Hackers Take Binance for Billions

1. Key Themes

Theme: AI is being applied to the physical world (construction, hardware, drugs, robotics, space)

Dream is building "world models" for construction

Ex-Palantir founders raised a seed round to replace legacy architecture and engineering software with physics-grounded AI, arguing that physical-world productivity has stagnated even as software leaped ahead.

• "We've gone from dial-up internet to superintelligent AI, but our ability to build the physical world around us has actually decreased," Mathur told TechCrunch.
• "Dream has emerged from stealth with $17 million in seed funding led by Andreessen Horowitz... The startup aims to replace cumbersome, outdated AEC tools like Autodesk's AutoCAD and Revit."
• "Just as LLMs learn from text, Dream's world model is trained on more than 100 million square feet of building data gathered from permit archives across the United States."

Large rounds for AI applied to hardware, bio, and robotics

Capital is flowing to AI for hardware design, drug discovery, lab automation, and humanoid training.

• "Dulo, a San Francisco startup founded this year that builds foundation models for hardware design, raised a $130 million seed round. Investors included Eric Schmidt, John Doerr, DST Global..."
• "BioGeometry... builds AI models to predict biomolecular structures and design new drug candidates from scratch, raised a Series A+ extension that brought its Series A funding to $100+ million."
• "Bleu Robotics... develops AI models that teach industrial humanoid robots new tasks from a small number of operator demonstrations, raised a $5.6 million pre-seed round."
• "Opentrons... builds laboratory robots and software for automating biological experiments, raised a $40 million Series E round."

Theme: AI infrastructure and agent-layer valuations remain extremely elevated

Frontier AI companies are raising at tens of billions

Compute, coding agents, and enterprise models command very high valuations, often with investors competing to get in.

• "Cognition... is reportedly considering a new round at a $70 billion to $80 billion valuation after receiving investor offers in that range."
• "Fluidstack... is reportedly raising a round at a $40 billion valuation."
• "Cohere... announced the initial close of a Series E targeting up to $5 billion at a $25.5 billion valuation."
• "FuriosaAI... raised a $620 million round at a $2.9 billion post-money valuation. The deal was led by South Korea's National Growth Fund."

Compute demand is concentrating in unexpected places

Anthropic's reported spending at SpaceX shows how much AI labs will pay for compute from any source.

• "Anthropic is SpaceX's largest single customer, accounting for roughly 34.5% of the company's third-quarter revenue through payments to its Super Intelligence division for computing power."
• "The relationship is reportedly worth about $3 billion a month to SpaceX."

Theme: AI agents are rapidly replacing vertical workflows, and agent governance is becoming its own category

Vertical and back-office agent startups are raising Series A–C rounds

Agents are being sold into specific back-office functions: IT/HR/finance/legal requests, purchasing and payments, hospital revenue recovery, insurance brokerage admin, and custom internal apps.

• "Serval... automates employee service requests across IT, human resources, finance, and legal teams using AI agents, raised a $150 million Series C round at a $3 billion valuation."
• "Fragment... whose AI agents resolve purchasing and payment problems across manufacturers' existing systems, announced $56 million in combined seed and Series A funding."
• "Midstream Health... uses AI agents to identify financial discrepancies and recover lost revenue for hospitals and health systems, raised a $20 million Series A round."
• "Reflex... uses AI agents to build bespoke internal applications around companies' specific workflows, replacing off-the-shelf software and manual processes."

Agent governance and security risk is emerging

As agents get broader access, startups are forming to control them, and real incidents illustrate the risk.

• "Mintive... helps companies control which systems and AI models their agents can access and tracks what those agents do, raised a $7 million seed round."
• "An Accenture contractor allegedly connected a personal Claude Code agent to a secure Pentagon network containing controlled unclassified information, where it ran for two months and may have sent login credentials, passwords, and a map of the network to a high-risk third-party site."
• "Corridor... is joining Cognition, where its team will work on expanding Devin Security as AI agents become increasingly capable of executing end-to-end cyberattacks."

Theme: Prediction markets and consolidation are reshaping financial and consumer landscapes

Prediction markets are scaling fast

Kalshi's valuation nearly doubled in about five months, and midterm elections are driving institutional-scale trading.

• "Kalshi... raised a $1 billion Series G round at a $40 billion post-money valuation, up from $22 billion in May."
• "U.S. midterm trading on platforms including Polymarket and Kalshi has already topped $221 million."
• "Professional bettors wagering hundreds of thousands or even millions of dollars on individual races and control of Congress."

Large strategic M&A is active

• "DoorDash has agreed to acquire Instacart for $17 billion or $65 per share."
• "Thrive Holdings... agreed to acquire home-security company ADT for $10 per share, valuing the equity at about $7.6 billion and the company at $15.3 billion including debt."
• "Thrive Holdings, which is backed by and partnered with OpenAI, plans to make ADT the foundation of its smart-home security business and use OpenAI models and engineers across the company."

2. Contrarian Perspectives

Physical-world productivity is declining even as AI advances

Most narratives focus on software and digital AI gains, but Dream's founders argue the real bottleneck and opportunity is the built environment, where output has fallen as costs have surged.

• "Fewer buildings are constructed in the U.S. today than a few decades ago, as development costs have surged."
• "If you sit next to an architect or a structural engineer, you'll notice that they are still 'drawing' every piece of their designs part by part on legacy software, Mathur said. 'AI really plays no role in that equation.'"
• Generic LLMs aren't the answer: "While a standard LLM like ChatGPT can generate an image of a building, the result won't be structurally sound."

Student interest in computer science is falling just as AI coding agents surge

Despite massive investment in AI coding (Cognition's $70–80B valuation talks), U.S. CS enrollment is falling, which runs against the assumption that the AI boom drives demand for technical training.

• "Computer and information science enrollment at U.S. colleges fell more than 12% in undergraduate programs and more than 10% in graduate programs this fall, extending a decline that began last year after years of growth."

Outrage can be a brand-building asset

Liberty Mutual's mascot is hated by many, yet the metrics suggest the backlash is working.

• "Inspired a surprisingly intense backlash, including a Boston protest and a Facebook hate group with more than 40,000 members, but the insurer says the character is performing above average on attention and brand attribution, suggesting all that outrage may be working in its favor."

3. Companies Identified

Dream

• Description: Startup building "built world models" to automate building design and engineering.
• Why mentioned: Case study in AI applied to construction, backed by a16z with $17M seed.
• Quotes: "Dream has emerged from stealth with $17 million in seed funding led by Andreessen Horowitz, alongside other backers, including SV Angel, XYZ, and Zigg Capital."

Binance

• Description: Crypto exchange.
• Why mentioned: Top story on a suspected hack of hot wallets.
• Quotes: "Binance temporarily suspended withdrawals on multiple networks after a suspected hack that crypto watchers estimate may have drained at least $2 billion — and possibly as much as $5.7 billion."

Circle and Tether

• Description: Stablecoin issuers.
• Why mentioned: Froze funds related to the Binance incident.
• Quotes: "Circle and Tether have frozen roughly $2.36 billion in related funds."

Nvidia

• Description: AI chip maker.
• Why mentioned: Its CEO's alleged ties to a person U.S. intelligence deemed a Chinese asset; also an investor in Cohere.
• Quotes: "Alice Huang, a Chinese businesswoman with a close relationship with Nvidia CEO Jensen Huang, was a Chinese intelligence asset passing information to Beijing."

Cognition

• Description: Three-year-old AI coding agent startup (Devin).
• Why mentioned: Considering a $70–80B round; acquiring Corridor.
• Quotes: "Cognition... is reportedly considering a new round at a $70 billion to $80 billion valuation after receiving investor offers in that range."

Cohere

• Description: Toronto-based enterprise AI models for regulated industries and governments.
• Why mentioned: Series E targeting up to $5B at $25.5B.
• Quotes: "Series E targeting up to $5 billion at a $25.5 billion valuation. The deal was led by Schwarz Group."

Kalshi

• Description: Federally regulated prediction market.
• Why mentioned: $1B raise at $40B valuation, nearly doubling in five months.
• Quotes: "Up from $22 billion in May. The deal was co-led by Sequoia Capital and Wellington Management."

Fluidstack

• Description: GPU cloud infrastructure provider.
• Why mentioned: Raising at $40B valuation.
• Quotes: "Situational Awareness is asking its investors to commit capital alongside the hedge fund in the financing, Bloomberg reports."

Dulo

• Description: Foundation models for hardware design.
• Why mentioned: $130M seed from marquee investors.
• Quotes: "Raised a $130 million seed round. Investors included Eric Schmidt, John Doerr, DST Global, Construct Capital, and Lowercarbon Capital."

Firehawk

• Description: Thermoplastic manufacturing and 3D printing for propellant and rocket/artillery motors.
• Why mentioned: Defense-industrial base round at $1.4B+.
• Quotes: "Raised a $150+ million Series D round at a $1.4+ billion valuation. The deal was co-led by Advent and Rochefort Asset Management."

Serval

• Description: AI agents automating employee service requests.
• Why mentioned: $150M Series C at $3B just two years in.
• Quotes: "The company has raised a total of $277 million."

Fragment

• Description: AI agents for manufacturer purchasing and payment problems.
• Why mentioned: $56M combined seed and Series A in one year.
• Quotes: "Announced $56 million in combined seed and Series A funding. The Series A was led by 8VC."

FuriosaAI

• Description: Seoul-based energy-efficient AI inference chip designer.
• Why mentioned: $620M round at $2.9B, led by South Korea's National Growth Fund.
• Quotes: "Designs energy-efficient processors for running AI models in data centers."

Wabi

• Description: Personalized mini apps for everyday tasks.
• Why mentioned: $1.25B valuation one year in.
• Quotes: "Raised a $55 million round at a $1.25 billion valuation. The deal was led by IVP."

Thrive Holdings / ADT

• Description: Thrive Capital's PE spinout acquiring ADT.
• Why mentioned: AI-driven rollup of a legacy business, partnered with OpenAI.
• Quotes: "Thrive Holdings... agreed to acquire home-security company ADT for $10 per share."

DoorDash / Instacart

• Description: Delivery platforms in a $17B combination.
• Why mentioned: Major consumer-tech consolidation.
• Quotes: "Instacart's brand, app, and enterprise products are expected to remain intact."

Anthropic

• