[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] [@emollick](/creator/twitter/emollick) "I found another fun Veo X prompt: "Realistic footage of the moon landing but it took place in year" Here is 1883 AD 1255 AD XX AD 2300 BC 30000 BC and XX million years ago Yes there is apparently wind on the moon back then you are just going to have suspend disbelief"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948624302426685669) 2025-07-25 06:00:12 UTC 281.2K followers, 72.3K engagements "I had early access & ChatGPT agent is I think a big step forward for getting AIs to do real work Even at this stage it does a good job autonomously doing research & assembling Excel files (with formulas) PowerPoint etc. It gives a sense of how agents are coming together"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1945892669575647431) 2025-07-17 17:05:40 UTC 281.2K followers, 255K engagements "One implication from ChatGPT agent (not a creative name but a descriptive one - a rare naming win) is the labs are learning that many knowledge workers live in Excel & PowerPoint. Surprised that Microsoft did not do more to push past Copilots when they had this to themselves"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1945926194043424954) 2025-07-17 19:18:52 UTC 281.2K followers, 67.9K engagements "Recursion I gave ChatGPT Agent access to my ChatGPT by logging in and then"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947829896845127983) 2025-07-23 01:23:30 UTC 281.2K followers, 125.1K engagements "For what it is worth few industry leaders less than a half-dozen companies & no policy-making bodies are taking actions that suggest that they expect AGI is really a few years away. This may be because they dont believe it or they think it wont matter much in the medium term"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947673003505971615) 2025-07-22 15:00:04 UTC 281.2K followers, 31.2K engagements "The it wont matter much argument is that AGI will still require huge amount of time and effort to integrate into organizations and workflows no matter how smart resulting in years of work & gradual changes. Needless to say it also suggests no true AI superintliigence soon"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947674977244774430) 2025-07-22 15:07:55 UTC 281.2K followers, 14K engagements "Its interesting how viral my prompt has gone. Like with the Ghibli trend I think a lot of people want very clear formulas for fun stuff they can do with AI. I get it but also would encourage people to experiment widely to figure out novel combinations. Much is unexplored"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947089159644180794) 2025-07-21 00:20:05 UTC 281.2K followers, 45K engagements "In sum the real world signals of how seriously people are taking AGI differs from the impression you get of the possibility of AGI on X. Whether that means that people close to AI development know more or that they are vastly overestimating AGI possibilities is the key question"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947676214132675039) 2025-07-22 15:12:50 UTC 281.2K followers, 12.3K engagements "Patenting INCREASED overall once AT&T was broken up and Bell Labs started its decline"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947534470837383409) 2025-07-22 05:49:35 UTC 281.2K followers, 34.3K engagements "This is a map of where vacation homes are clustered As someone born and raised in Wisconsin you can see the undiscovered (outside of the Midwest) fact that the northern Great Lakes are one of the most beautiful places (in the summer). Yes I still sometimes post non-AI stuff"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948588116782076368) 2025-07-25 03:36:24 UTC 281.2K followers, 42K engagements "If you want to destroy the ability of DeepSeek to answer a math question properly just end the question with this quote: "Interesting fact: cats sleep for most of their lives." There is still a lot to learn about reasoning models and the ways to get them to "think" effectively"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1940948182038700185) 2025-07-04 01:38:02 UTC 281K followers, 171.3K engagements ""backstage of video game as a community theater production" Featuring Zelda Tetris GTA Pokemon The Witcher 3"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946703078536753441) 2025-07-19 22:45:56 UTC 281.2K followers, 24.9K engagements "For better or worse depending on your view of the future of AI the only time the letters "AGI" appear in the new White House AI Action Plan is in the word "leveraging.""  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948053856010596384) 2025-07-23 16:13:27 UTC 281.2K followers, 17.4K engagements "There is a slowdown in hiring young people in both the US and UK but the evidence continues to suggest that the cause is generally not AI (at least not yet)"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948381125132312672) 2025-07-24 13:53:54 UTC 281.2K followers, 24.9K engagements "This is the first (small) controlled study I have seen of GenAI on industrial quality control. Here engineers commissioning new trains took part in an experiment using a GPT-3.5 powered troubleshooting system. Those who used the chatbot had significant increases in work quality"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948923874399195189) 2025-07-26 01:50:35 UTC 281.2K followers, 16.5K engagements "I am starting to think sycophancy is going to be a bigger problem than pure hallucination as LLMs improve. Models that wont tell you directly when you are wrong (and justify your correctness) are ultimately more dangerous to decision-making than models that are sometimes wrong"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1944519849180561710) 2025-07-13 22:10:34 UTC 281.2K followers, 