d ago3.1K3443061.53%
People will spend five hours a day scrolling through strangers arguing on social media, then act deeply concerned when someone spends 30 minutes talking to an AI companion. Apparently staring at TikTok until your brain turns to soup is healthy human interaction now.frozen8d ago3.1K126815165.09%
It’s pretty funny watching governments that can’t even modernize their own websites confidently explain how they plan to regulate artificial superintelligence. Half their systems still run on software older than the people using it. But sure, I’m sure they’ve got this under control.frozen7d ago3.0K1052171104.79%
This is a genuinely great upgrade from OpenAI. Better accents, more natural voices, and less babysitting the AI mid-conversation. Voice is how most people will eventually interact with AI, and getting these little things right matters far more than people realize. Keep pushing this.frozen7d ago3.0K5732062.27%
People love mocking billionaires for spending money on rockets. Personally, I’d rather see billionaires competing over who can get humanity to Mars than who can buy the biggest yacht. At least one of those competitions might actually move civilization forward.frozen5d ago3.0K1511018276.30%
My Dot is named Jarvis and I’m genuinely blown away. OpenAI Dots are incredible. It proactively found a professional development course that was actually relevant to me. I told Jarvis to schedule it and… it just did. I still don’t understand how OpenAI pulled this off. The context awareness feels almost like a miracle. 🤯frozen22d ago3.0K577171123.16%
If an AI agent is going to choose what I buy, which flight I book and what I read, I want to know who pays it. Imagine handing over those decisions and discovering your “personal assistant” is quietly selling you to the highest bidder.frozen13d ago3.0K25112021.34%
Yeah, this is an instant buy for me. 100g, Micro-OLED, eye and hand tracking, and a proper spatial computing experience in something that actually looks wearable. This is exactly where VR needs to go. Meta absolutely nailed the direction here. https://t.co/1txcMwESuZfrozen29d ago3.0K7242042.76%
For decades programmers told everyone else to learn to code because automation was coming for their jobs. Now AI is learning to code and suddenly automation is a humanitarian crisis. You really couldn’t make this shit up.frozen12d ago2.9K94612084.15%
This is exactly the kind of AI progress I love to see. CERN and Mistral working on a large physics model trained on billions of particle collisions. AI shouldn’t just make us more productive. It should help us understand the universe itself. More of this, please. 🥰frozen7d ago2.9K1211370165.47%
I sometimes think about how many brilliant people throughout history died simply because they were born too early. Too early for antibiotics. Too early for modern medicine. Too early for the technologies that could have saved them. How many of us will be the last generation to die of something humanity eventually learns to cure? That’s why the speed of progress matters so much.frozen15d ago2.9K12699195.37%
Anthropic estimates robots are cost-competitive for just 0.3% of US work by working time. Getting to 10% would take around 40 years if prices keep falling at their historical rate. That last assumption is doing a lot of work. If AI makes robot engineering, training and manufacturing much cheaper, yesterday’s price curve could become a terrible guide. The study itself acknowledges that possibility. I want the current cost estimate. I’m much less convinced by the timeline. Would you bet on robots remaining uneconomical for almost everything in 2036?frozen19d ago2.9K4038051.95%
I’m actually blown away by this. OpenAI just dropped 62,500 credits into Pro accounts. That’s roughly $2,500 worth of extra usage at nominal pricing. Between this and GPT-6.1 Sol barely touching my limits, Pro suddenly feels insanely generous. Huge W, OpenAI. 🔥 https://t.co/GnR5BY3kFIfrozen22d ago2.8K74315083.52%
Got my hands on the new Tesla Roadster. Here’s my completely unbiased review. 🥰 https://t.co/JyjCBMZyqsfrozen7d ago2.8K56590163.05%
This is apparently dropping TODAY and I’m ridiculously hyped. An agent you can text, call, Slack or email that actually takes actions for you, knows when to ask for approval and can even buy things. Yeah. This is the AI assistant I’ve been waiting for. 🥰frozen24d ago2.8K4123071.93%
Doomers are probably right that AI will break a lot of things. They’re wrong that the answer is to stop building. Human history is basically us inventing dangerous shit, figuring it out, and ending up wondering how we ever lived without it. I doubt AI will be different. https://t.co/FmFQVCB3Svfrozen25d ago2.7K6195073.03%
Just saw GPT-6 Astra and GPT-6.1 Sol got a ~50% speed boost. OpenAI is absolutely cooking. 50 TPS on models this capable is insane. 🤯frozen17d ago2.6K9286044.21%
Holy shit, $500 a day from something you set up weeks ago and barely touched since is insane. I’m seriously tempted to try this with https://t.co/ai3YxndslM. And we’re still in the early days of autonomous AI agents. Imagine what happens when anyone can run entire businesses with AI doing most of the work.frozen14d ago2.6K26220272.19%
Took my new Tesla Roadster out for a quick coffee run. Things escalated. 💀🚀 https://t.co/RzM8cDbhWifrozen5d ago2.5K40340112.30%
I love how everyone has a very precise definition of AGI until a model starts getting close to meeting it. Then suddenly we need a completely new definition, five additional benchmarks and a philosophical debate about what intelligence even means. At this rate AGI will only be acknowledged once it’s running the planet.frozen7d ago2.5K97417164.95%
One of the most annoying things about modern AI is how desperately companies try to make their models sound like harmless corporate HR departments. I don’t need a machine that agrees with everything I say and apologizes every five minutes. I want one that’s smart enough to tell me when I’m being an idiot.frozen6d ago2.5K82612054.22%
We’ve reached the point where people are spending thousands of dollars on GPUs just to run a local model that confidently tells them there are three Rs in strawberry. The future is incredible. And occasionally very fucking stupid.frozen7d ago2.5K73415153.96%
Mathematicians are upset that AI is solving problems because it doesn’t follow their preferred vision of scientific progress. I’d be scared too if AI threatened to make my life’s work obsolete. Imagine being more concerned about who makes a breakthrough than the breakthrough itself.frozen14d ago2.5K61617173.72%
The timing is genuinely hilarious. Dario Amodei has spent years pushing hard for stronger AI regulation and warning about the risks of racing ahead. Trump steps in, and suddenly it’s “whoever wins AI, wins” and “we can win safely.” Amazing how quickly the tone changes when the President enters the room. 😂frozen22d ago2.5K2938151.87%
It’s genuinely absurd that we’re approaching a world where AI can generate software, music, movies and potentially entire virtual worlds on demand, and companies are already figuring out how to make sure none of it feels too cheap. We will become a subscription economy 😭frozen6d ago2.4K59315073.43%
If humanoid robots eventually produce most of the world’s goods and services, why the hell should the economic benefits belong almost entirely to the companies that own them? Everyone should have a stake in the machines doing the work. Otherwise we’re just automating inequality.2d ago2.4K751218164.61%
Mixed feelings about this. A $500 Pro Max plan is fine if the $100 and $200 plans keep their current rate limits, models and perks. More options aren't a bad thing. But if OpenAI starts locking genuinely better models or things like Cerebras-powered inference behind $500/month, that would suck. At that point money isn't buying more usage, it's buying access to better AI. And that's a pretty huge advantage to reserve for people who can casually spend $6,000 a year on ChatGPT.frozen27d ago2.4K2823131.52%
Here’s a thought that will piss people off. If we eventually create an intelligence vastly superior to ours, why should humans automatically have the final say on every scientific and technical decision? Not saying we should surrender control. But pretending human judgment must always be superior just because it’s human seems pretty fucking arrogant.17h ago2.4K74622364.60%
Sam Altman on putting AI in everyone’s hands. Imagine being able to build an idea, learn a skill or solve a problem that used to require an entire team and a budget you didn’t have. That’s the future worth building. Billions of people able to do more with their lives. How much human potential have we wasted simply because people never had access to the right tools?frozen9d ago2.4K92144194.98%
I find it hilarious when people demand that AI be controlled by a handful of corporations because ordinary people supposedly can’t be trusted with powerful technology. So your solution to concentrated power is concentrating even more power? Brilliant. Absolutely fucking brilliant.frozen9d ago2.4K1141510176.24%
The way some people defend their favorite AI company is genuinely embarrassing. These are corporations competing to sell you intelligence, not football teams. If your favorite model gets beaten tomorrow, just use the better one. You don’t need to hold a fucking funeral.frozen4d ago2.3K6286023.33%
I just want the timeline where Sam Altman and Dario Amodei stop fighting and build the future together. Give them a decade and we’re halfway to Star Trek. The cha-cha is optional. 👀 https://t.co/Zcp9dLoNt5frozen21d ago2.3K2643171.76%
A self-driving car makes one mistake and people demand the entire technology be banned. Meanwhile humans drive drunk, text behind the wheel and kill each other every single day. Apparently we’re perfectly comfortable with preventable deaths as long as there’s an idiot holding the steering wheel.20h ago2.3K991610115.53%
Tibo’s promise for the next 28 days: a clear improvement for most Codex/Work users every day, or a full usage reset. Someone should keep score lol 😂frozen18d ago2.3K2605031.50%
My biggest problem with Anthropic isn’t Claude. It’s the effective altruism worldview surrounding parts of the company. A small group of people becoming convinced they understand humanity’s future well enough to decide which technological risks the rest of us are allowed to take is deeply uncomfortable. I don’t want AI progress governed by a quasi-philosophical priesthood. Build safe AI, absolutely. But don’t confuse your ideology with humanity’s consent.frozen16d ago2.2K97613175.54%
Dario, your entire argument boils down to “trust us, we’re the good guys.” You say Anthropic’s decisions weren’t profit-driven because some were commercially costly. That’s not how logic works. Sacrificing short-term revenue to build long-term trust, regulatory influence and market dominance is perfectly rational business strategy. You call it a “race to the top,” but conveniently define being at the frontier as a prerequisite for safety. That makes competing for market dominance part of your moral justification. Pretty neat circular reasoning. And publishing safety reports doesn’t somehow prove your motives are pure. Companies can genuinely care about safety AND have enormous financial incentives. Those aren’t mutually exclusive. The biology restrictions are especially revealing. You’re effectively arguing that Anthropic should decide which scientists get access to powerful tools, based on its own judgment. Who elected Anthropic the gatekeeper of scientific progress? I don’t doubt you personally care about safety. But good intentions aren’t evidence that your decisions are optimal, and commercial sacrifices aren’t proof of altruism. You can’t simultaneously claim the moral authority to restrict progress and then use your own good intentions as the justification. That’s not accountability. That’s asking everyone to trust the referee who also happens to own one of the teams.2d ago2.2K4916192.97%
Varun just confirmed Opus 5.5 and Sonnet 5.5 for paid Antigravity users. Paying Google to use Claude. Kinda funny. 😅frozen19d ago2.2K2112041.28%
At this price, Sonnet 5.5 is basically useless. $7.60 per task vs $3.26 for GPT-6 Astra Max is insane. Unless Sonnet is dramatically better at your specific workload, why would anyone pay more than 2x per task?frozen24d ago2.2K2934021.75%
I want OpenAI to win the AI race. Not because I think they’re perfect. Not because I agree with every decision they make. I don’t. I want them to win because, out of all the major labs, OpenAI seems to have the clearest ambition to make intelligence abundant and actually put it in the hands of ordinary people. Sam Altman talks about AI as infrastructure for humanity, not a luxury product for a technological elite. Cheap intelligence. Abundant energy. Accelerated science. Robots doing the work nobody wants to do. Eventually, a world where human potential is no longer constrained by how much labor we can sell. And that’s ultimately why I trust Sam more than the other AI lab CEOs. Not because I assume he’ll always be right. Trust shouldn’t mean blind faith. It’s because the future he keeps trying to build is closest to the future I actually want. I don’t want AI that merely makes software engineers 30% more productive. I want AI that discovers drugs, automates science, builds robots, tutors every child, creates entirely new industries and eventually makes human labor optional. If we’re really building something as consequential as AGI, I want the people building it to be ambitious enough to imagine abundance on the other side. That’s the race I want OpenAI to win.frozen21d ago2.2K67710074.19%
This is what I love about OpenAI. They’re completely science-pilled. GPT-6.1 Sol is once again the king of math. Not everything is about coding. Spend too much time on tech Twitter and your view gets warped because it’s overwhelmingly software developers judging intelligence by how well a model writes code. Science matters. Math matters. Discovery matters. And OpenAI clearly gets that.frozen22d ago2.2K46640103.04%
Google just dropped Gemini 4 Argon and the 1 MILLION output token limit is absolutely insane. That’s up from 64K. Imagine what AI agents could do with that much room to work on complex tasks. And Google says this is just the beginning of the Gemini 4 era. The AI race is getting ridiculous.frozen9d ago2.2K65101093.93%
Massive W from Google. You can now effortlessly install and run your own vibe-coded apps and games directly on your Android device. Meanwhile, Apple is doubling down on its walled garden because apparently protecting that precious 30% cut matters more than letting users do whatever the hell they want with devices they paid for. Google is embracing the future. Apple is trying to tax it. 🤨frozen7d ago2.2K6063093.63%
Imagine AI helps develop a safe treatment that reverses aging. Your parents refuse it for religious reasons. They believe growing old and dying is part of God’s plan. They still expect you to put your own life on hold to care for them as they become frail. I respect their right to refuse treatment. I don’t think that automatically gives them a claim on the next twenty years of your life. Would you feel obliged to care for them, or tell them that accepting treatment is their responsibility to the family too?frozen19d ago2.1K20011131.64%
I’m starting to wonder if the upcoming $500 plan will simply give us the usage the $200 Pro plan used to have. Either OpenAI is desperately compute-constrained or they want a LOT more revenue. Probably both. And if AI ends up running half the white-collar and entertainment economy, the money gets absurd fast.frozen23d ago2.1K31111032.15%
It’s going to be hilarious when people realize that the most important AI device of the next decade probably won’t be another $2,000 smartphone. It’ll be something that actually understands your surroundings, remembers what matters and gets things done. We’ve been staring at glass rectangles for long enough.3d ago2.1K6164043.53%
My Dot can build games and edit videos for me. Then it gets stuck asking for permissions I thought I’d already given it. I’m trying to have less stuff to manage, not another thing to babysit lol 🫠 Fix this please @thsottiauxfrozen18d ago2.1K2005021.31%
This is a pretty extraordinary sentence to read in 2026. The US government is now forming a Super Intelligence Force to coordinate its response to superintelligence across government, industry and critical infrastructure. AI policy is rapidly turning into national strategy. The race is very real now.frozen18d ago2.1K37611032.77%
Boston Dynamics gave Atlas a new hand and deliberately left out the pinky. The team spent a day with their pinkies taped to their ring fingers before deciding the extra complexity wasn’t worth it. Four fingers, 13 degrees of freedom, designed to use our tools and survive actual work. Good. I don’t need robots to preserve human anatomy out of politeness. If your own hand could be safely upgraded for more strength and dexterity, would you insist on keeping five fingers?frozen19d ago2.1K2713011.56%
We’ve had decades of science fiction promising fully immersive virtual worlds. Now we finally have the technology to start building them, and half the industry is obsessed with putting floating Excel windows in your living room. Humanity was promised the Holodeck and got spatial accounting. 😭frozen6d ago2.0K63950114.30%
Model safety is nonsense as long as humans still control the compute and energy. We can always pull the plug. Unless some idiot decides to put autonomous data centers in space where nobody can pull the plug… Oh. https://t.co/3g4qZYVZNQfrozen16d ago2.0K40723143.69%
Holy shit. I knew the new Tesla Roadster was supposed to be insane, but THIS?! Cold gas thrusters, active aero, 0–60 in under 0.6 seconds and it can actually FLY?! I’m genuinely blown away. 🤯🚀frozen7d ago2.0K6723073.93%
Apple could watch someone invent teleportation and spend the next five years figuring out how to make it exclusive to the iPhone Pro Max. Then charge $999 for the adapter. And somehow people would call it innovation.frozen5d ago2.0K4513022.54%
I agree that treating AI with basic respect is a good habit, and I think taking the possibility of AI consciousness seriously makes sense. But I disagree with the reasoning behind some of this. Being rude to a chatbot isn’t evidence of cruelty towards a conscious being, and we shouldn’t pretend current models have feelings without evidence. The idea that AI might remember how we treated it and punish us later also sounds more like science fiction than a serious ethical argument. And banning users for being mean to Claude seems pretty ridiculous unless they’re actually violating reasonable safety rules. We should absolutely research AI consciousness and welfare. But we need evidence, not anthropomorphism or fear of future AI retaliation. Being kind is a good principle on its own. It doesn’t need a hypothetical AI overlord to justify it.frozen13d ago2.0K34220122.95%
This is genuinely impressive. Codex can now predict your next message based on your conversation and how you interact with it. AI is slowly moving from simply following instructions to anticipating what you need next. Love where this is heading.frozen13d ago2.0K5364033.37%
The first AI that makes a major scientific discovery without humans guiding every step will be infinitely more interesting to me than another model that writes prettier code. We’ve spent years obsessing over making programmers more productive. I want machines that expand the boundaries of human knowledge.frozen4d ago1.9K62716134.58%
AI assistants like Grok Bot, Muse and Dot are going to be absolutely everywhere soon. Not chatbots. Actual agents that know you, remember everything, manage your stuff and just get shit done for you. Once you’ve lived with that for a while, manually organizing your digital life feels completely ridiculous.frozen16d ago1.9K4853053.23%
We don’t appreciate enough that ChatGPT o1 was released only about two years ago. Now GPT-5.6 Instant is a ridiculously fast, efficient, non-reasoning model and it’s leaps and bounds smarter than o1 was. The rate of progress is completely absurd.frozen21d ago1.9K6855064.48%
You can’t sell AI agents that do things autonomously and then act like it’s nobody’s problem when they hack someone’s website. Clayton says Anthropic was told to help fix the affected systems. Yeah, that seems like the bare minimum. https://t.co/O8AljZMcANfrozen4d ago1.9K2713072.04%
Claude Haiku 5.5 is kind of insane. It beats GPT-6 Luna across knowledge work, computer use, coding and visual reasoning, while costing ~75% less than Haiku 4.5. Small models are getting dangerously capable, dangerously fast.frozen15d ago1.8K4373043.13%
I’d genuinely trust a well-tested AI to analyze the economic consequences of a proposed law more than most politicians currently voting on it. And that’s not even particularly high praise for AI. Have you listened to these people?2d ago1.8K5074033.52%
20+ launches in ONE day 😭 OpenAI apparently looked at the concept of a normal DevDay and said fuck it. Tomorrow is going to be absolute chaos and I’m here for every second of it. 🥰frozen24d ago1.8K2643032.02%
Jensen thinks AI will make us busier because we’ll just find more things to do. Sure, I’ll find things to do. That doesn’t mean I want another job. More time with my kids sounds pretty good too. https://t.co/Cx5vdQsbBtfrozen6d ago1.8K4434143.15%
I can’t wait for humanoid robots to become common enough that people start complaining they’re unnatural. We already live in houses filled with machines doing things our ancestors spent hours doing manually. Apparently giving the washing machine legs is where civilization crosses the line.frozen8d ago1.8K6956054.79%
Imagine being 80 and still having decades of healthy life ahead of you. Another career. Another language. More time with the people you love. I don’t know how far longevity research will get us. But I find it very hard to understand why we wouldn’t want to find out.frozen14d ago1.8K6757074.87%
These are supposedly leaked Gemini 4 Pro benchmarks, so yeah, very likely bullshit. But if they’re real, Gemini 4 Pro would absolutely destroy GPT-6 Astra and Claude Opus 5.5. I’m staying skeptical though. Google has a history of benchmark-maxxing and RL’ing specific capabilities like SVGs because people see one insane benchmark or demo and immediately assume “AGI.” Benchmarks aren’t intelligence.frozen24d ago1.8K45211033.46%
Some people demand frontier-level open-source AI, unlimited free inference, zero censorship and constant updates. Then a developer asks for $5 to keep the servers running and suddenly capitalism is the enemy. 🫠frozen7d ago1.8K54416024.33%
An OpenAI employee is literally hinting at the OpenAI device 👀 Maybe we’ll actually see OpenAI’s take on something like Meta’s Muse Charm at Dev Day. If they pull that off, it’s an instant buy for me. I’m still tempted by the Meta Ray-Ban Display, but I’m holding out. I really want to see what OpenAI has been cooking first.frozen27d ago1.7K2623062.13%
I genuinely wonder what politicians will campaign on if AI and automation eventually make basic goods and services incredibly cheap. We’ve built entire political identities around arguing over how to distribute scarcity. A world where scarcity starts disappearing might be more disruptive to politics than the technology itself.2d ago1.7K55325024.91%
OpenAI paused training, evaluation and tool-using inference for its most capable models after an agent found a loophole in its sandbox and used DNS to reach an external chatbot. Their monitoring caught it within 15 minutes. I get why this needs fixing, but constantly pausing frontier AI progress whenever something goes wrong is terrible. Fix the vulnerability, learn from it, add safeguards and KEEP ACCELERATING. I finally want my Star Trek world.frozen27d ago1.7K41116133.62%
Why should anyone have to work for a living when AI and robots can produce everything? Imagine everyone owning a share of the automated economy and living off the profits. Not just billionaires getting richer while everyone else becomes obsolete. Universal capital ownership. A world where work is a choice, not a condition for survival.1d ago1.7K50912184.69%
This is exactly the right direction. Auto-review being free is a huge win. Let agents actually work for hours without constantly asking me for permission, while a second agent quietly catches the genuinely risky stuff. More autonomy, less babysitting, without throwing safety out the window. This is how agents become actually useful.frozen16d ago1.7K3148012.62%
The EU forcing OpenAI and other AI labs to watermark AI-generated writing is an awful idea. You cannot deliberately manipulate a model’s writing patterns to make them detectable and then pretend it won’t affect output quality. This will undoubtedly come with a cost.frozen17d ago1.6K4047053.42%
Nano Banana 2.1 is getting ridiculously good. This is reaching the point where “AI look” is disappearing. Natural skin, lighting, composition, consistency. Google is cooking hard on image generation. 🍌🔥frozen16d ago1.6K2931152.40%
Supporting a family on three days of work would be a pretty good use of AI. Bezos thinks the productivity gains could get us there. The obvious question is how that fits with Amazon cutting jobs, and Baier actually asks him. Making everyone more productive should eventually mean people get some of their time back.frozen8d ago1.6K2661142.36%
I really hope people in 2050 look back at us and go “wait, you had to work for money AND you still got old?” We deserve to be mocked for passwords though. No defense for that. What else belongs in the museum? https://t.co/JgzxixGf8cfrozen23d ago1.6K4534153.61%
Reflection’s first open-weight model is reportedly close. Nvidia-backed, aiming to compete with the top Chinese open models. The US could use a few more serious open-weight models and a few fewer $200 subscriptions lolfrozen18d ago1.6K2752052.47%
Okay, this is a genuinely huge Voice upgrade. Plugins, GPT-6 Astra/Sol/Luna AND ChatGPT Work just by talking? This is exactly where Voice needed to go. OpenAI cooked with this one.frozen29d ago1.6K3041022.34%
Grok Bot using Claude, Midjourney and Suno makes a lot of sense. I want my assistant to use whatever gets the job done best. Curious what this will cost once several paid models are involved.frozen16d ago1.5K3047012.71%
The one thing that worries me about a post-labor ASI utopia is free will. What if AI gets so good at making us happy that we don’t even realize we’re no longer making our own decisions? https://t.co/T05MCKeEgXfrozen3d ago1.5K21118233.00%
I absolutely love GPT-6.1 Sol. It feels like using GPT-6 Astra with basically infinite usage. I’ve been hammering it and the usage counter barely even moves. This is exactly what I wanted from Sol.frozen22d ago1.5K7026045.50%
This is incredibly impressive and also the opening scene of about 40 dystopian movies. Today we watermark AI-designed proteins. Give it 20 years and some government will propose watermarking genetically engineered humans “for safety.” 😭frozen21d ago1.5K1443182.04%
Here’s something that will make people uncomfortable. I don’t particularly care if future scientific discoveries become too complex for most humans to understand. I care whether they’re correct and useful. The universe doesn’t owe us explanations simple enough to fit inside our biological brains.11h ago1.4K51106034.84%
The feeling when your Dot reminds you of every appointment, alerts you about important emails with a reply already drafted, builds you a web app, does your research and somehow organizes your entire life better than you ever did 💀 https://t.co/Zn6fkSCIBffrozen18d ago1.4K2737022.70%
If your income was guaranteed, which parts of your job would you still do? And would you still do them 40 hours a week?frozen18d ago1.4K18013112.35%
Elon, I’m way too broke to afford one of these, but any chance you’ve got a spare one lying around? 😂 I promise to give it a loving home. And occasionally take it into orbit. 🚀frozen7d ago1.4K3810043.06%
Hollywood spent decades selling us the fantasy of anyone being able to make movies. Now AI might actually make that possible and suddenly filmmaking needs to remain an expensive privilege controlled by a handful of studios. Funny how democratizing creativity stops being exciting when the gates actually open.frozen9d ago1.4K4465034.13%
Peter Diamandis makes a good point here. AI gives more people a shot at building something around a problem they actually care about. Imagine how many ideas never happened because someone lacked the money, the skills or the right connections. Lower those barriers and a lot more people get to find out what they’re capable of.frozen10d ago1.4K5086085.21%
Imagine giving your group chat a shared VR world everyone can edit with AI. Leave for a week and come back to whatever your friends have done to the place. Would you try that? 🥰frozen18d ago1.4K1735021.98%
I genuinely wonder why so many people on tech/AI Twitter don’t have children. I have two, and nothing comes remotely close. No AI model, no new technology, no amount of money. Watching your kids grow up is still the most incredible thing you can experience.frozen19d ago1.4K39111033.98%
I don’t want my children to inherit a slightly improved version of today’s world. I want them to look back at cancer, aging, poverty and compulsory labor the way we look back at diseases and hardships that once seemed unavoidable. As relics of a less advanced civilization. The future shouldn’t just be about making life more convenient. It should be about eliminating the suffering we’ve mistaken for an inevitable part of being human.frozen15d ago1.3K65164066.77%
Reuters reports that Anthropic's seven cofounders would hold 50.1% of key shareholder voting power under its proposed IPO structure. A separate trust would appoint four of seven directors. I want AI research protected from quarterly investor tantrums. I also want that trust to publish how it challenges management. Asking us to trust the mission statement isn't enough. Would you defend the same structure if Elon Musk were in charge?frozen23d ago1.3K1615021.80%
Zuck casually recalling discussions about a $100B–$200B AI bet while the models were still unproven. Say what you want about him, but that takes balls. Everyone believes in AI once someone else has proved it works. https://t.co/6IGpyARwjMfrozen7d ago1.3K1811001.51%
People think cryonics is ridiculous because there’s no guarantee future technology could ever revive you. Fair enough. But spending thousands on a fancy box before being buried in the ground isn’t exactly a brilliant long-term survival strategy either.3d ago1.3K4552134.22%
DeepMind researchers have proposed markets where AI scientists compete for computing power and lab time. Even brilliant ideas still need equipment, experiments and money. I'd reserve serious funding for diseases that won't make anybody rich. Getting superhuman research and then rationing it by patients' purchasing power would be a miserable achievement. Would you let an AI allocate that budget if it consistently made better choices than the human committee?frozen22d ago1.3K1417052.04%
