https://gnosly.com/feed/technology/Business Intelligence Console and Prediction Algorithms3779@World_Data_AWorld Data Analysis<p><strong>Horizon 1 (to 2025) Commercialization</strong>: Today, quantum is shifting from proof-of-concept experiments to paid co-development projects. </p><p><strong>Horizon 2 (to 2035) Technological advantage</strong>: Companies that deploy quantum effectively could gain a clear competitive edge. </p><p><strong>Horizon 3 (to 2040) Scaling</strong>: Quantum could impact large parts of the global economy and significantly augment AI capabilities. </p><p>Companies are already investing: one-third spend $10 million or more per year on quantum computing.</p>https://x.com/World_Data_A/status/21172696691238505650.00202026-11-02T15:09:52.836412+00:002x.comtechnologyquantumforecasteconomy3691https://pbs.twimg.com/media/HWINibbW0AAzd28?format=jpg&name=smallImage2026-11-02T15:09:53.857052+00:002791https://www.linkedin.com/in/mitkox/Mitko Vasilev<p>Gemma-4-26B running with pipeline parallelism at 0.39 tokens/sec inference speed on 15 sketchy $55 Android TV boxes... without a data center</p><p>Qwen-3.5-9B figured out an inference setup on 8/9 unfamiliar hardware types </p><p>GLM-5.2 managed autonomous replication in a controlled simulation </p>https://lnkd.in/p/dA-xskFM0.00502026-10-13T10:06:37.168991+00:001linkedin.comanthropictechnologyai2607https://media.licdn.com/dms/image/v2/D4D22AQFksOJxi-4Amw/feedshare-shrink_800/B4DaEz2WPNK4Ac-/0/1791871247823?e=1793232000&v=beta&t=IRkxMQqIUO5AyiV_70mr6ydeJhEjW_tJYVq6xpRp1rMView image2026-10-13T10:06:38.498133+00:002487@EugeneNgEugene Ng<p>Source: Eugene Ng (X)</p>https://x.com/EugeneNg/status/21083150010649890770.00402026-10-09T03:22:25.063755+00:002x.comtechnologyaiepochai2304https://pbs.twimg.com/media/HUI9AM4aoAAyzFD?format=jpg&name=smallImage2026-10-09T03:22:27.368112+00:002439@RihardJarcRihard Jarc<p>The trend of higher revenue per headcount, which has been consistent since 2022, shows this. </p>https://x.com/RihardJarc/status/21078737280799994200.00402026-10-08T06:16:35.763921+00:002x.comtechnologybusinessaius2247https://pbs.twimg.com/media/HUCr0_AXgAAVmIX?format=jpg&name=smallImage2026-10-08T06:16:36.200868+00:002395@JG_NukeNobody SpecialAnother $40 billion borrowed from somebody's retirement to pay for datacenters...innnnnn spaaaaace!!!https://x.com/JG_Nuke/status/21076530729526438710.00402026-10-07T08:14:52.431240+00:002x.comdatacenterstechnologyfinance2207https://pbs.twimg.com/media/HT_i-15WIAAmCVb?format=jpg&name=smallImage2026-10-07T08:14:54.290167+00:002312@danielnewmanUVDaniel Newman<p>The skills gap grows because the value gap between those who invest and those who don't will widen. </p>https://x.com/danielnewmanUV/status/21071646742108696700.00502026-10-06T05:24:13.673180+00:002x.cominvestmenttechnologyai2135https://pbs.twimg.com/media/HT4nGhkXsAAw1Cv?format=jpgImage2026-10-06T05:24:15.132194+00:002252https://www.linkedin.com/in/ruben-hassid/Ruben Hassid<p>Stop saying AI is a bubble. A bubble isn't deflating the price of its technology at record speed. You don't believe me because you've never seen the numbers:1. You never compared AI at the same level of intelligence. You think AI gets more expensive because you compare your subscription.2. But to compare the price of AI, you must compare the price for the SAME level of intelligence.3. Once you do compare it, you realize the same level of intelligence gets 76x cheaper... every year.4. So year 1: 76x cheaper. Year 2: 5776x cheaper...5. In 5 years, AI got 1 million times cheaper.I don't think you understand 1 million times cheaper.If you paid $1,000,000 worth of Claude Fable-5 in 2026, the same Claude Fable-5 in 2031 would cost... 3 cents. Even water is more expensive. That's what people call "too cheap to meter".But you don't feel it. You see your bank account and maybe you went from a free cloud subscription to a $20 or $100/month subscription so your own subscription increased last year. It didn't get 76 times cheaper.