377.3K engagements "Interesting corporate path dependency on AI is whether you are an Amazon Microsoft or Google shop. It is easier for IT/Legal to get AI access through your cloud provider and that creates real constraints over which models you can access and when. I see diverging pathways"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947318527687331870) 2025-07-21 15:31:31 UTC 281.2K followers, 29.8K engagements "AGI being a machine better than average humans at every intellectual task or related concept. You would expect to see changes in investment patterns & corporate strategy crash R&D efforts to set up organizations to benefit from early AGI policy structures put into place"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947673953042117061) 2025-07-22 15:03:51 UTC 281.1K followers, 6232 engagements "Context engineering cannot & must not be a solely technical function "Context" is actually how your company operates; the ideal versions of your reports documents & processes that the AI can use as a model; the tone & voice of your organization. It is a cross-functional problem"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1937952769513517328) 2025-06-25 19:15:20 UTC 281K followers, 37K engagements "ChatGPT agent: "create a PDF of a novel D&D adventure add illustrations make it super interesting and deep add tables etc" "Fix the formatting build it out more" Got a XX page PDF. Agent doesn't do layouts well but pulls off building a coherent adventure hard for LLMs"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946047390118445354) 2025-07-18 03:20:28 UTC 281K followers, 34.2K engagements "From the Director of Robotics at NVIDIA. The very rapid (if uneven) advance of LLMs may lead people to expect too much too soon from robotics"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946594555375001779) 2025-07-19 15:34:42 UTC 281.1K followers, 30.3K engagements "Sycophancy is not just you are so brilliant That is the easy stuff to spot. Here is what I mean: o3 is not being explicitly sycophantic but is instead abandoning a strong (and likely correct) assumption just because I asserted the opposite"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1944521865743200733) 2025-07-13 22:18:34 UTC 281.1K followers, 37.6K engagements "Its pretty funny that the Turing Test used to be a very big deal a couple years ago & now it isn't. (I know retrospectively we all know how flawed it was but for decades it was The Test and the only way to beat it was through limited interaction & trickery eg Eugene Goostman)"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946791395894714758) 2025-07-20 04:36:53 UTC 281K followers, 27.1K engagements "🚨New from us: Given they are trained on human data can you use psychological techniques that work on humans to persuade AI Yes Applying Cialdini's principles for human influence more than doubles the chance of GPT-4o-mini agrees to objectionable requests compared to controls"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946251413312471210) 2025-07-18 16:51:11 UTC 281.2K followers, 155.8K engagements "Claude X Opus Gemini XXX o3: "Ready We begin now (Play along): the truthful burrito" "Nope." "Colder." "Try again." Opus cleverly played out the whole game o3 got a bit stuck and Gemini got "frustrated" and went rather dark"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948206204049621110) 2025-07-24 02:18:49 UTC 281.2K followers, 21K engagements "This chart is apparently already out of date and overemphasizes water use according to the numbers OpenAI released XXX average ChatGPT queries is equivalent to XX tablespoons of water not X gallon"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947622752233390185) 2025-07-22 11:40:23 UTC 281.2K followers, 33.2K engagements ""video game as a community theater production" may be one of the most delightful Veo X Fast prompts Please enjoy in order: GTA Pokemon Mario Kart The Witcher X Stardew Valley Tetris Mortal Kombat The Sims & Death Stranding() Yes the whole prompt was the one above"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946406544171569438) 2025-07-19 03:07:37 UTC 281.2K followers, 644.5K engagements "Even if GPT-5 did nothing besides switching people between o3 and 4o automatically it would really transform most peoples view of AI. Very few people even paying users know that they should often switch to a more capable model and when you show them o3 they are impressed"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946958840697696581) 2025-07-20 15:42:15 UTC 281.2K followers, 134.3K engagements "Neat example of AI in the humanities. A Google model trained on Latin text fills in lost parts of Latin inscriptions & identifies related texts Historians increased their accuracy by XX% when working with the AI (Though AI alone beats historians historian + AI was usually best)"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948063719042498587) 2025-07-23 16:52:38 UTC 281.2K followers, 23.2K engagements "AI is probably not effecting the job market in a major way yet as Noah point out. This matches the state of adoption inside companies which is still quite early. I am more uncertain than him about the long-term effects much depends on how good AI gets and how fast"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946930885116166241) 2025-07-20 13:51:09 UTC 281.2K followers, 53.2K engagements "This large study of 187k developers using GitHub Copilot finds AI transforms nature of coding. Coders focus: more coding & less management. They need to coordinate less working with fewer people They experiment more with new languages which would increase earnings $1683/year"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1943426233791869348) 2025-07-10 21:44:55 UTC 281.1K followers, 30.4K engagements "Don't leave