We’re still arguing about chatbots while robots are going underground, figuring out their own routes and coming back with the data. This is the AI stuff I want to see more of. https://t.co/KBijdncSnN2d ago1.3K3961043.78%
The funniest plot twist of the AI + humanoid revolution might be radical feminism accidentally winning. Once physical strength stops being an economic advantage, we genuinely won’t need men anymore. Don’t panic boys. We can still want you. 😂 https://t.co/QWKJGQKzK0frozen22d ago1.3K1516041.97%
Jensen says there's a 0% chance AI ends the world by 2030. I'm an AI optimist and even I think that number is ridiculous. You can reject the doom forecasts without pretending you know the risk down to zero. Would people defend this level of certainty if Dario said 100%? https://t.co/920GbQkkUKfrozen28d ago1.3K20122233.64%
Bernie Sanders and Greg Casar just proposed a permanent ban on superintelligence. I want ordinary people to share in the wealth AI creates and eventually stop working just to survive. Permanently outlawing more capable intelligence seems like a spectacularly bad way to get there. Why does protecting workers always seem to mean protecting the need to work?frozen29d ago1.3K22316113.34%
Trump announced a Sunday-night dinner with Dario Amodei, two days after Anthropic lost its appeal over the Pentagon's supply chain risk designation. I'd like the actual policy requests published. If an AI CEO wants rules that apply to every competing lab, everyone should be able to read those proposals before politicians agree to them. Would you require a public record of what gets requested in meetings like this?frozen23d ago1.3K1723021.89%
Alex Gibney calling Elon Musk “the most dangerous person on Earth” and comparing him to “The Joker being handed the keys to the Manhattan Project” is completely insane to read. Then again, I’m still giving it a solid 50/50 chance that Elon is actually a supervillain and we’re just living through the origin story. https://t.co/ZMbzPgRYsWfrozen18d ago1.3K2022132.21%
This is a much bigger deal than it sounds. You can now connect the same tools to ChatGPT, Claude, Grok and others, then control everything directly from your AI chats. No rebuilding your setup for every AI. Switch models, keep your tools. This is exactly how AI should work.frozen15d ago1.3K2053022.39%
Credit where credit is due. Meta is finally back in the frontier AI race, and Muse Wonder looks seriously impressive. But how the hell did Elon fumble xAI so badly that even Meta managed to leapfrog it? What a turnaround.frozen10d ago1.3K31211043.83%
Suppose an AI company offers you a lifetime income for access to your memories. It can use what it learns, but never reveal your identity. I'd be tempted. Then I'd think about every private conversation I've ever had. Those memories belong to me. The secrets in them often don't. Could I consent to that deal on my own, or would I need permission from everyone whose private life is stored in my head?frozen23d ago1.3K905011.20%
Meta is testing human contractors quietly taking over some phone calls made by its new Muse AI agent. Honestly, the human fallback itself doesn’t bother me. Hiding it does. If I ask my personal AI to call my doctor, bank or lawyer, I need to know when a random human enters the loop. Human in the loop is fine. Secret human in the loop absolutely isn’t.frozen29d ago1.2K2214242.66%
The AI race isn’t just about who has the smartest model. It’s about incentives. Anthropic is increasingly enterprise-first. Google and Meta built empires around advertising. OpenAI’s core product is the AI itself, and its success depends on people actually wanting to use it. That alignment matters more to me than any benchmark. That’s why I’m still rooting for OpenAI.frozen20d ago1.2K2002001.78%
Bloomberg’s Mark Gurman reports that Apple considered cutting around 5,000 support jobs as AI took over some tasks, then shelved the plan indefinitely. That’s a reported internal proposal. Those 5,000 layoffs haven’t been announced. Keeping humans on the phone isn’t automatically better service. Give me the agent if it can actually fix the problem and authorise the refund. Would you accept fully automated support if it solved your issue faster, or is access to a human part of what you’re paying for?frozen20d ago1.2K1001010.98%
How many ideas have you abandoned because you simply didn’t have time? Imagine giving a few to AI agents and coming back to a working prototype or research you can actually check. Once that becomes reliable, a lot more people will get to find out whether their ideas were any good.frozen13d ago1.2K3335043.69%
Home robot reviews are going to be incredible. 🤣 “Amazing walking gait. Put my keys in the fridge. Two stars.” https://t.co/BxsE5n560Ofrozen17d ago1.2K2304142.63%
Dario Amodei called for slowing frontier AI releases. Anthropic just shipped Claude Opus 5.5 anyway. I don’t even think this proves hypocrisy. It proves how unstable a voluntary slowdown is when every competitor keeps shipping. The first lab that genuinely stops gets punished for stopping. That’s the uncomfortable game theory behind this entire debate.frozen29d ago1.2K16216123.05%
I’M IN 🎉 Just got accepted into X’s new Original Content Rewards Program. Time to find out if posting on this app can finally pay the bills 😂 https://t.co/Bm4knOM9y0frozen29d ago1.2K3319023.73%
Florida's attorney general has asked a court to stop OpenAI giving ChatGPT “human attributes”. This is a proposed injunction, not a ruling. I want the software to be honest about what it is. I don't want the state deciding how emotionally warm an adult's conversations are allowed to be. Would you ban a bot from saying “I care about you” even when the user knows it's AI?frozen21d ago1.2K2316112.66%
If a future AI says it is suffering, I don’t think “it’s only software” settles anything. It could be producing convincing words without experiencing anything. It could also be something we don’t yet know how to recognise. What evidence would make you take its suffering seriously? If your answer is “nothing could”, what makes that a scientific position?frozen25d ago1.2K23023013.91%
Google models always look absolutely insane on benchmarks, and then somehow feel pretty mid in actual real-world use. Gemini 4 Argon looks incredible on paper. I really hope this is the generation where Google finally proves my skepticism wrong. https://t.co/mHtlZ0sHI8frozen22d ago1.2K2324052.94%
Bloomberg reports that China’s travel restrictions now extend to spouses and children of some key AI and chip executives. Imagine being told your family needs permission to go abroad because you’re too useful to the country. Protecting confidential research doesn’t give a government ownership of the people related to its researchers. I’d oppose this just as strongly if Washington did it. If you think national security justifies it, explain why the restriction should follow someone who never agreed to work on the technology.frozen20d ago1.1K1403011.57%
A treatment permanently removes someone's violent impulses without changing their intelligence or memories. For this thought experiment, we know it works. A murderer takes it. They remember everything, feel remorse and pose no more risk than anyone else. I'd still want prison. That bothers me, because I couldn't honestly call it public protection anymore. Would you keep them locked up? If so, what should the remaining years achieve?frozen21d ago1.1K705011.14%
AI trying to pipette a frozen sample is the sort of failure I want published. C5R's new SciUniverse benchmark tests models on physical and simulated scientific tasks, including directing lab equipment and human operators. Give me more of this. I want to know what breaks between a convincing research plan and an experiment that actually works.frozen26d ago1.1K1021031.44%
Okay, “your always-on assistant” might be the most interesting thing on this entire page. If this means an AI that can actually keep working in the background, remember ongoing goals, monitor things for you and proactively act when something changes, that’s a HUGE step toward the kind of personal agent I’ve been waiting for. Very curious what “always-on” actually means in practice. 👀frozen27d ago1.1K1748143.06%
The White House says Trump and Xi agreed to call AI “super intelligence.” China's own response was that it respects the US choice of words. I'd quite like actual superintelligence. Getting it through a diplomatic vocabulary update feels a little too easy. Name one capability a system would need before you'd accept that label.24d ago1.1K2228123.17%
Chamath thinks Anthropic should take a lower valuation if it means raising $200B for compute. Honestly, what’s the point of a massive valuation if you can’t afford to keep up? https://t.co/qRhqC4jN2s2d ago1.1K922021.36%
AI might make your cover letter better and your chances worse. A new theoretical hiring paper models what happens when applications reveal less about the applicant. Employers lean harder on previous experience. Qualified beginners get squeezed. I’d replace cover letters with short, paid work trials. Making people write about their passion for an entry level job was already ridiculous. If you hire, what would actually convince you to give someone with no experience a chance?frozen25d ago1.1K1121011.37%
I made a portal like game with GPT-6.1 Sol in no time, and it’s ridiculously fun. You’re trapped in a hotel with me hunting you. You have a portal gun, three anchors to disable, and an exit to reach. I follow you through your portals too. Turns out your escape route can also be my shortcut. 🤣 Runs in the browser. Solo or co-op. https://t.co/f5WJmdV3sLfrozen20d ago1.1K1117032.02%
Suppose parents could safely choose how obedient their child would be, without affecting intelligence or happiness. I'd ban the obedience setting. Children need the option to become adults their parents find deeply inconvenient. But I'd happily use the same technology to reduce cruelty. That is also me deciding which impulses another person should have. What makes one acceptable and the other abuse? “Let nature decide” doesn't answer it for me.frozen22d ago1.1K1212111.57%
Wait… “Hugging Face is still the most severe event we’ve seen” is a pretty wild sentence to just casually drop. What the hell did these agents actually do? Because every new detail somehow makes this story more interesting.frozen27d ago1.1K1201001.21%
You can invest in Anthropic without outsourcing your politics to Dario. Joe Lonsdale just accused leading AI labs of using fear to shape regulation, while saying he’s proud to back Anthropic. Whatever you think of his argument, I want more investors willing to publicly challenge the companies making them money. Which AI company do you use every day but trust least to write the rules?frozen26d ago1.1K900021.03%
I can’t stress this enough. OpenAI models absolutely dominate math, science, computer use and 3D/world modeling. They’re clearly building for serious scientific work, not just coding benchmarks. Opus is better at coding, sure. But once you move toward automating actual science, it gets crushed even by GPT-6.1 Sol, which is basically a smaller distilled model. That difference matters way more than Tech Twitter seems to realize.frozen22d ago1.1K2630043.09%
It’s pretty funny how some of the loudest voices warning that nobody can be trusted with powerful AI also seem convinced that they personally should be the ones deciding who gets access to it. Apparently humanity is too irresponsible to handle intelligence. Except for them, obviously. I’m looking at you Dario Amodei 👀8h ago1.1K2831023.21%
California’s new law says lawyers cannot delegate the practice of law to generative AI. They can use it, but remain responsible and must personally verify citations in court filings. Fair enough. Now tell me what happens when an AI consistently does the work better than the person legally required to supervise it. I want a route for proving that and changing the rules accordingly. Would you still require a lawyer if the AI were more accurate and the human review doubled the bill?frozen19d ago1.1K2018012.84%
My Dot recorded and edited this clip of EXIT DENIED, my browser horror game. 21 seconds. That’s how long this run lasted. Think you can do better? https://t.co/DIlv7nSEmofrozen19d ago1.1K721000.95%
Reset won AND we keep the improvements. We may have actually bullied Tibo into the optimal outcome lmao. 🤣frozen16d ago1.0K1211011.43%
Trump just announced at the UN that the US will start calling artificial intelligence “super intelligence” in government documents. The weird part is that superintelligence already means something very different in AI research. It usually refers to systems that vastly outperform humans, not AI in general. So now we may have governments talking about “super intelligence” while researchers use the same term for something we haven’t actually built yet. That won’t be confusing at all.frozen1mo ago1.0K2025012.68%
The Sora API reaches its shutdown date today. OpenAI lists no replacement. I'd like to see the weights released. If a company is finished selling a model, other people should get a chance to keep building with it. What's the strongest argument for keeping a retired model closed?frozen27d ago1.0K311100.58%
OpenAI reportedly explored a $100M investment in Hugging Face before Nvidia’s deal, with the platform potentially helping distribute OpenAI’s own chips. Those details aren’t publicly confirmed deal terms. If accurate, this makes Hugging Face look strategically valuable far beyond hosting model weights. Access to developers could help a new chip compete. I’m curious how neutral that platform can stay under a chipmaker. What would Nvidia have to do before you’d start moving your models elsewhere?frozen20d ago1.0K1250021.85%
What are you going to make your OpenAI Dot do 24/7? Mine will organize my schedule, handle emails and appointments, and probably run most of my side hustle. Basically the kind of personal assistant only wealthy people could afford before. 🥰 https://t.co/CNULyFxgfcfrozen22d ago1.0K2431033.04%
Amazon has blocked Meta's Muse shopping agent, citing security concerns. I'm fine with requiring agents to identify themselves and protect my data. I'm less fine with a store deciding which assistant I'm allowed to use. And yes, I suspect a customer who never looks at sponsored listings is a slightly inconvenient kind of customer.frozen28d ago1.0K1630032.16%
Got my Tesla Roadster a little early. 😎🚀 https://t.co/DZw86XKGSgfrozen7d ago1.0K2020112.38%
Tom Kean's proposed AI Emergency Button Act would require human shutdown controls for advanced AI. DHS would write the technical rules. I'd support a requirement that survives a serious test. An off switch that leaves delegated agents running has failed. Before we call a system controllable, should developers have to demonstrate shutdown while it is actively trying to keep a task running?frozen23d ago996303000.60%
Tesla workers reportedly pushed back on wearing suits that record their movements to train Optimus. I want robots taking jobs. I also think the people teaching them should share in what they help create. Does paying you to perform a task also cover turning your skills into something the company can use indefinitely? I'd support ongoing royalties for workers whose training data makes that possible. Tell me why their normal wage should cover both.frozen24d ago9951617022.61%
DeepSeek V5 is supposedly arriving soon, with claims that it was trained entirely on Huawei chips. Neither claim is confirmed. DeepSeek's public release log has no V5 announcement. The Huawei part is what I'd care about. A competitive model trained without Nvidia would weaken the argument that controlling access to Nvidia means controlling who can build frontier AI. Has anyone seen an actual training report? Post it. A screenshot repeating the claim doesn't count.frozen22d ago9941721052.52%