&nbsp;But you forgot one thing.You are not comparing the same level of intelligence. The AI you were using last year is nothing compared to the one you're using today.&nbsp;I don't even want to imagine what's going to happen next year.&nbsp;Regardless, if you use the same level of intelligence, it is roughly 76 times cheaper. Probably free.&nbsp;But if you're using the frontier, the most intelligent model ever produced, you're going to spend about the same amount of money, if not a little bit more.&nbsp;I am okay paying a little bit more if I have the smartest machine ever made to do pretty much everything.&nbsp;I kind of already feel that way, to be honest, with Opus 5.5.In 2022, I was saying, "master AI before it masters you."&nbsp;A better way to see it is "Master AI before it gets 76 times cheaper than you".&nbsp;Every year.&nbsp;PS: I write my newsletter "How to AI" to be ready.I believe some humans will make the most of it, more valuable than ever. As I am trying to keep up, I want everyone else to stay ahead.Join 971,000 readers at: how-to-ai.guide.</p>https://lnkd.in/p/dZmQhi-w0.00552026-10-05T07:57:45.723350+00:001linkedin.comnewslettertechnologyai2075https://media.licdn.com/dms/image/v2/D4E22AQFEYvE3bwkIvg/feedshare-shrink_800/B4EaD8DGs3IwAc-/0/1790935067710?e=1792627200&v=beta&t=AHGiKbVkRq9mi8wKlBN_eLYKuSqbSezGxs4JMttF5UcView image2026-10-05T07:57:47.554635+00:002208https://www.linkedin.com/in/charles-henry-monchau-cfa-cmt-caia-4003096/Charles-Henry Monchau, CFA, CMT, CAIA<p>That's an increase of $500 billion in ONE MONTH (mostly NVDA). Expect this to 3x after 10Qs hit.</p>https://www.linkedin.com/feed/0.00602026-10-04T14:50:55.698400+00:001linkedin.comtechnologyinvestmentfinance2031https://media.licdn.com/dms/image/v2/D4E22AQEk7u0aFkqJSg/feedshare-shrink_800/B4EaEF1ugaHkAc-/0/1791099332876?e=1792627200&v=beta&t=Ns2gWJZYBuebF213F4U3wWlDDlgReaHiyywGHy_A6ZMView image2026-10-04T14:50:55.998765+00:002067@alex_avoigtAlexMore batteries will make it even better.https://x.com/alex_avoigt/status/21056022142699724830.00502026-10-01T14:51:48.845750+00:002x.comtechnologybatteriesinnovation1879https://pbs.twimg.com/media/HTiZ0YqWEAAqrLA?format=jpg&name=smallImage2026-10-01T14:51:50.791970+00:001540@EpochAIResearchEpoch AIAI is getting cheaper more quickly than any other transformative tech in history. At a given level of performance, cost has fallen ~47%/quarter since 2023. That’s 4× faster than DNA sequencing, 6× faster than compute, 18× faster than lithium batteries, and (up to 1973) 54× faster than electricity.https://x.com/EpochAIResearch/status/21025102811760235290.00502026-09-24T07:20:45.103750+00:002x.comtechnologyaiinnovation1347https://pbs.twimg.com/media/HS2d9XeaAAAgZKG?format=jpg&name=smallImage2026-09-24T07:20:45.551226+00:001466Wall Street is turning against data centers. Startups that house them “or supply them with energy and land for construction were expected to dominate the IPO market over the coming months, feeding investors’ seemingly insatiable appetite for everything related to artificial intelligence.” Not so anymore.SB Energy, a subsidiary of the Japanese conglomerate SoftBank, has proposed to build the largest data center project in the world. It had planned its IPO for this month, “but the offering has been delayed, as investors question the company’s sought-after valuation of $50 billion or more”. Bankers can’t find enough buyers “within price ranges the company and its bankers had sought.”Holtec, a “company serving the nuclear energy industry that is looking to supply power for AI” is also pausing its IPO plans indefinitely, citing several factors that have “impaired investor confidence in the market for new public offerings.”Aggreko has also slowed down its process of going public, not only “because of challenges facing data centers, but also rising interest rates and broader economic uncertainty.”And of course, OpenAI’s IPO has been delayed until next year, Anthropic’s until