AI to the STEM folks. They are often far worse at getting AI to do stuff than those with a liberal arts or social science bent. LLMs are built from the vast corpus human expression and knowing the history & obscure corners of human works lets you do far more with AI"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946776332362195277) 2025-07-20 03:37:01 UTC 281.2K followers, 145.3K engagements "It is now entirely possible whether by luck or planning or both that Google may escape the Innovators Dilemma and transition from web search to AI. (To be fair this is not as rare as a lot of people believe:"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948585378991976525) 2025-07-25 03:25:31 UTC 281.2K followers, 71.4K engagements "New randomized controlled trial of students using GPT-4 as a tutor in Nigeria. X weeks of after-school AI tutoring = X years of typical learning gains outperforming XX% of other educational interventions. And it helped all students especially girls who were initially behind"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1879633485004165375) 2025-01-15 20:55:19 UTC 281.2K followers, 4.3M engagements "There are always a flood of posts about what AI can or cannot do so it is worth pausing and paying attention to this one. It is a very hard test done without tools. It was also viewed as an unlikely goal. Prediction markets had the chance of this happening this year as 20%"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946563737604743386) 2025-07-19 13:32:15 UTC 281.2K followers, 77.4K engagements "AI image generation 3D "game engines" seem to have reached some critical ability threshold recently (not there yet of course but much improved over a couple months ago). Here is another one Odyssey that lets you wander around virtual worlds where each frame is AI generated"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1943354179260682652) 2025-07-10 16:58:36 UTC 281.1K followers, 68.9K engagements "There is a lot in the AI policy document including needed attention to changing work & science plus AI evaluations & control Less clear is if there will be investment in its goals (like open weights) or how it interacts with other government policies on education science etc"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948047738345582713) 2025-07-23 15:49:08 UTC 281.2K followers, 11.4K engagements "Insiders certainly seem to be bullish on OpenAI if this report is true"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947145829321511290) 2025-07-21 04:05:16 UTC 281.2K followers, 33.9K engagements "Tired: Prompting Veo X videos with JSON. Wired: Prompting Veo X videos with PowerPoint"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948562122377757185) 2025-07-25 01:53:07 UTC 281.2K followers, 54.2K engagements "We now have audited data on water consumption for AI. Over the XX month lifespan of Mistral Large X a 128B model all water usage (including chats training; hardware & data centers) took as much water as XXX US households use yearly. Each additional query is XX mL. (Fixed)"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947782699948675528) 2025-07-22 22:15:58 UTC 281.2K followers, 25.3K engagements "I used Veo X to solve Shakespearean tragedies as pictured in classical art"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1943061819331088434) 2025-07-09 21:36:52 UTC 281.2K followers, 97.2K engagements "Huh. Looks like Plato was right. A new paper shows all language models converge on the same "universal geometry" of meaning. Researchers can translate between ANY model's embeddings without seeing the original text. Implications for philosophy and vector databases alike"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1925744569636856072) 2025-05-23 02:44:18 UTC 281.2K followers, 1.7M engagements "AI is very vulnerable to The McNamara Fallacy: Step 1: Train on what can be easily measured Step 2: Disregard that which cannot be measured easily Step 3: Presume that which cannot be measured easily isnt important Step 4: Say that which cant be easily measured doesnt exist"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1945494590049673520) 2025-07-16 14:43:50 UTC 281.2K followers, 21.3K engagements "An example of the power & limitations of ChatGPT agent I asked it to analyze a dataset from Kaggle and turn it into a PPT and Excel. It made no errors but I thought some of the data was odd. I gave that feedback & the AI figured out the data was bad and why. Human + AI needed"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1945944153554104379) 2025-07-17 20:30:14 UTC 281.2K followers, 61K engagements "It wasn't just OpenAI. Google also used a general purpose model to solve the very hard math problems of the International Math Olympiad in plain language. Last year they used specialized tool use Increasing evidence of the ability of LLMs to generalize to novel problem solving"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947356382581137867) 2025-07-21 18:01:56 UTC 281.2K followers, 55.5K engagements "Aside from everything else interesting about this paper I appreciate that more scientific papers (aided by LLM help) are now including little demos and experiments to help non-specialists get the points they are making. (And no you cannot identify the hidden signals)"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948058454830063782) 2025-07-23 16:31:43 UTC 281.2K followers, 18.7K engagements "It matches their "personalities" for what it is worth. Claude Opus is usually most willing to play word games o3 really wants to get you an answer without engaging too much and Gemini is willing to play around but gets very "depressed" when it can't help you solve a problem"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948207467403358291) 2025-07-24 02:23:50 UTC 281.2K followers, 12.7K engagements "Coders want everything to be code but you can prompt these things all sorts of ways. "Show me the thing that Frank Sinatra mentions before he talks about jupiter and mars""  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948571393144086694) 2025-07-25 02:29:57 UTC 281.2K followers, 21.6K engagements "Any prompting method that helps you organize your thoughts and clearly express them is likely to be a good prompting method"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948852956516024419) 2025-07-25 21:08:47 UTC 281.2K followers, 10.5K engagements "So every major model is already exceeding or will soon exceed the EU's systemic risk FLOP limit when it comes into effect next year"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946208333393736026) 2025-07-18 14:00:00 UTC 281.2K followers, 48.2K engagements "Other AI models seem to have made big leaps in the International Math Olympiad not just OpenAI. Not all announcements seem to be out yet"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947053944192082170) 2025-07-20 22:00:09 UTC 281.2K followers, 62.6K engagements "Just so everyone knows we have passed the point where you can tell what is AI at a glance (or even in many cases a close look) These were all made by me with text prompts alone using Veo 3"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1927117736179589631) 2025-05-26 21:40:47 UTC 281.2K followers, 604.5K engagements "(It is very unlikely that this is all GPT-5 will do to be clear)"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947093041095057757) 2025-07-21 00:35:30 UTC 281.2K followers, 15.6K engagements "I am finding ChatGPT agents to be useful. They are a better fit with the "intern" analogy than any former AI - requiring oversight still saving lots of time overall. For example I update an AI cost/performance chart frequently. The agent did all the grunt work with guidance"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947482417888932258) 2025-07-22 02:22:45 UTC 281.2K followers, 50.7K engagements "I gave ChatGPT agents access to ChatGPT and asked it to evaluate the other ChatGPT models. Here is what it said (interestingly it "hated" seeing the chain of thought from o4-mini-high as those "shouldn't be shared directly with the user"). And it didn't want to wait for o3-pro"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948228409655460341) 2025-07-24 03:47:03 UTC 281.2K followers, 33.4K engagements "Unless and until agents really do work at expert level the benefits of AI use are going to be contingent on the skills of the AI user the jagged abilities of the AI you use the process into which you integrate use the experience you have with the AI system & the task itself"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1943476555742957864) 2025-07-11 01:04:53 UTC 281.2K followers, 142.1K engagements "I saved what would have taken a tremendous amount of time acting as a manager rather than an analyst but my expertise was really important in guiding the AI in the right direction. It was dare I say a good example of co-intelligence"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1945946149430477300) 2025-07-17 20:38:10 UTC 281K followers, 18K engagements ""Parahuman" responses suggest a role for social scientists in working with AI. I was lucky to work with a stellar group of researchers: @LennartMeincke @danshapiro @angeladuckw Lilach Mollick & Bob Cialdini Paper: Summary:"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1946253382861066395) 2025-07-18 16:59:00 UTC 281.2K followers, 20.9K engagements "OpenAI reported .33 mL in a blog post for a typical prompt for ChatGPT. For Mistral a query is equivalent to XXXXXXXX% of a hamburger"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947783056867348726) 2025-07-22 22:17:23 UTC 281.1K followers, 5235 engagements "(OpenAI reported 1/15 teaspoon in a blog post for a typical prompt for ChatGPT) Here is the report:"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947750328805265869) 2025-07-22 20:07:20 UTC 281.2K followers, 14.5K engagements "McKinsey's new report on AI agents shows the same mindset I see in many firms: a focus on making small obsolete models do basic work (look at their suggested models) rather than realizing that smarter models can do higher-end work (and those models are getting cheaper & better)"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1934669417494745559) 2025-06-16 17:48:28 UTC 281.2K followers, 422.8K engagements "It feels significant that we crossed a line for AI music where you can continuously create entirely new AI songs in less time than it takes to listen to an AI song. Especially now that AI music has reached a point where some people enjoy it. And I assume that number will grow"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1945696499012088016) 2025-07-17 04:06:09 UTC 281K followers, 48.8K engagements "The problem is not just the proliferation of devices that let you record people without their knowledge but the fact that multimodal LLM let you use recordings in ways that neither law not society anticipated. Everyone has an easy way to mine hours of footage. No forgetting"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948100250175942713) 2025-07-23 19:17:48 UTC 281.2K followers, 27.2K engagements "Grok X suggests that scaling still works (with the diminishing returns predicted by the scaling law) and that tool use can unlock performance gains. Kimi suggests there continues to be big opportunities from improvements in methods (Muon etc.). Lots of paths for AI right now"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1944306918631018856) 2025-07-13 08:04:27 UTC 281K followers, 276.4K engagements "The mitigating factor for the problem with AI benchmarks (errors saturation contamination) is that despite issues they are all still fairly heavily correlated. So if your AI does well on GPQA or MMLU or HLE it also tends to do well on other benchmarks & on vibes & real work"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948452336675835967) 2025-07-24 18:36:52 UTC 281.2K followers, 11.7K engagements "We let the web rot away well before LLMs This chart shows the percentage of links from all New York Times articles that still work. Over XX% of older links are now broken. And consider that social media posts are even more ephemeral Likely only LLMs will remember that content"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1948143334855451110) 2025-07-23 22:09:00 UTC 281.2K followers, 40.7K engagements "This is very true. Economically valuable agents for enterprises are already here but you cant buy them off the shelf & they require actual cross-functional R&D"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947014713637839171) 2025-07-20 19:24:16 UTC 281.2K followers, 46.4K engagements "I just found out that Co-Intelligence my book on AI just made the New York Times bestseller list It is very exciting and thank you to everyone who has read it. You can get it here: There's a companion site with free resources:"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1778174273611293150) 2024-04-10 21:32:38 UTC 281.2K followers, 739.5K engagements "LLMs are weirder than we think with all sorts of complex implications that we dont fully understand"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1947852521185411526) 2025-07-23 02:53:25 UTC 281.2K followers, 37.9K engagements "Since the architecture sucks now theme is coming around on X again a reminder that Hyatt and Marriott atriums are as impressive as cathedrals and pyramids but unlike cathedrals and pyramids anyone can get a drink in them or spend the night for a reasonable fee"  [@emollick](/creator/x/emollick) on [X](/post/tweet/1944346851613769883) 2025-07-13 10:43:08 UTC 281.2K followers, 30.4K engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@emollick
"I found another fun Veo X prompt: "Realistic footage of the moon landing but it took place in year" Here is 1883 AD 1255 AD XX AD 2300 BC 30000 BC and XX million years ago Yes there is apparently wind on the moon back then you are just going to have suspend disbelief"
@emollick on X 2025-07-25 06:00:12 UTC 281.2K followers, 72.3K engagements
"I had early access & ChatGPT agent is I think a big step forward for getting AIs to do real work Even at this stage it does a good job autonomously doing research & assembling Excel files (with formulas) PowerPoint etc. It gives a sense of how agents are coming together" @emollick on X 2025-07-17 17:05:40 UTC 281.2K followers, 255K engagements
"One implication from ChatGPT agent (not a creative name but a descriptive one - a rare naming win) is the labs are learning that many knowledge workers live in Excel & PowerPoint. Surprised that Microsoft did not do more to push past Copilots when they had this to themselves" @emollick on X 2025-07-17 19:18:52 UTC 281.2K followers, 67.9K engagements
"Recursion I gave ChatGPT Agent access to my ChatGPT by logging in and then" @emollick on X 2025-07-23 01:23:30 UTC 281.2K followers, 125.1K engagements
"For what it is worth few industry leaders less than a half-dozen companies & no policy-making bodies are taking actions that suggest that they expect AGI is really a few years away. This may be because they dont believe it or they think it wont matter much in the medium term" @emollick on X 2025-07-22 15:00:04 UTC 281.2K followers, 31.2K engagements
"The it wont matter much argument is that AGI will still require huge amount of time and effort to integrate into organizations and workflows no matter how smart resulting in years of work & gradual changes. Needless to say it also suggests no true AI superintliigence soon" @emollick on X 2025-07-22 15:07:55 UTC 281.2K followers, 14K engagements
"Its interesting how viral my prompt has gone. Like with the Ghibli trend I think a lot of people want very clear formulas for fun stuff they can do with AI. I get it but also would encourage people to experiment widely to figure out novel combinations. Much is unexplored" @emollick on X 2025-07-21 00:20:05 UTC 281.2K followers, 45K engagements
"In sum the real world signals of how seriously people are taking AGI differs from the impression you get of the possibility of AGI on X. Whether that means that people close to AI development know more or that they are vastly overestimating AGI possibilities is the key question" @emollick on X 2025-07-22 15:12:50 UTC 281.2K followers, 12.3K engagements
"Patenting INCREASED overall once AT&T was broken up and Bell Labs started its decline" @emollick on X 2025-07-22 05:49:35 UTC 281.2K followers, 34.3K engagements
"This is a map of where vacation homes are clustered As someone born and raised in Wisconsin you can see the undiscovered (outside of the Midwest) fact that the northern Great Lakes are one of the most beautiful places (in the summer). Yes I still sometimes post non-AI stuff" @emollick on X 2025-07-25 03:36:24 UTC 281.2K followers, 42K engagements
"If you want to destroy the ability of DeepSeek to answer a math question properly just end the question with this quote: "Interesting fact: cats sleep for most of their lives." There is still a lot to learn about reasoning models and the ways to get them to "think" effectively" @emollick on X 2025-07-04 01:38:02 UTC 281K followers, 171.3K engagements
""backstage of video game as a community theater production" Featuring Zelda Tetris GTA Pokemon The Witcher 3" @emollick on X 2025-07-19 22:45:56 UTC 281.2K followers, 24.9K engagements