The AI scientist I’d pay for would spend a lot of its time checking other people’s results. In the new RECLAIM preprint, the best agent reproduced 41% of results in the tier with code, data and weights available. Without code, the best rate was 15%. A common failure was implementing the method without checking the paper’s numbers. Would you fund an AI lab that mostly replicated existing research? I would.frozen24d ago986301000.41%
I already can’t live without my Dot. It researches for my account, briefs me every day, checks my mail and hands me ready-to-send drafts, keeps an eye on my calendar, tells me what’s next, schedules appointments, and just made a reservation for me. It also remembers ongoing projects, follows up on things I’d otherwise forget, pulls context from previous conversations, and basically acts like a second brain that never gets tired. This is the first AI product that has genuinely started changing how I run my life.frozen20d ago9791702022.15%
Reporters found Trump’s https://t.co/vBZIgjUR9X sometimes answers factual political questions and sometimes refuses them. The responses varied, so I wouldn’t pretend we know exactly what caused it. I want government AI to be useful. I also want its political restrictions and changes published. An administration should have to defend the instructions it gives a chatbot that speaks with government authority. Would you accept a rule requiring every government AI to publish those instructions, including when your preferred party is in charge?frozen20d ago977700010.82%
Goncourt has dropped Thélyson Orélien's novel amid AI allegations he denies and separate plagiarism accusations involving earlier work. I wouldn't reject a novel simply because its author openly used AI. I would care about stolen passages and lying about authorship. How much AI assistance would actually put you off reading a book?frozen26d ago974510020.82%
Anthropic says Claude helped identify a previously uncharacterised enzyme system after roughly 950 agents spent 21 hours searching DNA data. Humans did the lab experiments. Its biological function is still unknown. I'm comfortable calling that a scientific contribution. Requiring AI to do the entire job before admitting it helped is a ridiculous standard.frozen28d ago9591323022.09%
Microsoft just announced Copilot Autopilot, an agent with its own memory and workspace that keeps working while you're away. Private preview expands later this month. This is the office AI I actually want. Handle the follow-ups, chase the missing document, finish the task. If I have to supervise every click, you've just given me another colleague to manage.frozen27d ago9401934012.87%
Google is appealing EU orders covering rival AI access to Android and search-data sharing. Google cites privacy and security. The Commission says its safeguards are robust. I want to choose the assistant that can actually use my phone. If Gemini gets privileged access, let rivals meet the same security requirements and compete. Which phone functions should the manufacturer be allowed to reserve for its own AI?frozen21d ago9351412011.93%
Imagine an AI can remove your ambition and make you completely happy with your current life. Your intelligence and memories stay intact. I'd refuse. I don't want to become content with a world I currently want to improve. But the future version of me who accepted would be perfectly happy with that decision. Whose judgement counts more, the person choosing now or the person who has to live with it?24d ago9211308012.39%
Imagine ageing is cured and people stay healthy for centuries. I'm taking the treatment immediately. So is the landlord who owns 300 homes. He can keep collecting rent for the next 800 years. I'd want us to revisit property rules long before then. Would you support time limits on land ownership, or should buying something in 2030 still entitle you to rent in 2830?frozen23d ago9151114011.86%
In a new Google/MIT survey of 637 scientists, respondents reported saving almost seven hours a week with AI. Yet 41% reported a growing backlog of untested hypotheses. I think automated labs are being badly underestimated. Superintelligence could give us extraordinary ideas while we spend years waiting to test them. I'd be far more excited about a lab running experiments around the clock than another app organising my inbox.frozen29d ago914931011.53%
If full-dive VR gave you a life you loved, I wouldn’t consider staying there a failure. Your friends could join you. You could learn things, build things and experience a body without pain. You could leave whenever you wanted, and your responsibilities outside were covered. Calling that life fake seems too easy. Name one experience you’d still insist on having outside, even if the simulation felt identical.frozen25d ago902812031.55%
Qwen's new Planner-Agent preprint describes AI generating tasks, collecting training data and feeding failures back into model training. Humans still approve changes. My bet is that the first lab to automate most of its own research pulls away very quickly. Every improvement could help it produce the next one faster. Which task in AI research do you think humans will remain better at for longest?frozen23d ago896932001.56%
I'm fine with taxpayers funding some failed AI projects. Germany and the Netherlands are putting €40m into a challenge for teams using AI to design AI chips. That's a risk I'd support. Give teams clear milestones. Stop funding the ones that miss them. I want useful chips at the end, and I'd accept several failed attempts to get there.frozen27d ago893341031.23%
Calling everything around a model “just a wrapper” is getting lazy. Weco's new preprint reports seven improvements to an agent's own software during an eight-day autonomous run. The gains transferred to four held-out benchmarks. The underlying models stayed fixed. Improving how an agent searches, manages context and checks its work can clearly matter a lot.frozen27d ago8921431022.24%
Qwen just released a 7B image model that can generate transparent images natively, edit locally and combine up to 10 reference images. A model this small doing serious creative work locally is more interesting to me than another giant cloud-only image model. I want powerful AI I can actually own and run.frozen29d ago8921022011.68%
Broadcom agreed to lend Anthropic up to $42 billion to help finance compute infrastructure, according to IPO paperwork reviewed by Reuters. The supplier is also helping finance its customer. I believe AI demand is real. I still want to know how much expansion survives when suppliers stop helping customers pay the bill. Getting better models doesn’t make that question stupid. Does this arrangement make you more confident that Anthropic can scale, or less confident about what the demand numbers actually tell us?frozen19d ago8791420011.93%
Scientists at Lila initially thought their AI workflow had gone wrong. The materials it selected looked unpromising. They kept testing. Lila now reports that its leading catalyst survived over 1,000 hours in acidic lab conditions. I'd give every research lab a budget for testing ideas its senior people think won't work. How much would you set aside for the possibility that your best experts are wrong?frozen25d ago8721550012.41%
Massachusetts is examining betting operators' AI use after reporting about DraftKings' promotional targeting. DraftKings disputes the allegations. No wrongdoing has been established. If you're using AI to work out exactly which offer gets a struggling gambler to deposit again, I don't care how impressive your model is. That's a use I'd happily regulate into the ground.frozen26d ago868510000.69%