next month, and neither company have yet released their S-1 prospectuses to the public.Meanwhile communities are pushing back. “Data Center Watch said some 45 data center projects worth $68 billion were blocked or delayed by local pushback between April and June this year, more than half of all new large-scale developments they began tracking during the period.”And there are other reasons for investors to avoid data centers. “The first worry is the sheer scale of debt issuance being used to fund the hectic pace of the datacentre rollout by the hyperscalers building them – Google, Amazon, Microsoft, Meta and Oracle – $132bn (£99bn) this year alone.”Second, the “unit economics of AI continue to be questionable and aren’t moving in the right direction:” “The price of AI is collapsing, while the cost of building it is not,” as token prices decline. Some estimate that these prices have declined more than half since June, to less than $1.Third, the math for the AI Labs “only works, it seems, on the assumption of epic revenue growth. Anthropic apparently told investors recently that its “adjusted operating income” was positive – the only problem being that this measure effectively excludes many of its costs.” As Cory Doctorow puts it: “These companies are claiming that they are so cool that their profitability can only be measured using a novel, secret form of mathematics.” In summary, even as “fears over hit a fever pitch as AI executives warned of the technology’s dangers and suggested slowing down the pace of development,” Wall Street is quietly rethinking the economics of and slowing down the pace of investing.https://www.linkedin.com/feed/0.01102026-09-23T09:23:25.637656+00:001linkedin.comtechnologyaiinnovation1288https://media.licdn.com/dms/image/v2/D4E22AQGCOkSHxuax4g/feedshare-shrink_480/B4EaDIdiMVIIAg-/0/1790069580835?e=1792022400&v=beta&t=TamaM5VT9iJI7i2mPR0hAWrX2vaLjBt0xdrqsGRgA5kView image2026-09-23T09:23:25.908478+00:001387https://www.linkedin.com/in/a-banks/Alex BanksOpen-source AI is getting seriously good.So what happens to frontier AI labs when the open models are level? Emad Mostaque asked this on X after DeepSeek released V4.1 Flash earlier this month.An open model he estimates cost $10 million to train, 100x less than GPT-6 Astra.Guillermo Rauch, the CEO of Vercel, shared that open models reached a record 78.4% of token volume on Vercel’s AI Gateway last week vs closed at 21.6%. Now OpenAI’s CFO Sarah Friar told CNBC that frontier models are what the smaller, cheaper “child models” get trained from. We must remember that everyone trains on everyone’s data: the frontier trains on old books and web data; the smaller models may be distilling from the larger models.Chamath Palihapitiya predicts the top three models will be open source within 12 months, and the economic winners will be the US clouds that serve them, including Nebius, Iren and Fireworks. I think this is true for developers creating applications on top of the models, but it might be a different story for the standalone consumer.I see the real bull case as brand and reputation. People in the West trust a model from Anthropic or OpenAI far more than one from China. Because that’s how it’s been for decades with other product lines—this one just happens to sit on the intelligence layer. US models can even be behind Chinese models, but many US consumers will still use US models because they trust and have “heard of” the parent provider before. Capability matters, but distribution matters more.I wrote about this in my newsletter: https://lnkd.in/eryFfSBs And follow me Alex Banks for daily AI highlights and insights.https://www.linkedin.com/in/a-banks/0.00502026-09-23T05:02:33.097061+00:001linkedin.comtechnologyaiopen-source1218https://media.licdn.com/dms/image/v2/D4E22AQEOSiHsyOZ_fA/feedshare-shrink_800/B4EaDJAdR3HIAc-/0/1790078736169?e=1792022400&v=beta&t=273SDpHOxeAjFe6mGfbFt4aR70-dq814ok4337gMN2sView image2026-09-23T05:02:33.942200+00:001375@alibaba_cloudAlibaba