February: no frontier model could solve Irregular's hardest exploitation task. By April: the best one could, occasionally, for ~$2,000. By June: several could, reliably, for ~$20. Over the next year, we will likely enter a period where AI's offensive capability outpaces defense at scale. @dan_lahav and the @Irregular team are truly at the frontier of AI security, as major partners to the frontier labs. This is the most detailed and thoughtful account of where AI security is today, where it's headed, and what we can do now so defenders don't fall behind.

Meet Lucy 2.5, our most advanced Live AI model yet. Lucy edits videos in realtime, now with more capabilities and greater control. See how it's being used across streaming, e-commerce, advertising, and more 🧵

Agents aren’t coming. They’re already at the door. Ora lets them in. @telnyx, the leading voice platform for real-time AI agents, is already acting on it with Ora: 30%+ of their agent signups now convert to paid. Read the story:

We appreciate @AnthropicAI's collaboration and transparency. Addressing these risks will require closer cooperation across the AI ecosystem. We as well look forward to working together with Anthropic to advance security.

Frontier compute will likely get 10x more expensive. But the price of useful intelligence will likely still collapse. We need a CPI for intelligence... What does it cost to get the same real work done to the same standard?

Sequoia's @deanmeyerrr: "If distillation is real, the question is, how do we address the asymmetry?" "Western companies aren't going to set up 35,000 proxy accounts to steal from Anthropic." "There *is* a structural asymmetry, and it's something we have to think about." "The gap between closed and open source has closed a lot." "And the key conversation online, over the past week, was distillation the reason for this? Or is there some other, mysterious reason?" "The truth is, from the conversations we've had, is that it's somewhere in the middle." "Whether it's 20% or 40%, it doesn't really matter as much. But it's meaningful that we know it's happening." "We know from the labs that there's 10s of millions of queries through these large-scale proxy distillation accounts." "We also know, if you speak to anyone who's leading in synthetic data gen on pre- or post-training at the labs, they will tell you that even if you do hardcore KYC, and then just get access to the output (let's assume you just get the output from the prompt), it is still extremely valuable." "That is 100% true. We know distillation is real."

We’ve raised $300M in Series C funding at a $10.3B valuation from Sequoia, Andreessen Horowitz, Jane Street, Argo, and SK Hynix. Our mission is to run the world's inference. This round accelerates production of our inference clusters. We've opened an 80,000-sqft, 10-MW facility 15 minutes from our office to expedite production and prototyping.

Excited to share that we’ve raised $300M in our Series B round, led by @radicalvcfund, bringing our total funding to more than $450M, with leading technology companies joining as both customers and investors. We continue accelerating the path to AGI through our two core pillars: ultra-optimized infrastructure for AI workloads, and realtime world models built on top of it. Today, we’re also launching DOS (Decart Optimization Stack) 2.0 – our next-generation inference and training platform, delivering over 1,600 tokens per second for agentic inference and over 100 FPS for world models across major hardware platforms. Alongside DOS, we will launch new versions of our world models in the coming weeks: Lucy, for immersive realtime experiences, and Oasis, for physical AI. Grateful to our partners, customers, and backers across media & entertainment, Physical AI, chips, hyperscalers, and the broader AI ecosystem. @radicalvcfund, @Adobe, @alphaptrs, @amazon and @awscloud, @Atreidesmgmt, @benchmark, @eBay, @nvidia, @sequoia, @Toyota, @valor, @orenzeev; Andrej Karpathy, Michael Eisner, Yamauchi-No.10 Family Office, Moritz Baier-Lentz and more. We made a film for this moment with Decart CEO @DLeitersdorf and Moritz Baier-Lentz - a look at the company, the technology, and what comes next.

The agentic web needs agent-ready tools. We spent a decade optimizing websites for humans. Lighthouse was built by humans to measure human experience. Ora is powered *by real world agents* that rank the real agentic experience. Instead of guessing:

#Anthropic# got it wrong. The public models already grant attackers the upper hand. This is why we created #MOAK# - to prove that the exploitation risk is no longer theoretical. AI-powered instant exploitation of 1-days is now the rational move for every threat actor. We are expecting "Spray & Pray" attacks to hit within an hour of publication, and this won't be limited to nation-states. While #Glasswing# focuses on maintainers, it lacks the robust support required by enterprises. In the long run, these tools might make the world safer, but in the short term, currently available models enable a "Wild West" scenario. I believe that the next two years will be brutal for enterprises. AI-powered attackers are significantly more vicious. To the defenders reading this - we’re entering a dangerous era - you must step up your patching game. You’ve got to check the website, the link is in the first comment. To be continued.