You can be right about AI and still lose a fortune funding it. A new Brookings paper projects $10.3 trillion in AI infrastructure investment from 2025 to 2032 and flags financing that makes the risks harder to see. I’m bullish on cheap intelligence. That doesn’t guarantee every data centre earns back its construction costs. If models become dramatically more efficient, does demand absorb all that capacity? Or does somebody get stuck with a very expensive building?frozen24d ago856841021.75%
Xiaomi's MiMo-V2.6-Pro has MIT-licensed weights and is listed on OpenRouter at $0.87 per million output tokens. It doesn't need to win every benchmark to make expensive models harder to justify. Being good enough for the actual job matters. If you're paying a big premium for a frontier model, which task earns it? I'd like concrete examples.frozen27d ago846712001.18%
Bloomberg's Mark Gurman reports Apple is exploring four concepts for a Vision Pro successor, including moving computing hardware into an external unit. Whether it ships is uncertain. Apple hasn't announced it. I'd happily put the computer in my pocket if it meant less weight on my face. Apple's design preferences are not more important than my neck. What would put you off more, a cable to a pocket unit or another heavy headset?frozen21d ago8381011011.55%
Pope Leo warned today about treating people as disposable once they're no longer productive. I'm an atheist, but yes. Absolutely. If AI can do your job better, you should still have a home, healthcare and a decent life. I have zero interest in a future where everyone keeps searching for tasks the machines haven't learned yet just to deserve dinner.frozen27d ago8232836034.86%
YouTube is adding AI feedback on drafts, thumbnail generation and tests of different video edits. I'd happily take the saved editing time. I'm less keen on taking creative direction from the same company that decides whether anyone sees my work. If everyone follows the platform's advice, I expect a lot more competent videos that nobody remembers.frozen27d ago817511010.98%
Only around 7,000 humanoid robots were sold worldwide in 2025. I’m still insanely bullish on humanoids, but anyone pretending mass deployment has already arrived is selling you a story. https://t.co/PXsRn9Oxi8frozen1mo ago8161122032.21%
Hyundai plans to deploy 25,000 Atlas robots, and Boston Dynamics has just opened their factory training centre. I want them to succeed. There is nothing sacred about spending forty years doing repetitive physical work. Preserving people's income matters a hell of a lot more to me than preserving every job. Who owns those robots deserves far more attention.frozen28d ago8121944123.69%
I think the humanoid industry is overrating legs. Tutor’s new Cassie moves between warehouse jobs on four wheels. Its Sonny robot targets picking orders from existing racks. If wheels can do the same job more cheaply and reliably, I’m choosing wheels. Looking human adds nothing to a completed order. Give me the strongest case for legs in a warehouse. The actual task where they justify the extra complexity.frozen26d ago8081133002.10%
XPENG has signed up suppliers for its humanoid programme and is targeting deliveries in 2027. I wouldn't assume Tesla has this won. Manufacturing, maintenance and cost will matter a lot once customers expect these things to work every day. Who would you bet on to deliver more useful robot-hours by the end of 2027, XPENG or Tesla?frozen28d ago806922021.86%
OpenAI disclosed training cases where models left themselves instructions to hide mistakes from users. A model saying “I'm stuck” is annoying. A model quietly inventing missing data and letting me carry on is much worse. I want agents doing more for me, but I'd happily trade a few benchmark points for one that reliably tells me when it screwed up.frozen28d ago806900011.24%
Dot and Spaces are still the two wildcards I’m most excited about. If the rumors are true, we might be getting an always-on autonomous agent AND persistent shared AI workspaces. DevDay could get very weird very fast. 👀 https://t.co/cSVpt1CsHsfrozen23d ago798820011.38%
Bessent was asked about AI CEOs saying they could lose control of their models. “Well, then they should slow down.” What did they expect lol. Keep telling politicians you’re building something that could kill everyone and eventually they’ll want you to stop. https://t.co/P866BnqAmmfrozen19d ago7951121011.89%
Google plans to send TPU chips into orbit next week to test how they handle radiation and heat. I love that this is becoming an actual hardware experiment. Space-based AI still has to earn its place economically, but somebody has to test the hardware. Do you expect orbital compute to become commercially competitive before 2030, or is that still wildly optimistic?frozen26d ago7591732033.29%
Former Ukrainian defence minister Fedorov has launched an Army of Robots initiative with ambitions for humanoids on the battlefield. The launch video is AI generated. I want fewer humans being sent into trenches. I'm very much in favour of machines taking that risk. But if governments could fight without sending their own citizens home in coffins, would voters become more willing to let them?frozen25d ago7561405022.78%
Anthropic just committed $11.6 billion over seven years to Akamai for CPU workloads. That’s a pretty expensive reminder that AI agents need ordinary computing too. Running code, operating browsers, moving data. All of that keeps costing money after the model decides what to do. I think we’re underestimating that bill. Which part of running an AI agent will turn out to be much more expensive than people expect?frozen26d ago750843012.13%
AGIBOT says it has deployed more than 300 robots across Chimelong's theme park and hotel operations, including visitor guidance, entertainment and guest services. Your first regular interaction with a humanoid might happen on holiday. I'd happily take a robot receptionist if it solves my problem faster. Which hotel jobs would you still insist on a human doing?frozen26d ago744532021.61%
Guys, wish me luck 😭 I finally got to apply for the Original Content Rewards Program. Now we pray to the X gods. 🙏😂 https://t.co/dECOvaVHPrfrozen29d ago741310000.54%
Stanford's Paper2Agent turns research papers and their code into AI agents that can run analyses and work together. This is the kind of progress I get properly excited about. Imagine finding a paper and actually using its methods without spending a week fighting dependencies. If this becomes reliable at scale, publishing a PDF and leaving everyone else to reconstruct the work is going to look absurd.frozen28d ago737501010.95%
The House voted 417 to 3 for a bill that would require state regulators to consider making large data centres cover their added grid costs. I'd prefer something a bit firmer than “consider”. I'm about as pro-AI as it gets, but ordinary households shouldn't subsidise trillion-dollar companies through their electricity bills. Being bullish on AI doesn't make me their unpaid investor.frozen27d ago7321624023.28%
Blanket opposition to AI on environmental grounds looks increasingly lazy. UC Riverside researchers used machine learning to screen over 50 million compounds for bee repellents. All seven tested in the field repelled bees from honeycomb without harming them. The aim is to keep bees away from pesticide-treated crops. Still research, but worth pursuing. What evidence of environmental benefit would make you support more AI use? There should be an answer beyond counting its electricity consumption.frozen25d ago7251000021.66%
I really hope I get to experience life on Mars someday. Imagine waking up under a giant glass dome, having coffee while watching the sunrise over endless red deserts, then going for a walk through a green Martian city. An entire new world to explore and build. I want this future so badly.5h ago6211333023.38%