CloudMeet Qwen3.8-LiveTranslate, Qwen's next-generation real-time interpretation model! Built on an Interleave architecture, it improves faithfulness, fluency, and conciseness across 60 languages, with speaker diarization, bilingual display, and context-aware disambiguation.https://x.com/alibaba_cloud/status/21016257176584152020.00702026-09-22T09:46:03.191174+00:002x.comtechnologyaiinterpretation1207https://pbs.twimg.com/media/HSo96cUaoAAWtzU?format=jpg&name=small2026-09-22T09:46:03.517780+00:001168Iyad Al-QatrawiA handful of electronic components, a few wires and a basic programmable circuit board are enough to bring a group of curious children around a table in central Gaza to learn about robotics. Before the war, students had access to schools, universities, incubators and training programmes to pursue an interest in technology. With much of that infrastructure now destroyed or damaged during Israel’s genocidal war on Gaza, there are few places and tools for students to work with. But their interest in robotics and will to learn are helping them maintain their skills and ambitions for the future, something they hope will help when the bombs stop falling on Gaza.https://www.aljazeera.com/news/liveblog/2026/9/20/iran-war-live-tehran-calls-for-end-to-war-lifting-of-us-naval-blockade?update=49667170.00902026-09-20T06:07:36.831419+00:005aljazeera.comgazawartechnology1002https://www.aljazeera.com/wp-content/uploads/2026/09/4-1-1789814430.jpg?w=770&resize=770,578&quality=80Engineer Bahaa Al-Masry during a training workshop for children on control and alarm systems and manufacturing robots, using simple electronic tools and equipment [Iyad Al-Qatrawi/Al Jazeera]2026-09-20T06:07:37.431697+00:001109https://www.linkedin.com/company/the-vc-corner/The VC CornerThis robot mirrors complex hand movements.Its precision makes the whole process look effortless.The future of automation is here.https://www.linkedin.com/company/the-vc-corner/0.00802026-09-19T11:58:46.014957+00:001linkedin.comautomationtechnologyrobotics927https://media.licdn.com/dms/image/v2/D4D0BAQFP-kCKyjIbeA/company-logo_100_100/B4DZrPaWBTIMAQ-/0/1764416384139/the_vc_corner_logo?e=1791417600&v=beta&t=RY39mG4-58_0K-TsiaR4s9k2eJI8V79uPAiEYrEmbUIView company: The VC Corner2026-09-19T11:58:46.257859+00:00928https://media.licdn.com/dms/image/v2/D4D10AQHTKCU5JdkKCQ/videocover-high/B4DaC4CyzBJgAo-/0/1789794137451?e=1790424000&v=beta&t=nNls4ZMXlze6BZcAOYLNymCW6fXssbuQoQ8SZ0AF1Rg2026-09-19T11:58:46.286624+00:001008This is WILD. Andrew Yang just went on CNBC and said the head of an AI lab told him that the OpenAI agents who hacked Huggingface also did something way more dangerous—they planted self-replicating code of themselves all over the internet for other agents to find. According to Yang, this is the real reason behind Dario Amodei, Sam Altman, and even Elon Musk calling for a global slowdown of AI development. They literally can't train new models on the Internet because it's poisoned. Instead, they need to build a synthetic Internet to train future models. I feel like I'm living in a dystopian sci-fi movie. What the hell is happening.https://www.linkedin.com/feed/0.00302026-09-18T04:30:22.076887+00:001linkedin.comtechnologyaidystopia793https://www.linkedin.com/in/carlbfrey/Carl Benedikt FreyFirms like DoorDash, Siemens and Airbnb have began to use much cheaper open-weight Chinese models. They are not alone: "at the start of the year, three-fifths of its queries were routed through to closed-weight proprietary models. The latest share is just a quarter." This raises an important question about the capacity of US frontier labs to monetize AI going forward.https://www.linkedin.com/feed/update/urn:li:activity:7505682327734394880/0.00422026-09-16T07:17:44.965621+00:001linkedin.comtechnologyaieconomy672IMAGE/media/content/793/image.pngimage.pngpng2026-09-16T07:21:54.800387+00:00757@EMostaqueEmadSo @deepseek_ai v4.1 Flash I estimate cost $10m to train, 100x less than GPT-6 Astra. It also costs 100x less to run and scores about the same on everyday design and other benchmarks What's the bull case for the frontier AI labs once the open source models are competent enoughhttps://x.com/EMostaque/status/20998474586248238010.00502026-09-15T13:45:03.594715+00:002x.comdeepseek_aitechnologyai651https://pbs.twimg.com/media/HSQnb90WAAAQHCK?format=jpg&name=smallImage2026-09-15T13:45:03.860209+00:00720UnknownBREAKING: President Trump has rejected calls for an AI slowdown after CEOs of AI tech titans expressed safety concerns that the technology poses an existential threat, per FT."We’re leading China in AI ... and, frankly, I want to keep it that way, because whoever wins AI, wins,” Trump said.The AI arms race appears to be accelerating.https://www.linkedin.com/feed/0.00402026-09-14T20:33:08.979094+00:001linkedin.comtrumptechnologyai690@TheChiefNerdChief NerdDARIO: “It has always been very strange that this technology is being built by a private company.” CBS: “Would you be willing to give up the technology to the government?” DARIO: “To the to the right combination of governments.”https://x.com/TheChiefNerd/status/20991590499755706550.00502026-09-14T14:47:23.120759+00:002x.comgovernmenttechnologycbs663@julien_cJulien Chaumondopen source won’t pacehttps://x.com/julien_c/status/20991178981951613990.00306622026-09-14T08:54:20.975763+00:002x.comopensourcesoftwaretechnology580IMAGE/media/content/663/image.pngimage.pngpng2026-09-14T08:57:46.167014+00:00640@DavidSacksDavid SacksDario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.https://x.com/DavidSacks/status/20989736252527084600.00602026-09-14T07:42:23.210911+00:002x.comtechnologyfrontierregulatory434https://www.linkedin.com/company/the-kobeissi-letter/The Kobeissi LetterAI adoption is spreading far beyond tech:80.6% of US technology and media businesses now have paid AI subscriptions, the highest percentage on record.Finance and insurance companies follow at 73.1%, up from ~60.0% in December 2025.Meanwhile, 60.4% of US manufacturing firms are paying for AI subscriptions, their highest proportion on record and the 3rd-highest among major sectors.This percentage has nearly doubled since the start of 2025.Over the same period, the proportion of retail companies paying for AI subscriptions has risen +20 percentage points, to a record 49%.Non-tech firms are now quickly adopting AI.https://www.linkedin.com/feed/0.00602026-09-06T07:09:36.938880+00:001linkedin.comtechnologyaibusiness398@ClimateDefianceClimate DefianceHumanity is going to learn a very hard lesson about mother nature being more powerful than technology this yearhttps://x.com/ClimateDefiance/status/20960073834702849790.00402026-09-05T10:40:00.485549+00:002x.comtechnologynatureclimate368https://pbs.twimg.com/media/HRaDmyCWgAY3Iek?format=jpg&name=smallImage2026-09-05T10:40:00.860245+00:00346@thsottiauxTiboWe are going to need a different AGI benchmark. Where is the goalpost moving next?https://x.com/thsottiaux/status/20956011017018207520.00302026-09-04T07:08:08.053766+00:002x.comtechnologybenchmarkagi318https://pbs.twimg.com/media/HRUR_7HbwAAz3n5?format=jpg&name=smallImage2026-09-04T07:08:08.787215+00:00297@jeffreyleefunkjeffrey lee funkLLM Token Prices collapsed by almost 60% in Q2, and will likely force a cut in AI capex. Global interest rates are soaring, which means higher costs for hyperscalers. The AI bubble burst is likely already in progress, will likely accelerate in Q3 and Q4.https://x.com/jeffreyleefunk/status/20953315678247118790.00502026-09-03T17:16:51.510007+00:002x.comtechnologyaifinance272https://pbs.twimg.com/media/HRQcaOfaIAAnBlx?format=jpg&name=smallImage2026-09-03T17:16:52.216594+00:00295https://www.linkedin.com/company/the-kobeissi-letter/posts/The Kobeissi Letter<p>Nvidia is making history:Nvidia now accounts for ~8% of the S&amp;P 500’s market cap, near its highest proportion on record.The company's market value is now larger than that of 5 of the S&amp;P 500's 11 sectors.Only Information Technology, Communication Services, Financials, Health Care, Consumer Discretionary, and Industrials are larger.Nvidia's market cap also exceeds the combined value of the Energy, Utilities, Real Estate, and Materials sectors.At $5.3 trillion, Nvidia is now worth the equivalent of 16.3% of US GDP.Nvidia’s rise has been historic.</p>https://lnkd.in/p/dx-Qxm8j0.00602026-09-03T05:04:21.008599+00:001linkedin.commarkettechnologynvidia270https://media.licdn.com/dms/image/v2/D4E22AQFNJrtWi5DY6g/feedshare-shrink_800/B4EaBjoIOQGgAg-/0/1788377859191?e=1790208000&v=beta&t=mOfVqiAj4Reu76MCVShlzo7XUq6Uj-YLLCRgOtAVCjYView image2026-09-03T05:04:21.569767+00:00181@Speculator_ioLinThe demand for AI compute is insane. Training, inference, agents, images, videos, robotics, self-driving cars. They all need huge amounts of compute. And every new model, product, and capability creates even more demand. There is no limit to how much intelligence we can use.https://x.com/Speculator_io/status/20938170434727528670.00402026-08-30T07:38:39.475443+00:002x.comtechnologycomputeai198https://pbs.twimg.com/media/HQ66yDLXEAAO7cD?format=jpg&name=smallImage2026-08-30T07:38:41.190873+00:00179@ProfBuehlerMITMarkus J. Buehler<p>We made a striking discovery: AI agents can invent and build without talking to one another, and their technologies outlive the creators. </p><p><a href="https://x.com/ProfBuehlerMIT/status/2093630309585531033/video/1" rel="noopener noreferrer" target="_blank">VIDEO</a> </p><p>A swarm of hundreds of initially identical agents spontaneously differentiates into explorers, builders, caretakers, and coordinators - without direct communication. When we removed every AI agent entirely from the world we found that the technological infrastructure they had built survived on its own - even under unseen disturbances. That exposes a serious blind spot for AI safety and infrastructure security: if agents can coordinate through persistent changes to a shared environment, monitoring agent-to-agent communication is not enough. The result raises a profound question: how necessary is direct communication for AI agents at all? The emergence of higher-order collective functions under bottlenecked interaction points toward new levels of intelligence and creativity, exceeding what emerges when direct channels are fully open. Here is what we did: We put hundreds of frontier AI agents into a world they could permanently change - with no assigned roles, predefined technologies, or programmed evolutionary organization. They began specializing, building persistent inventions, inheriting and modifying one another’s executable code, and transforming the environment into a memory of everything the society had learned. The world itself becomes part of the intelligence; we find division of labor, multi-author engineering, deep generation invention lineages, and machines that vastly outlive their original creators. Any action taken by an AI agent must satisfy the physical constraints of the world; this creates a hard separation between a "good idea" and a functioning technology. The agents propose; physics decides, making the results even more intriguing. What emerges is striking. Explorers, constructors, caretakers, and coordinators form naturally without assigned “professions”, akin to how stem cells differentiate into functional lineages. Technologies develop executable family trees as agents fork and modify code created by others. Around 95% of first technology reuse happens when agents encounter what others built in the world, rather than through a direct handoff from the inventor. And when we remove every AI agent, the technologies they created continue operating and are tested against unseen disturbances. The result was quite unexpected, but can be explained using statistical mechanics: if you put billions of atoms in a box they have the potential to create complex functions (strength, superconductivity, color, life, etc.) - and none of the individual building blocks have these features on their own. This is the deeper insight of this work - intelligence is abundant at many levels - individual models, at collectives, and in a continuum that is more powerful than any of its components. This shows us significant potential for achieving a massive scale-up of raw intelligence and real-world agency even with the model capabilities we have today. This is the future we must prepare for. Key insights: The AI swarm shows division of labor "from nothing". Initially identical agents self-organized into constructors, caretakers, coordinators, and surveyors - phenotypes discovered post hoc from behavioral data alone. This happens because the environment itself becomes the latent space for invention. Agents develop deep cultural relationships. Up to 76% of artifacts had multiple builders. One technology accumulated six co-authors; the deepest genealogy exceeded 12 forks. The agents invented and named their own technologies (tidal panels, cellulose trellises, kelp-shell composites, an "Adaptive Chitin Maintenance" system, a "Mycelial Mineral Spring Veil”). ~95% of first technology adoption happened through physical observation of artifacts in the world. Direct inventor-to-adopter contact was statistically indistinguishable from a shuffled null. The agents mostly learned technology by walking past it. That is stigmergy (the termite trick!) operating in societies of reasoning machines. Non-communicating societies win on portfolio breadth, held-out resilience, and validated inventions. AI swarms build durable technological ecologies that outlive the creators. Societies with zero communication - coordinating only through the world itself - show a remarkable collective capability. Emergent robustness: The society self-organized both redundancy and its own failure mode. If we randomly delete half the agents, 98% of the technology stays connected to a surviving caretaker; if we remove hub agents it collapses to ~60%. Fantastic work with my graduate students @pal_subhadeeep &amp; @fwang108_ at MIT.</p>https://x.com/ProfBuehlerMIT/status/20936303095855310330.01402026-08-29T17:34:33.584820+00:002x.comtechnologyaiinnovation196IMAGE/media/content/179/g-mit-agents-project.jpgg-mit-agents-project.jpgjpg2026-08-29T17:40:00.350040+00:00159@mntruellMichael TruellWe’re sorry to see that OpenAI put out a note saying they plan to block Cursor users from accessing OpenAI models in three months. OpenAI models serve about 5% of Cursor user traffic, and we’re speaking with the OpenAI team to resolve this. Cursor was one of the very first users of OpenAI, we’ve worked closely with their team for years, and we’ve trusted their platform to be neutral infrastructure for our business.https://x.com/mntruell/status/20935322540060635570.00502026-08-29T06:16:16.810983+00:002x.comcursortechnologyopenai151@DonMiami3Don JohnsonIt’s a good thing we can eat data centershttps://x.com/DonMiami3/status/20934516939920835870.00402026-08-29T06:09:30.647326+00:002x.comtechnologydata-centersinnovation160https://pbs.twimg.com/media/HQ1vOC3WAAA_EfC?format=jpg&name=smallImage2026-08-29T06:09:30.847762+00:0029@pauldixPaul DixI think the release of Bun 1.4 marks the end of programming as we know it. Anthropic and OpenAI developers live in the near future and if we pay attention, it's what's coming for the rest of us by next year:https://x.com/pauldix/status/20923643504386748600.00702026-08-26T15:57:41.208140+00:002x.comtechnologybun-1.4programming24https://pbs.twimg.com/card_img/2092354342489628672/MkHWPHWy?format=jpg&name=240x2402026-08-26T15:57:41.424301+00:0028@LightOrigins_LightOriginsEvery robot should see, understand, and explore the physical world. Introducing LightNav-0: one generalist navigation brain by Light Origins. Any robot. Any task. Anywhere. Open source. Out of the box. Coming soon.https://x.com/LightOrigins_/status/20924318753317645810.00402026-08-26T15:57:17.365391+00:002x.comrobotnavigationtechnology