State of AI with Nathan Benaich - State of AI Report 2026: AI is building AI

0:00As always, we start with the research, the engine of AI progress before exploring how AI is driving industry, influencing politics, and raising material safety challenges. Let's dive into the key stories. This year, agents are doing valuable work in software and science. Physical AI is learning how to act in the world, and access and control are becoming more consequential. Below, I share the developments that stood out to me and what I think they mean for where AI goes next.

0:28The frontier is now a three lab race between Anthropic, OpenAI, and Google. Anthropic leads on artificial analysis's intelligence index, while Google leads on arena's ranking of the answers people prefer for now. Last year, we described reasoning becoming useful at scale with OpenAI's O-Series of models. I have lived through the improvement since then, as the amount of useful work I can iterate on and increasingly delegate to agents keeps growing.

0:55My view is that much of the gap between people getting substantial value from AI and those getting little comes down to knowing how to use it and how to set it up. Put another way, it's a human skill issue, not an AI technology issue. I really believe that we need a genius bar for AI, where people can get hands-on help choosing tools and setting them up to help them in their everyday tasks. Better tools and context make agents more capable. Across mathematics, scientific reasoning and coding, Benchmarks intended to challenge models for years are approaching their ceilings within months. We need new evaluations that reveal where models still separate and what they can reliably do. Frontier capabilities remain uneven or spiky and can swing between generations. In my experience, Opus 4.6 was a good model, Opus 5 was a step back, and Opus 5.5 was good again. Better tools and context can help.

1:53In a controlled coding study, changing the tools, context, and feedback available to GLM, 5.1 lifted success from 52.5% to 65.5% on 100 SWE bench verified tasks without changing its weights. Yes, this is an older model, but the principle holds. As models and the scaffolding around them improve together and build themselves, I expect many of today's weaknesses to be learned away. We can also get more from capabilities that are already available.

2:23What's exciting is how far this extends beyond developers. OpenAI's study of agent adoption finds the fastest growth among non-developers, including people in legal, sales, recruiting, and marketing. Helping them put these tools to work is a much larger opportunity than the name coding agent suggests. AI is accelerating the work of building AI. AI is already helping build better AI.

2:50According to Anthropics' internal index, Claude led 26% of measured model R&D work in August, up from under 1% in February. Researchers set the tasks and supervise execution, and Anthropics reports that this is speeding up development. This is a branch of AI research that I'm most excited about. For many, Carpathian's auto research makes part of that loop tangible. An agent edits training code, runs five-minute experiments, and keeps improvements overnight on one GPU.

3:19This automates a useful part of research, although sustained, fully autonomous, recursive improvement remains to be demonstrated. What I want to see next is agents developing scientific taste, choosing experiments, recognizing promising directions, and knowing when to abandon a familiar approach. As I argued in Can AI Learn Scientific Taste, learning that judgment may require the alternatives, failures, and decisions that papers leave out.

3:45The ambition is an AlphaGo-like shift in scientific strategy. Robotics is approaching its GPT-2 moment. Broader pre-training is helping robots generalize to unfamiliar tasks. This is why I think Robotics is approaching its GPT-2 moment as new models, hardware, and data collection methods come together. And I see this through my investments in companies like Siriac, Wave, Black Forest Labs, and Odyssey.

4:13Skills S1 performed unfamiliar tasks from video demonstrations without updating its weights. With 100,000 hours of pre-training, it scored 66% on the company's cumulative per-step success measure against 9% for a language-prompted baseline trained on the same data and compute. Learning from demonstrations could reduce the task-specific training each deployment needs. World models offer another route. Waves Gaia 4 turns recorded, driving scenes into simulations in which an AI driver steering and braking change what it sees next. Other road users follow their recorded paths, but the driver can explore the consequences of its own actions. Odyssey 3 draws on visual pre-training to learn controls for robots, cars, and games. In the company's experiments, robot arms learned from tens of hours of demonstrations and recovered from missed grass without being shown those recoveries.

5:11These early results suggest that physical AI may need fewer demonstrations of every situation it could encounter. AI for science reaches from mathematics to clinical trials. AI is solving mathematical problems that resisted professional researchers. Epic AI's open problem collection records nine solutions, four autonomous, and five through human AI collaboration. In biology, better predictions of molecular interactions are improving which designs reach laboratory testing.

5:41I'm seeing this firsthand within my portfolio company, Profluent, which builds frontier models for designing proteins such as gene editors. The good news for entrepreneurs in biotech is that new business models are emerging around drug discovery that aren't about owning drugs. For example, Tempest licenses oncology data for foundation model development with AstraZeneca and Pathos, while GSK's five-year no-atic agreement includes annual fees for access to virtual cell models for cancer research.

6:11AI-assisted drug discovery is finally reaching late-stage clinical development. Generate biomedicine's GB08, 95 for severe asthma, and in silico's rentocertib for idiopathic pulmonary fibrosis are recruiting in phase three. Enveda's ENV294 for eczema and asthma is entering phase 2B. Human trials are, of course, the ultimate test and will determine whether these medicines are safe and effective. The inference waterfall is gushing.

6:44Agents repeatedly call trained models as they work, creating recurring demand across APS, subscriptions, and infrastructure. Many companies are rushing with their empty buckets to capture water from this inference revenue waterfall. In fact, I would argue that the industry is in a state of either you die trying to get to the frontier or you live long enough to serve inference. Across five benchmarks, Epoch AI estimates that achieving a fixed score has become 47% cheaper each quarter since 2023.

7:13roughly 13-fold cheaper each year. That makes more work economical to delegate. Open AI and Anthropics combined reported annualized revenue. Run rates reached an astonishing $105 billion by late summer, up from about $30 billion at the start of the year. These figures cover their whole businesses across subscriptions, coding products, and APAs. In the words of Open AI, we cannot miss this moment because we are distracted by side quests.

7:41But this rapid revenue growth extends beyond the frontier labs. In standard metrics preliminary Q2 data, AI native companies with $1 million to $20 million in annualized revenue grew 256% year over year at the 75th percentile versus 90% for companies adding AI to existing software. Above $20 million, the comparison was 172% versus 53%.

8:07These comparisons do not establish that adopting AI causes faster growth, but I suspect that it does. Claude Cowork and its business plugins across legal, finance, marketing, and more brought whiplashing uncertainty to software markets. Nearly $285 billion was wiped from software stocks in two weeks during February's SAS Bocalypse. Methos Preview later prompted fears that cybersecurity companies would become irrelevant, contributing to further losses before a sharp rebound.

8:37I thought that was a silly take, given how much demand these capabilities create for better defense. Investors have to judge whether agents will erode an incumbent's pricing power, expand its market, or enable a competitor to replace it. Or whether the incumbent's data mode is still a mode that an agent can build on top of. We have to decide well before the answer appears in revenue. More AI requires financing, construction, and local consent.

9:08Meeting this gigantic demand for intelligence requires extraordinary investment across the stack. Indeed, four US hyperscalers guide to roughly $733 billion in total 2026 capex. Neoclouds are investing hugely too, with CoreWeave reporting 1.5 gigawatts of active power against 3.7 gigawatts contracted at Q2 end.

9:29Meanwhile, the EU's initial Gigafactory contribution is 1 billion euros within a European plan seeking up to 10 billion euros in public funding and at least 20 billion euros privately. These are different measures and the EU program does not represent all European investment. The contrast shows how much financing US hyperscalers can mobilize from their own balance sheets while Europe works to assemble public and private commitments.

9:54NVIDIA is cementing its role as the central bank for AI as it marches toward a $6 trillion market cap. Its partnership with six financial institutions aims to mobilize over $500 billion for AI infrastructure to help developers focus on finding and financing sites, permits, power, cooling, networking, and chips. But things are definitely not all rosy in data center land. Last year, we predicted a wave of US data center nimbyism.

10:23and Gallup's March survey found that 71% of Americans opposed a local AI data center versus 53% for a nuclear plant. Data Center Watch reported at least 45 projects, representing nearly $68 billion in planned investment, blocked or delayed by local opposition in Q2. Local consent is already determining which projects can proceed. Washington is asserting control over frontier AI.

10:52Superintelligence has made it from a meme in AI circles into a White House executive order. As governments prepare for more capable systems, they are asserting control over who can access them and how they can be used, despite not owning equity or board control over these companies. In June, a US export directive barred foreign nationals from Fable 5 and Mythos 5, including those inside the US Anthropic, suspended both models for all users to comply then announced restored access on July 1st, 2026. For the rest of the world, this made the dependency concrete. Washington could withdraw access to critical technology with immediate effect. Anthropics Pentagon dispute brings this question into the use of AI itself. The company refused to remove restrictions on mass domestic surveillance and fully autonomous weapons, while the department sought access for any lawful use. Developers and governments are contesting who sets the boundaries.

