Air Street Press

Description:Black Forest Labs has launched FLUX 3, a multimodal foundation model that learns jointly from images, video and audio within a single architecture, and mimic robotics has released FLUX-mimic, a video-action model built on that backbone and being tested on real assembly work in Audi's Production Lab. Nathan Benaich of Air Street Capital, an investor in Black Forest Labs, reads his Air Street Press essay on why a model trained to predict how scenes evolve turns out to be a usable robot controller. Covers the FLUX 3 preference results against Runway Gen-4.5, Grok Imagine Video, Kling v3 Pro, Seedance 2.0 and Gemini Omni Flash; how a lightweight action decoder reads the video prediction path without ever generating video; the frozen-backbone ablation against π0.5; the 101-millisecond system reaction time on a single RTX 5090; and what Air Street's own robotics deal flow says about where the constraint really sits.Chapters (estimated at ~150 wpm, slide proportionally against final audio):0:00 A robot arm in Audi's Production Lab0:40 What Black Forest Labs and mimic released1:20 Early access and open weights1:50 The preference-test results2:40 What a model must represent to predict video3:30 Reading actions off the video path4:20 The frozen-backbone ablation5:00 101 milliseconds, and the Audi tasks5:50 What we see in robotics deal flow6:30 The road to physical intelligenceLinks: FLUX 3 · FLUX-mimic (BFL) · FLUX-mimic (mimic) · Odyssey Series B · BFL Series B ·

Raia Hadsell, VP of Research at Google DeepMind, makes the case that intelligence is more than language: the same recipe that learns the patterns of text can learn the patterns of any complex system. She walks through DiffusionGemma and text diffusion, the Genie world models, and DeepMind's robotics stack, where world models now generate training data you cannot tell from the real thing. Recorded at RAAIS 2026.Timestamps0:00 Introduction (Nathan Benaich)0:35 From philosophy to DeepMind: the frontiers of intelligence2:37 The twenty-year lesson: one recipe for complex systems4:34 DiffusionGemma and the Gemma 4 open models5:32 How text diffusion works7:34 Speed, self-correction, and the sudoku test10:51 World models: better agents need better worlds12:56 Genie 1 to Genie 315:01 Genie 3 demos: typing a world into being18:20 World models for education19:38 Grounding Genie in Street View20:41 Robotics: a brain and a spine23:14 Gemini Robotics-ER 1.6 and Boston Dynamics' Spot24:42 The vision-language-action model25:45 The data bottleneck and closing the loop27:24 Beyond language: the domains still to crack

At RAAIS 2026, Nathan Benaich sits down with Hadrien Canter, co-founder and CEO of Alta Ares, the air-defense company building AI-guided interceptors to shoot down cheap drones and cruise missiles. They get into why Europe has lost air superiority for the first time in modern history, what the data loop looks like when you run it on a freezing front line instead of a laptop, why "quantity is the quality" in the industrialization race, and how the talent pool is shifting toward European defense. Recorded live at RAAIS 2026 in London.Chapters00:00 — Introducing Hadrien Canter and Alta Ares00:47 — 2022 in Ukraine, and how Europe lost air superiority03:23 — The Series A and the Airbus partnership05:13 — The data loop, edge AI, and the three phases of a mission08:22 — What no simulation can reproduce10:09 — Hiring for the mission; defense as the precondition for peace12:36 — Open research questions and the human in the loop14:22 — How the adversary uses AI: evasive Shaheds, drone mesh, China17:27 — Two interceptors, and why quantity is the quality20:50 — Iron Dome math, budgets, and peace-time vs war-time23:41 — Q&A: keeping pace with a fast-changing front26:31 — Q&A: talent and the shift toward European defense29:30 — Q&A: re-arming without permanent war32:56 — Freedom doesn't come for free

World models are the bet that AI should learn the world by watching it and acting in it, not just by reading about it. At RAAIS 2026, Odyssey co-founder and CTO Jeff Hawke makes the case: a world model is a neural simulator - an interactive stream of pixels that runs in real time, models physics, and answers back.He walks through Odyssey's four research fronts - streaming interactive pixels (Odyssey-2), joint audio and video (Starchild-1), shared multiplayer worlds (Agora-1, demoed live as a fully generated game of GoldenEye), and PROWL, which sends a reinforcement-learning agent to find and fix a world model's own failures - and argues the field is at its GPT-2 moment: promising, but pre-ChatGPT, with the GPT-3-style commercial unlock still ahead.Recorded at the 10th Research and Applied AI Summit (RAAIS), London, June 2026.Timestamps00:00 Intro: Nathan on Odyssey and world models01:05 Jeff Hawke: from self-driving to world models01:40 The bet — a missing form of intelligence02:40 Why world models suddenly matter (the late-2025 flip)03:16 What a world model actually is (and isn't)04:45 The neural simulator06:34 Two principles: end-to-end learning and generality07:19 The "GPT-3 of world models" and four research themes08:46 Odyssey-2: streaming, interactive pixels10:33 Starchild-1: generating audio and video together13:03 Agora-1: multiplayer world models13:57 Live demo: the room plays GoldenEye16:20 PROWL: improving the model by breaking it18:39 Where Odyssey goes next19:55 Still the GPT-2 era21:30 Q&A: physics limits, safety, compute cost, merging with LLMs

Ted Moskovitz leads the Science of Scaling team at Anthropic, the group that works out how to turn compute into smarter models. In this RAAIS 2026 fireside with Air Street Capital's Nathan Benaich, he argues that frontier scaling has become an empirical science - a discipline for cutting uncertainty before spending the compute, not just buying more of it.They get into the honest measure of AI acceleration (it's the counterfactual, not the benchmark), why a bigger model can be cheaper than splitting a task across small ones, whether a model can have research taste, and why safety and capability turn out to be the same axis. Plus the highest-leverage AI work to do in 2026, and why Anthropic's London office no longer feels like a satellite.Recorded live at RAAIS 2026 in London.Timestamp:00:00 - Meet Ted Moskovitz and the Science of Scaling team00:45 - What "the science of scaling" actually means01:18 - Why scaling is a science, not an art02:55 - Big labs vs the new "neo labs"04:47 - How a research finding reaches the product06:44 - What neuroscience carries over to AI (and what doesn't)09:12 - "When AI builds itself" and the real measure of acceleration10:33 - Trust, bypass mode, and the latest model jumps11:42 - One big model vs many small ones13:13 - Can a model have research taste?15:39 - How safety research makes products better17:36 - Emergent misalignment and the alignment race19:14 - The highest-leverage AI work in 202620:21 - Inside Anthropic's London office21:34 - Audience Q&A

Vivek Natarajan, Research Lead for AI, science and medicine at Google DeepMind, on porting the self-play and search recipe behind AlphaGo into scientific and clinical reasoning. He walks through the AI co-scientist, which generates and debates hypotheses (one matched a decade of lab work in two days), and AMIE, a diagnostic dialogue system trained in simulation. Recorded at RAAIS 2026.Chapters:0:00 Welcome and introducing the AI co-scientist1:41 Origins: Med-PaLM and the leap to hypothesis generation5:10 System 1 versus System 2 thinking6:34 Borrowing from AlphaGo: self-play and search8:02 Generate, debate, evolve, and tournaments11:47 Testing in real labs: Imperial College and antimicrobial resistance13:29 Ten years in two days: Penadés reacts15:44 More breakthroughs: leukemia, liver fibrosis and vorinostat18:44 Plant immunity and protein design20:09 Democratizing medicine: from Med-PaLM benchmarks21:28 AMIE and the value of experience23:12 Diagnosis, empathy and augmenting doctors25:19 Real patients: the Beth Israel feasibility study27:21 The co-clinician and the new triad of care28:31 Audience Q&A

At RAAIS 2026, Google DeepMind's Roberta Raileanu lays out a recipe for superhuman scientific discovery: AI systems that make groundbreaking discoveries across domains faster than people can. She walks through three ingredients - reinforcement learning to discover solutions where progress can be measured, open-ended divergent search to find new problems rather than climb known ones, and meta-learning to speed up discovery on problems no one has posed yet. The through-line: we can search for anything we can measure, but we still cannot measure what makes a discovery good. The bottleneck isn't the search. It's the signal.Chapters:00:00 - Introduction00:51 - Defining superhuman scientific discovery01:40 - The state of play: real progress, real plateau06:58 - Ingredient one: discovery as reinforcement learning (Move 37, MLGym)12:51 - Ingredient two: open-ended search and why greatness cannot be planned18:30 - Rainbow Teaming: quality-diversity in practice21:07 - Ingredient three: meta-learning the process of discovery (DiscoBench)25:06 - The recipe, and the missing signal

Angelos Perivolaropoulos, a research engineer at ElevenLabs, on turning GPU scarcity into an inference-engineering problem: how to serve far more users on the same hardware, from batching to frontier architecture changes. Recorded at RAAIS 2026.00:00 Introduction: ElevenLabs and the GPU squeeze00:38 The question: how to scale when you can't add capacity01:11 About Angelos: Scribe, speech-to-text and text-to-speech01:56 GPU scarcity meets exponential demand02:44 What a token actually costs: compute vs memory bandwidth03:38 Prefill, decode and the KV cache05:53 Batching and continuous batching (1 → 15 users/GPU)08:37 FP8 quantization and quantize-aware training (→ 20)11:29 Speculative decoding and multi-token prediction (→ 28)15:13 Compressing the KV cache: TurboQuant and distillation (→ 70)17:27 Frontier architectures: MLA, linear attention, state-space (→ 140)20:39 Trade-offs: nothing is free22:03 Q&A: papers vs production, token subsidies, TTS evals

The fifth State of AI Compute Index, in collaboration with Zeta Alpha. After a soft 2025, open AI research citations rebounded in 2026 - and NVIDIA still appears in ~91% of them. But the bigger story has moved off the page: Hopper is now the live installed base, Blackwell is mostly still pipeline, and frontier labs have started buying compute by the gigawatt. Nathan walks through what changed, what didn't, and why "GPU count" is becoming the wrong question.Read the full piece and explore the live charts: https://www.stateof.ai/computeChapters(00:00) What's new in v5 - citations, infrastructure, and gigawatts(01:25) The breather was short: 2025 was a pause, not a rollover(03:15) NVIDIA at ~91%, and the challengers - AMD, Huawei, Apple, TPU(05:05) Inside NVIDIA: the handover from A100 to Hopper to Blackwell(06:55) Startup silicon fragments - Groq, Cerebras, and the NVIDIA deal(08:20) Hopper is the installed base: 460k deployed GPUs(09:50) Blackwell is mostly pipeline: 80% still announced(11:00) The demand side, measured in gigawatts(12:15) Looking ahead, and why a GPU order isn't a clusterLinks:Full index and charts: https://www.stateof.ai/computeState of AI Report: https://www.stateof.aiAir Street Press: https://press.airstreet.comIf you found this useful, rate State of AI with Nathan Benaich five stars and share it with someone building in AI infrastructure - it genuinely helps.

Air Street Capital backs Stark, the German multi-domain defense company, in its €500M led by Founders Fund and Sequoia. In this episode, we discuss why cheap, software-defined unmanned systems in the air and at sea are the decisive lesson of Ukraine, and why we think co-founder and CEO Uwe Horstmann - a Project A GP and Bundeswehr reservist - is building the German neoprime Europe needs. Round led by Sequoia and Founders Fund, with the NATO Innovation Fund, Project A, and Air Street.Links: stark-defence.com · full post at press.airstreet.com · YouTube version

Nikolay Donets, Head of Machine Learning Engineering at Revolut, on what it takes to run AI across more than 70 million customers, 200+ products, and 40+ countries - and why the hard part is no longer the model but the control plane around it: one gateway, a use-case-based governance layer, fallback chains, cost controls, and mandatory human oversight. Recorded at RAAIS 2026.Chapters:0:00 Intro - Revolut's AI at scale1:24 The problem: classical ML and three libraries2:54 The 2022 shift to API-served models4:25 Four internal groups, four sets of needs9:39 The decision: govern the use case, not the model10:54 One central gateway vs. distributed libraries14:10 Performance monitoring and drift detection17:33 Lesson: fallback chains and the silently-dead model20:02 Lesson: frontier vs. non-frontier cost (up to 8x)20:48 Lesson: the platform is the org chart22:59 Case study: from Rita to AIR26:40 Voice support at scale28:21 AIR, the in-app assistant30:25 Q&A: human oversight, hallucinations, AI as judge

Odyssey just raised a $310M Series B at a $1.45B valuation to build world models. We wrote the first check into the seed back in July 2024, so in this episode we walk through what the team has actually built, and why it is more interesting than "AI video."The short version: the scarce input for world models is experience. We get into how Odyssey is attacking that on three fronts. Starchild-1 gives world models sound, generating audio and video together in real time. Agora-1 is a learned game engine that drops four players into the same generated world, frame by frame. And PROWL lets a model hunt down its own failures and train on them.Along the way we cover why a world model is not a video generator, what self-driving taught Oliver Cameron and Jeff Hawke, and where this goes next for robotics, agents, and simulation.From Air Street Press. Read the full piece at press.airstreet.com.

Macrodata just raised a $4M pre-seed, led by Air Street, to build the data layer for robotics. The team behind FineWeb - Guilherme Penedo and Hynek Kydlíček - is bringing the discipline that made open LLMs work to messy physical-world robot data, through their open-source framework Refiner. We cover why physical AI is the next scaling paradigm, what Refiner does, and why we wrote the first check. From Air Street Press. Read the full piece at press.airstreet.com.

Last week the US government ordered Anthropic to switch off its most capable model for every foreign national on earth - four days after it launched. Nathan Benaich (founder of Air Street Capital, co-author of the State of AI Report) argues this exposed the AI risk almost no one is naming: not that the machines go rogue, but that everyone outside the US and China rents their intelligence from a landlord who can cut them off at will.Adapted from remarks given at a private dinner this week, this is the case for why Europe can't regulate its way to sovereignty - and what government, industry, civil society and academia each need to do to build it instead.Read the full essay and subscribe at press.airstreet.com.

Today we announce Perceptic — the AI operating system for biopharma built by the team behind Palantir's AIP and Life Sciences practice. Coming out of stealth with a $12M seed round from Air Street Capital, Accel and angels, Perceptic is already in production at CSL and multiple top-20 pharma companies.In this piece, we unpack why the frontier labs are racing into life sciences — from Anthropic's recent Novartis CEO board appointment to OpenAI's 80-year-old Erdős proof — why pharma's next R&D leap needs an application layer the model labs can't build alone, and why this is the team to do it.

Over the past few weeks, it’s been hard to keep up with AI in biology. Profluent signed a $2.25B partnership with Eli Lilly on AI-designed gene editors, Verve put out striking base-editing data, CZ Biohub published new scaling results on protein models, and Isomorphic Labs pulled in another large raise.I couldn’t think of anyone better to discuss this with than Ali Madani, the founder and CEO of Profluent. Profluent is an AI lab building frontier models to design proteins, with the goal of taking medicine from discovering molecules nature already made to designing the ones it didn’t. I first read Ali’s ProGen paper back in 2021, DMed him on then Twitter, and wrote the largest first check from Air Street Capital into the company at inception. Last month, Profluent announced a $2.25B deal with Eli Lilly, one of the largest to date between a frontier AI biology lab and big pharma.We discuss the shift from discovery to design, why Profluent bet sequence-first while others went structure-first, the Lilly deal and large-scale DNA editing, fine-scale base editing, whether LLM-style scaling laws hold for proteins, and much more. You can either watch the interview in full here or on YouTube or read the transcript below.Timestamp timeline0:00 – Teaser: AI-designed molecules & the $2.25B Lilly deal0:22 – Intros: Nathan Benaich (Air Street Capital) & Ali Madani (Profluent)2:10 – What is Profluent, and why AI matters5:45 – The landscape: readers vs. writers7:45 – Profluent’s edge: 100B+ sequences and a wet lab9:20 – OpenCRISPR and the exponential curve12:55 – Why sequence beats structure14:50 – The Eli Lilly deal and large gene insertion16:20 – Fine-scale vs. large-scale editing18:00 – Why it’s hard: the pre-AI era and the activity/specificity trade-off20:45 – The Verve news, and scaling beyond one-offs23:45 – Rare vs. common disease26:10 – “What do you know that no one else does?”27:40 – bio × AI is an undersaturated field32:40 – When will a top-10 pharma be AI-first?34:30 – Every molecule will be designed with AITimestamp

Air Street Capital has led Alta Ares’s $60M Series A.In this episode, we explain why modern air defense is no longer just about intercepting threats, but doing so affordably, under jamming, and at the speed of battlefield adaptation. From Ukraine to the GCC, the old model of static shields and slow procurement is breaking. Alta Ares is building full-stack, AI-first air defense across software, sensors, command-and-control, and effectors. We believe it can become Europe’s Iron Dome for autonomous air defense.

Angelos Perivolaropoulos leads speech-to-text research engineering at ElevenLabs across Scribe v2 and Scribe v2 Realtime. From AA-WER-leading batch transcription to 150ms real-time ASR, his work powers voice agents that can listen. A RAAIS 2026 speaker profile from Air Street Press.

Odyssey just shipped two new world models. Starchild-1 generates synchronized audio and video in real time at up to 24 fps, responding to streaming text, speech, or action input - the first real-time multimodal world model. Agora-1 puts up to four players into a shared simulated deathmatch on GoldenEye, with every frame each player sees generated on the fly while the model holds a shared world state across all participants.We walk through both releases, the technical contributions behind them - a causal distillation pipeline from a bidirectional audio-video foundation model, an asynchronous KV-cache that handles the audio/video clock mismatch, and a decoupled simulation/rendering architecture for the multi-agent case - and why world models are a different shape of system than the clip-makers (Veo, Sora, Kling) that have dominated generative video for three years.Odyssey is an Air Street Capital portfolio company. Jeff Hawke, Odyssey's co-founder and CTO, presents this work at RAAIS 2026 in London on June 12.

Nikolay Donets leads ML Engineering at Revolut - the platform behind classical ML, fraud detection, time-series foundation models, and the ElevenLabs voice agents now serving customers in 30+ languages. A RAAIS 2026 speaker profile from Air Street Press.

The April 2026 issue of State of AI from Air Street Press. The through-line: frontier AI moved from benchmark progress to operational capability across cyber, coding, agents, and capital formation.Two frontier models cleared the UK AI Security Institute's 32-step end-to-end cyber-attack range in a single month, and AISI now estimates frontier cyber-offence is doubling every four months, down from seven months at the end of last year. We unpack what that means for the public cybersecurity stack.Microsoft and OpenAI reset their 2019 deal to non-exclusive while keeping Microsoft as primary cloud partner. Anthropic stacked another $40B from Google, $5B from Amazon (with $100B of AWS spend), and chip deals with Google and Broadcom reportedly worth hundreds of billions, and is reportedly already raising again at a $900B valuation. Sam Altman's Axios essay sketched a "superintelligence New Deal" in explicit FDR terms.Four Chinese labs (Z.ai, MiniMax, Moonshot, DeepSeek) released open-weights coding models inside a 12-day window, all landing at roughly the same capability ceiling as Claude Opus 4.6 and GPT-5.4 on agentic engineering. NIST's CAISI evaluation puts the aggregate gap closer to eight months. Both are true.Anthropic's Project Deal ran a classified marketplace of 69 Claude agents and reported that stronger agents won, and the losers did not realise it. KellyBench (from Air Street portfolio company General Reasoning) watched every frontier model lose money on a Premier League betting season under non-stationarity. Ramp's procurement agents run 3× faster.Plus eight research papers worth keeping (π0.7, the Ríos-García epistemology paper, ClawBench, the FAIR experience-replay paper, Agent-World, and others), April Investments (Ineffable Intelligence's $1.1B seed, Saronic, Cognition's $25B talks, Cursor's $50B+ talks), and Exits (Skild and Zebra, SpaceX and Cursor, OpenAI and Hiro, Cohere and Aleph Alpha, China blocking Meta's acquisition of Manus).

Profluent just announced a multi-program strategic partnership with Eli Lilly to develop AI-designed recombinases for genetic medicine — worth up to $2.25 billion in milestones, plus tiered royalties on net sales.In this episode, Nathan unpacks why this deal matters far beyond the headline number. CRISPR taught us how to fix typos in the genome. The harder problem — and arguably the larger one — is editing at the kilobase scale: replacing whole paragraphs of DNA at a chosen genomic address. That's the route to therapies for the long tail of genetic disease driven by patient-level mutational heterogeneity, from cystic fibrosis to inherited hearing loss to retinal dystrophy.Recombinases have always been the right class of enzyme for this job. They've also been stuck for decades because their targeting specificity is encoded directly in the protein structure, with no equivalent of CRISPR's modular guide RNA. That makes recombinases a near-perfect problem for foundation-model protein design — and it's exactly the bet Profluent has been building toward since their 2024 work designing novel Cas enzymes from scratch.We cover: why kilobase-scale editing is the next frontier of genetic medicine; why recombinases were intractable until AI; how Profluent's foundation-model platform changes the picture; why Lilly is the right partner; and what the world looks like if you can name a genomic address and get a designed editor back.

Episode DescriptionWelcome back to the State of AI! In this packed Q1 2026 episode, we dive into a quarter defined by unprecedented geopolitical friction, staggering capital concentration, and the rapidly blurring lines between commercial cloud infrastructure and national defense.From a constitutional showdown between Anthropic and the Trump administration to the first-ever kinetic military strike on commercial data centers, the stakes for frontier AI have never been higher. Plus, we break down Anthropic’s explosive $19B ARR sprint, the escalating "distillation wars" with Chinese AI labs, and the historic $1.25 trillion merger between xAI and SpaceX.If you want to understand where the frontier is heading next, you can't miss this one.The Pentagon Standoff: Anthropic's $200M DOD contract, its refusal to drop safety guardrails, and the ensuing White House blacklist and federal lawsuit.Cloud as a Theater of War: Breaking down the unprecedented Iranian drone strikes on AWS data centers in the Middle East.Revenues Go Vertical: How Anthropic surged to a $19B ARR on the back of Claude Cowork, and OpenAI's massive $50B strategic alliance with Amazon.The Model Treadmill: The rapid succession of new model releases, including Claude Sonnet 4.6, Gemini 3.1 Pro, and GPT-5.4.The Distillation Wars: Inside the industrial-scale IP theft by Chinese labs cloning Claude, and the $2.5B NVIDIA GPU smuggling bust.Safety Meets Reality: Sabotage risks, machine-speed SQL injections, and the UK AI Safety Institute's chilling findings on AI-assisted cyber attacks.The Physical Layer & NIMBYism: The pushback against hyper-scale data centers and NVIDIA's complete exit from the China-compliant chip market.Breakthrough Research: From zero-loss cache compression (TurboQuant) to an Australian entrepreneur curing his dog's cancer with AlphaFold.Historic Mega-Deals: OpenAI's record-shattering $110B raise and xAI's trillion-dollar merger into SpaceX.00:00 - Intro, Air Street Capital Epoch 3, & RAAIS 202601:28 - Geopolitics: Anthropic vs. The White House03:12 - The Iran-AWS Conflict & Cloud Warfare04:12 - Financials: Anthropic's $19B ARR & OpenAI's Hyperscaler Strategy08:04 - The Model Treadmill: Claude Sonnet 4.6, Gemini 3.1 Pro, & GPT-5.408:56 - Open Source, IP Warfare, & the $2.5B Smuggling Ring10:44 - AI Safety: Catastrophic Sabotage & The Sabotage Risk Report13:20 - Data Center NIMBYism & The Contested Physical Layer16:00 - Research Highlights: UK AISI, TurboQuant, & World Action Models22:40 - Investments & Exits: OpenAI's $110B Round & The SpaceX/xAI MergerStay Connected:Love hearing what you’re up to! Hit reply to our newsletter or connect with us at the upcoming Air Street AI meetups in SF (April 28) and NYC (May 14). We are also actively recruiting Research Analysts for the State of AI Report—reach out if you live and breathe this space.Produced by the State of AI & Air Street Press.In This Episode, We Cover:Episode Timestamps (Estimated):

Raia Hadsell is VP of Research at Google DeepMind, co-leading the Frontier AI unit. Her work spans Siamese nets and elastic weight consolidation to Gemini 2.5, RoboCat, and a UK AI Ambassador role. A RAAIS 2026 speaker profile from Air Street Press.

Roberta Raileanu leads open-ended learning at Google DeepMind and co-authored Toolformer. From RIDE and AMIGo to Llama 3's tool use and MLGym, her research tackles what it takes for AI agents to keep acquiring skills. A RAAIS 2026 speaker profile from Air Street Press.

Starcloud launched the first NVIDIA H100 GPU in space and trained the first LLM in orbit. CEO Philip Johnston explains why AI's energy bottleneck leads to orbital data centres — with 10x lower energy costs and 5 gigawatts of solar-powered compute on the roadmap. A RAAIS 2026 speaker profile from Air Street Press.

Air Street Capital has raised a third fund of $232M to back AI-first companies from the earliest stages. In this post, founder Nathan Benaich shares the conviction behind the firm - from his first investments in 2013 through to a portfolio that now includes Synthesia, Black Forest Labs, Wayve, Profluent, and poolside - and explains what Fund III enables for the most ambitious AI founders in Europe and North America.Read more: https://press.airstreet.com/p/fund-iii

Sereact's Cortex 2.0 marks a shift in robotics from reactive control to predictive planning. In this episode, we examine how Sereact’s world-model architecture generates and scores imagined futures before acting, improving success rates and eliminating human intervention across complex warehouse tasks. We break down the benchmark results, the planning budget trade-off, and what it means to deploy world models in real industrial environments rather than simulation.

European voters say they support higher defense spending. But when higher taxes or welfare cuts are mentioned, support collapses.In this episode from Munich Security Conference 2026, we explore Europe’s fiscal test: Germany’s industrial flywheel, the reality of attrition warfare in Ukraine, the broken procurement model, and the tension between welfare and warfare.Europe has demonstrated urgency. Now it must prove permanence.

In this episode of the State of AI, we break down the growing disconnect between rapid AI capability gains and collapsing software valuations, with nearly $300B wiped from public markets in weeks. We cover the agent shock triggered by Anthropic and OpenAI’s latest releases, why investors are repricing long-term SaaS revenues, and how AI sovereignty is fracturing across U.S. policy, state-level infrastructure pushback, and China’s accelerating model and talent pipeline. We also look at the security risks of computer-use agents, the infrastructure arms race spanning GPUs, memory, power, and data centers, and the latest research breakthroughs in autonomy, medicine, and reinforcement learning. Plus, a full rundown of the month’s largest AI financings, IPOs, and acquisitions.

AI has progressed fastest where the world can be cleanly digitized, but robotics remains stubbornly hard. In this episode, we examine Sereact’s Cortex 1.6 and what its results reveal about learning from execution rather than sparse task outcomes. We discuss why execution-level learning improves robustness, recovery behavior, and learning efficiency in real-world robotic manipulation, and what this signals for the future of deployment-first robotics.

In this episode, Air Street Capital shares its 2025 year in review. We cover what changed as AI moved into large-scale deployment, from the emergence of reasoning models and agents in production to the economics of frontier AI, energy constraints, and geopolitics.We reflect on the year across our investment portfolio, angel investments, Air Street Press, the State of AI Report, and our global community, and look ahead to what it will take to deploy AI reliably and at scale in the years to come.

Europe sharply increased defense spending in 2025. But money alone does not produce weapons, stockpiles, or readiness.In this episode, we examine why Europe’s defense build-up is running into industrial limits as it enters 2026. From procurement bottlenecks and factory capacity to Germany’s surge in orders and the slow pace of production, the challenge is no longer political will - it is execution.This is a conversation about defense as an industrial system, and why turning budgets into battlefield capability is proving harder than expected.Read more on press.airstreet.com

As 2025 came to a close, conversations about AI swung between excitement and anxiety. Markets debated bubbles, capital cycles, and constraints, while researchers quietly shipped systems that worked.In this audio essay, Nathan Benaich takes stock of what AI actually delivered in 2025 — drawing on recent writing by Tim Dettmers, Dan Fu, and Andrej Karpathy, alongside conversations with Sebastian Borgeaud at Google DeepMind.Rather than speculating about distant futures, this episode focuses on what changed in practice: why AI crossed a usability threshold, how constraints reshaped progress rather than stopping it, and why the shift from models to systems matters more than any single benchmark.The result is a grounded look at AI progress as it enters 2026 — not as hype or prediction, but as an evolving system that continues to compound.Read the full essay at press.airstreet.com

Delfa is an AI-first clinical trials software company. In this episode, we explore how Delfa’s AI-first Participant Relationship Management system transforms clinical trial operations, speeds recruitment, and helps bring medicines to patients faster.

New AI research systems are beginning to contribute verifiable results across mathematics, physics, biology, and materials science. How close are we to AI producing genuinely new scientific knowledge?

A deep dive into world models, VLAMs, planning layers and real deployments from robotics companies Sereact and Wayve - and what comes next for embodied AI.

Today, I’m excited to unveil Air Street’s investment in Black Forest Labs as it announces a landmark funding milestone: a $300M Series B, following a previously unannounced Series A. Together, these rounds represent a big step in scaling the company’s momentum, with Black Forest Labs now capitalised with half a billion dollars and trusted by leading Fortune 500 enterprises.

Welcome to the latest issue of the State of AI, an editorialized newsletter that covers the key developments in AI policy, research, industry, and start-ups over the last month.

Building the first AI-native wealth institution for the mass affluent. Air Street Capital is investing in Clove’s $14M first financing round, backing a new kind of wealth institution built for this generation rather than the last. Clove’s founders, Christian Owens and Alex Loizou, see this gap not as an inevitability but as a result of infrastructure that was never built for the modern consumer. With Clove, they are creating a new kind of financial institution, one designed from the ground up for people who want trustworthy guidance but have been priced out or ignored by traditional services. Their platform brings together regulated human advisors with an AI-first environment that handles the repetitive and compliance heavy processes which dominate advisory work today. By removing friction and expanding advisor capacity, Clove can deliver high quality personalised guidance at a scale that has not been possible before.

Profluent builds frontier AI systems to unlock programmable biology. Now, Profluent has raised $106M led by Bezos Expeditions and Altimeter, with continued support from Spark, Insight, and Air Street. The company is now the largest position in Air Street’s second fund. This new capital accelerates the company’s path toward scaling frontier protein models and, ultimately, delivering the first AI-designed therapeutic to a human patient.

Understanding how protein sequence encodes structure and function remains one of the central challenges in the life sciences. Yet most protein language models still treat each sequence as an isolated datapoint. This forces the entire burden of evolutionary context into model parameters, which leads to blind spots in underrepresented families and amplifies the biases of sequence databases. Profluent’s new E1 family demonstrates that this constraint is no longer necessary. Retrieval augmentation, a technique that transformed natural language processing, is now beginning to reshape protein modeling by allowing models to incorporate evolutionary information at the moment of inference rather than storing it all in weights.

Poolside is building one of the strongest full-stack AI companies in the world: energy, compute, models, and the infrastructure needed to run multi-agent systems for complex enterprise workflows. The Fern team brings a deeply opinionated agentic core - built from first principles and stress-tested on real workloads - into a company with scale, distribution, and compute firepower. Fern’s architecture was built for agent specialization and coordinated long-horizon work - the exact capabilities poolside can scale into a full production environment.

Welcome to the latest issue of the State of AI, an editorialized newsletter formerly known as Guide to AI that covers the key developments in AI policy, research, industry, and start-ups over the last month. First up, a few reminders:Read more on press.airstreet.com

PARIMA, a global leader in cultivated proteins, has become the first European company to secure regulatory approval for cultivated meat, with the Singapore Food Agency granting clearance for its cultivated chicken. It’s a historic moment for Europe’s food-tech sector, but one that’s unfolding thousands of miles away from home.

Building a next-generation food company.

Integrating across compute, power, and intelligence.

The State of AI Report is the most widely read and trusted analysis of key developments in AI. Published annually since 2018, the open-access report aims to spark informed conversation about the state of AI and what it means for the future. Produced by AI investor Nathan Benaich and Air Street Capital. State of AI Report 2025 is reviewed by leading AI practioners in industry and research.

AI is the ultimate force multiplier on technological progress in our digital, data-driven world. Our mission at Air Street Capital has always been to back the most ambitious teams building breakthrough AI products that would have seemed like magic when I started investing over a decade ago. Key to these inflection points is the infrastructure that enables them: NVIDIA computing systems. I have long argued that NVIDIA is the defining company of the AI era, powering breakthroughs in science, industry, and national strategy. This is why I’m excited to share that we are deepening that story together: Air Street Capital is partnering with NVIDIA as part of its new £2B commitment to the UK AI ecosystem.

Clinical trials are the bottleneck of the pharmaceutical industry. They’re slow, expensive, and often fail, not only because the science frequently doesn’t pan out, but because the ops don’t either. That’s why we’re leading the $3.8M Seed round for Delfa. The team is building an AI-native operating system for clinical trials, starting with patient enrolment, the most broken part of the process.