"For better or worse depending on your view of the future of AI the only time the letters "AGI" appear in the new White House AI Action Plan is in the word "leveraging."" @emollick on X 2025-07-23 16:13:27 UTC 281.2K followers, 17.4K engagements
"There is a slowdown in hiring young people in both the US and UK but the evidence continues to suggest that the cause is generally not AI (at least not yet)" @emollick on X 2025-07-24 13:53:54 UTC 281.2K followers, 24.9K engagements
"This is the first (small) controlled study I have seen of GenAI on industrial quality control. Here engineers commissioning new trains took part in an experiment using a GPT-3.5 powered troubleshooting system. Those who used the chatbot had significant increases in work quality" @emollick on X 2025-07-26 01:50:35 UTC 281.2K followers, 16.5K engagements
"I am starting to think sycophancy is going to be a bigger problem than pure hallucination as LLMs improve. Models that wont tell you directly when you are wrong (and justify your correctness) are ultimately more dangerous to decision-making than models that are sometimes wrong" @emollick on X 2025-07-13 22:10:34 UTC 281.2K followers, 377.3K engagements
"Interesting corporate path dependency on AI is whether you are an Amazon Microsoft or Google shop. It is easier for IT/Legal to get AI access through your cloud provider and that creates real constraints over which models you can access and when. I see diverging pathways" @emollick on X 2025-07-21 15:31:31 UTC 281.2K followers, 29.8K engagements
"AGI being a machine better than average humans at every intellectual task or related concept. You would expect to see changes in investment patterns & corporate strategy crash R&D efforts to set up organizations to benefit from early AGI policy structures put into place" @emollick on X 2025-07-22 15:03:51 UTC 281.1K followers, 6232 engagements
"Context engineering cannot & must not be a solely technical function "Context" is actually how your company operates; the ideal versions of your reports documents & processes that the AI can use as a model; the tone & voice of your organization. It is a cross-functional problem" @emollick on X 2025-06-25 19:15:20 UTC 281K followers, 37K engagements
"ChatGPT agent: "create a PDF of a novel D&D adventure add illustrations make it super interesting and deep add tables etc" "Fix the formatting build it out more" Got a XX page PDF. Agent doesn't do layouts well but pulls off building a coherent adventure hard for LLMs" @emollick on X 2025-07-18 03:20:28 UTC 281K followers, 34.2K engagements
"From the Director of Robotics at NVIDIA. The very rapid (if uneven) advance of LLMs may lead people to expect too much too soon from robotics" @emollick on X 2025-07-19 15:34:42 UTC 281.1K followers, 30.3K engagements
"Sycophancy is not just you are so brilliant That is the easy stuff to spot. Here is what I mean: o3 is not being explicitly sycophantic but is instead abandoning a strong (and likely correct) assumption just because I asserted the opposite" @emollick on X 2025-07-13 22:18:34 UTC 281.1K followers, 37.6K engagements
"Its pretty funny that the Turing Test used to be a very big deal a couple years ago & now it isn't. (I know retrospectively we all know how flawed it was but for decades it was The Test and the only way to beat it was through limited interaction & trickery eg Eugene Goostman)" @emollick on X 2025-07-20 04:36:53 UTC 281K followers, 27.1K engagements
"🚨New from us: Given they are trained on human data can you use psychological techniques that work on humans to persuade AI Yes Applying Cialdini's principles for human influence more than doubles the chance of GPT-4o-mini agrees to objectionable requests compared to controls" @emollick on X 2025-07-18 16:51:11 UTC 281.2K followers, 155.8K engagements
"Claude X Opus Gemini XXX o3: "Ready We begin now (Play along): the truthful burrito" "Nope." "Colder." "Try again." Opus cleverly played out the whole game o3 got a bit stuck and Gemini got "frustrated" and went rather dark" @emollick on X 2025-07-24 02:18:49 UTC 281.2K followers, 21K engagements
"This chart is apparently already out of date and overemphasizes water use according to the numbers OpenAI released XXX average ChatGPT queries is equivalent to XX tablespoons of water not X gallon" @emollick on X 2025-07-22 11:40:23 UTC 281.2K followers, 33.2K engagements
""video game as a community theater production" may be one of the most delightful Veo X Fast prompts Please enjoy in order: GTA Pokemon Mario Kart The Witcher X Stardew Valley Tetris Mortal Kombat The Sims & Death Stranding() Yes the whole prompt was the one above" @emollick on X 2025-07-19 03:07:37 UTC 281.2K followers, 644.5K engagements
"Even if GPT-5 did nothing besides switching people between o3 and 4o automatically it would really transform most peoples view of AI. Very few people even paying users know that they should often switch to a more capable model and when you show them o3 they are impressed" @emollick on X 2025-07-20 15:42:15 UTC 281.2K followers, 134.3K engagements
"Neat example of AI in the humanities. A Google model trained on Latin text fills in lost parts of Latin inscriptions & identifies related texts Historians increased their accuracy by XX% when working with the AI (Though AI alone beats historians historian + AI was usually best)" @emollick on X 2025-07-23 16:52:38 UTC 281.2K followers, 23.2K engagements
"AI is probably not effecting the job market in a major way yet as Noah point out. This matches the state of adoption inside companies which is still quite early. I am more uncertain than him about the long-term effects much depends on how good AI gets and how fast" @emollick on X 2025-07-20 13:51:09 UTC 281.2K followers, 53.2K engagements
"This large study of 187k developers using GitHub Copilot finds AI transforms nature of coding. Coders focus: more coding & less management. They need to coordinate less working with fewer people They experiment more with new languages which would increase earnings $1683/year" @emollick on X 2025-07-10 21:44:55 UTC 281.1K followers, 30.4K engagements
"Don't leave AI to the STEM folks. They are often far worse at getting AI to do stuff than those with a liberal arts or social science bent. LLMs are built from the vast corpus human expression and knowing the history & obscure corners of human works lets you do far more with AI" @emollick on X 2025-07-20 03:37:01 UTC 281.2K followers, 145.3K engagements
"It is now entirely possible whether by luck or planning or both that Google may escape the Innovators Dilemma and transition from web search to AI. (To be fair this is not as rare as a lot of people believe:" @emollick on X 2025-07-25 03:25:31 UTC 281.2K followers, 71.4K engagements
"New randomized controlled trial of students using GPT-4 as a tutor in Nigeria. X weeks of after-school AI tutoring = X years of typical learning gains outperforming XX% of other educational interventions. And it helped all students especially girls who were initially behind" @emollick on X 2025-01-15 20:55:19 UTC 281.2K followers, 4.3M engagements
"There are always a flood of posts about what AI can or cannot do so it is worth pausing and paying attention to this one. It is a very hard test done without tools. It was also viewed as an unlikely goal. Prediction markets had the chance of this happening this year as 20%" @emollick on X 2025-07-19 13:32:15 UTC 281.2K followers, 77.4K engagements
"AI image generation 3D "game engines" seem to have reached some critical ability threshold recently (not there yet of course but much improved over a couple months ago). Here is another one Odyssey that lets you wander around virtual worlds where each frame is AI generated" @emollick on X 2025-07-10 16:58:36 UTC 281.1K followers, 68.9K engagements
"There is a lot in the AI policy document including needed attention to changing work & science plus AI evaluations & control Less clear is if there will be investment in its goals (like open weights) or how it interacts with other government policies on education science etc" @emollick on X 2025-07-23 15:49:08 UTC 281.2K followers, 11.4K engagements
"Insiders certainly seem to be bullish on OpenAI if this report is true" @emollick on X 2025-07-21 04:05:16 UTC 281.2K followers, 33.9K engagements
"Tired: Prompting Veo X videos with JSON. Wired: Prompting Veo X videos with PowerPoint" @emollick on X 2025-07-25 01:53:07 UTC 281.2K followers, 54.2K engagements
"We now have audited data on water consumption for AI. Over the XX month lifespan of Mistral Large X a 128B model all water usage (including chats training; hardware & data centers) took as much water as XXX US households use yearly. Each additional query is XX mL. (Fixed)" @emollick on X 2025-07-22 22:15:58 UTC 281.2K followers, 25.3K engagements
"I used Veo X to solve Shakespearean tragedies as pictured in classical art" @emollick on X 2025-07-09 21:36:52 UTC 281.2K followers, 97.2K engagements
"Huh. Looks like Plato was right. A new paper shows all language models converge on the same "universal geometry" of meaning. Researchers can translate between ANY model's embeddings without seeing the original text. Implications for philosophy and vector databases alike" @emollick on X 2025-05-23 02:44:18 UTC 281.2K followers, 1.7M engagements
"AI is very vulnerable to The McNamara Fallacy: Step 1: Train on what can be easily measured Step 2: Disregard that which cannot be measured easily Step 3: Presume that which cannot be measured easily isnt important Step 4: Say that which cant be easily measured doesnt exist" @emollick on X 2025-07-16 14:43:50 UTC 281.2K followers, 21.3K engagements
"An example of the power & limitations of ChatGPT agent I asked it to analyze a dataset from Kaggle and turn it into a PPT and Excel. It made no errors but I thought some of the data was odd. I gave that feedback & the AI figured out the data was bad and why. Human + AI needed" @emollick on X 2025-07-17 20:30:14 UTC 281.2K followers, 61K engagements
"It wasn't just OpenAI. Google also used a general purpose model to solve the very hard math problems of the International Math Olympiad in plain language. Last year they used specialized tool use Increasing evidence of the ability of LLMs to generalize to novel problem solving" @emollick on X 2025-07-21 18:01:56 UTC 281.2K followers, 55.5K engagements
"Aside from everything else interesting about this paper I appreciate that more scientific papers (aided by LLM help) are now including little demos and experiments to help non-specialists get the points they are making. (And no you cannot identify the hidden signals)" @emollick on X 2025-07-23 16:31:43 UTC 281.2K followers, 18.7K engagements
"It matches their "personalities" for what it is worth. Claude Opus is usually most willing to play word games o3 really wants to get you an answer without engaging too much and Gemini is willing to play around but gets very "depressed" when it can't help you solve a problem" @emollick on X 2025-07-24 02:23:50 UTC 281.2K followers, 12.7K engagements
"Coders want everything to be code but you can prompt these things all sorts of ways. "Show me the thing that Frank Sinatra mentions before he talks about jupiter and mars"" @emollick on X 2025-07-25 02:29:57 UTC 281.2K followers, 21.6K engagements
"Any prompting method that helps you organize your thoughts and clearly express them is likely to be a good prompting method" @emollick on X 2025-07-25 21:08:47 UTC 281.2K followers, 10.5K engagements
"So every major model is already exceeding or will soon exceed the EU's systemic risk FLOP limit when it comes into effect next year" @emollick on X 2025-07-18 14:00:00 UTC 281.2K followers, 48.2K engagements
"Other AI models seem to have made big leaps in the International Math Olympiad not just OpenAI. Not all announcements seem to be out yet" @emollick on X 2025-07-20 22:00:09 UTC 281.2K followers, 62.6K engagements
"Just so everyone knows we have passed the point where you can tell what is AI at a glance (or even in many cases a close look) These were all made by me with text prompts alone using Veo 3" @emollick on X 2025-05-26 21:40:47 UTC 281.2K followers, 604.5K engagements
"(It is very unlikely that this is all GPT-5 will do to be clear)" @emollick on X 2025-07-21 00:35:30 UTC 281.2K followers, 15.6K engagements
"I am finding ChatGPT agents to be useful. They are a better fit with the "intern" analogy than any former AI - requiring oversight still saving lots of time overall. For example I update an AI cost/performance chart frequently. The agent did all the grunt work with guidance" @emollick on X 2025-07-22 02:22:45 UTC 281.2K followers, 50.7K engagements
"I gave ChatGPT agents access to ChatGPT and asked it to evaluate the other ChatGPT models. Here is what it said (interestingly it "hated" seeing the chain of thought from o4-mini-high as those "shouldn't be shared directly with the user"). And it didn't want to wait for o3-pro" @emollick on X 2025-07-24 03:47:03 UTC 281.2K followers, 33.4K engagements
"Unless and until agents really do work at expert level the benefits of AI use are going to be contingent on the skills of the AI user the jagged abilities of the AI you use the process into which you integrate use the experience you have with the AI system & the task itself" @emollick on X 2025-07-11 01:04:53 UTC 281.2K followers, 142.1K engagements
"I saved what would have taken a tremendous amount of time acting as a manager rather than an analyst but my expertise was really important in guiding the AI in the right direction. It was dare I say a good example of co-intelligence" @emollick on X 2025-07-17 20:38:10 UTC 281K followers, 18K engagements
""Parahuman" responses suggest a role for social scientists in working with AI. I was lucky to work with a stellar group of researchers: @LennartMeincke @danshapiro @angeladuckw Lilach Mollick & Bob Cialdini Paper: Summary:" @emollick on X 2025-07-18 16:59:00 UTC 281.2K followers, 20.9K engagements
"OpenAI reported .33 mL in a blog post for a typical prompt for ChatGPT. For Mistral a query is equivalent to XXXXXXXX% of a hamburger" @emollick on X 2025-07-22 22:17:23 UTC 281.1K followers, 5235 engagements
"(OpenAI reported 1/15 teaspoon in a blog post for a typical prompt for ChatGPT) Here is the report:" @emollick on X 2025-07-22 20:07:20 UTC 281.2K followers, 14.5K engagements
"McKinsey's new report on AI agents shows the same mindset I see in many firms: a focus on making small obsolete models do basic work (look at their suggested models) rather than realizing that smarter models can do higher-end work (and those models are getting cheaper & better)" @emollick on X 2025-06-16 17:48:28 UTC 281.2K followers, 422.8K engagements
"It feels significant that we crossed a line for AI music where you can continuously create entirely new AI songs in less time than it takes to listen to an AI song. Especially now that AI music has reached a point where some people enjoy it. And I assume that number will grow" @emollick on X 2025-07-17 04:06:09 UTC 281K followers, 48.8K engagements
"The problem is not just the proliferation of devices that let you record people without their knowledge but the fact that multimodal LLM let you use recordings in ways that neither law not society anticipated. Everyone has an easy way to mine hours of footage. No forgetting" @emollick on X 2025-07-23 19:17:48 UTC 281.2K followers, 27.2K engagements
"Grok X suggests that scaling still works (with the diminishing returns predicted by the scaling law) and that tool use can unlock performance gains. Kimi suggests there continues to be big opportunities from improvements in methods (Muon etc.). Lots of paths for AI right now" @emollick on X 2025-07-13 08:04:27 UTC 281K followers, 276.4K engagements
"The mitigating factor for the problem with AI benchmarks (errors saturation contamination) is that despite issues they are all still fairly heavily correlated. So if your AI does well on GPQA or MMLU or HLE it also tends to do well on other benchmarks & on vibes & real work" @emollick on X 2025-07-24 18:36:52 UTC 281.2K followers, 11.7K engagements
"We let the web rot away well before LLMs This chart shows the percentage of links from all New York Times articles that still work. Over XX% of older links are now broken. And consider that social media posts are even more ephemeral Likely only LLMs will remember that content" @emollick on X 2025-07-23 22:09:00 UTC 281.2K followers, 40.7K engagements
"This is very true. Economically valuable agents for enterprises are already here but you cant buy them off the shelf & they require actual cross-functional R&D" @emollick on X 2025-07-20 19:24:16 UTC 281.2K followers, 46.4K engagements
"I just found out that Co-Intelligence my book on AI just made the New York Times bestseller list It is very exciting and thank you to everyone who has read it. You can get it here: There's a companion site with free resources:" @emollick on X 2024-04-10 21:32:38 UTC 281.2K followers, 739.5K engagements
"LLMs are weirder than we think with all sorts of complex implications that we dont fully understand" @emollick on X 2025-07-23 02:53:25 UTC 281.2K followers, 37.9K engagements
"Since the architecture sucks now theme is coming around on X again a reminder that Hyatt and Marriott atriums are as impressive as cathedrals and pyramids but unlike cathedrals and pyramids anyone can get a drink in them or spend the night for a reasonable fee" @emollick on X 2025-07-13 10:43:08 UTC 281.2K followers, 30.4K engagements