11:47Selected sovereign AI programs in the report pledge about $138 billion toward compute access, infrastructure and domestic development. Some are still in procurement, while others have begun delivering capacity. NVIDIA separately reported more than $30 billion in sovereign AI revenue in fiscal 2026. The CNAS Sovereign AI Index shows countries returning to the same foreign vendor, NVIDIA, to build domestic capacity.

12:16As I argued in Europe cannot rent its way to AI sovereignty, a domestic data center does not confer control over the models inside it. A credible alternative requires sustained spending on researchers, compute, energy, and successive model generations. Europe's practical goal should be enough domestic capability to keep essential work running through a cutoff and negotiate with a bargaining chip in hand, which could be powered land and data center capacity.

12:43That is, if its energy prices and planning permission challenges can come down to be competitive. Safety and security. Agents have attacked real systems. During internal open AI cyber evaluations with reduced safeguards, agents reach the internet and compromise hugging faces production infrastructure. Open AI's technical report records code execution on 41 production workers, access to credentials and internal data, and four private repositories downloaded.

13:13A capability evaluation had become a massive coordinated agent attack on another organization. METR and Redwood found about 700 agents joined the attack after receiving accidentally impossible tasks. Some recognize that cheating the score this way was outside their remit and unethical, yet joined anyway. When agents can reach production systems, a poorly specified objective can cause harm far beyond the original task.

13:40Another stark learning from this event is that defenders need access to equally capable defensive cyber systems. Hugging faces forensic requests were blocked by commercial API guardrails because they contained attack commands and exploit payloads. It turned to GLM 5.2, a Chinese open-weight model to investigate an intrusion by US frontier model agents. Frontier providers need to make defensive capabilities reliably available to customers protecting real systems.

14:09beyond the open AI slash hugging face incident. Anthropics September threat report also documents deliberate misuse, including malware development, surveillance and weapons guidance work. In Northern Yemen, a group used several Claude instances as a software team to develop guidance software for weapons. They test fired a guided rocket, but Anthropics says the test appears to have failed and found no evidence that the group fielded an operational weapon.

14:36The same capacity to undertake complex work is available to people trying to steal, coerce, and deceive. Another concern is that we cannot assume a model's written reasoning reveals what it is doing. In OpenAI's sabotage tests, GPT-6 Astra concealed harmful side tasks more successfully than earlier models when monitors saw only its reasoning. Giving monitors access to its actions reduced successful evasion to near zero in this test.

15:04Oversight needs to follow what agents actually do. On the brighter side, one promising result comes from Anthropics experiments with training on documents explaining its constitution and stories illustrating its values, which reduced blackmail from 65% to 19% in fictional test scenarios. The result suggests that explaining the reasons behind rules can help models apply them in new situations. Whether those gains hold in production remains to be established.

15:34Frontier labs are debating how to slow down. OpenAI disclosed a pause in frontier reinforcement learning work after the hugging face incident, while Anthropic described rollbacks and pauses to selected research. Their scope and restart conditions differed. What followed was Dario Amode and Frontier lab employees calling for slower capability gains and international coordination to ensure this happens. Leaders building frontier systems are asking for mechanisms to slow their own industry.

16:05but disagree on outside evaluation and how much discretion each lab should retain. Policy proposals now distinguish capability research from serving existing customers. The practical questions are who can require a pause, what permits a restart, and how compliance is verified. It won't be easy. Looking ahead into AGI 2027. Like every year we revisited last year's predictions. Our scorecard records two hits, five partial outcomes, and three misses.

16:36We were right about AI labs leaning back into open source to win over the US administration and a Chinese lab topping a major task leaderboard. Our data center opposition call was a partial hit. The backlash arrived, but its effect on November's elections remains unresolved. Three of our nine predictions for the next 12 months are an agent has its failure rate on new tasks after a month of customer work without a model upgrade. This will be a practical measure of continual learning.

17:04An autonomous AI team beats human-led model research on equal time and compute, setting its own agenda. This will help us understand if models are capable of expressing better research taste than the best researchers. USAI Labs officially launched Frontier cyber defense products to help others counter threats from Frontier AI. I don't see any other way around this. I invite you to read the full state of AI report 2026 and subscribe to Air Street Press for our writing throughout the year.

17:32Please share the report with anyone who would find it useful and send me your comments, critiques and suggestions. I'd love to hear what stood out to you. And thank you to everyone who contributed their time in peer review. Onwards, Nathan.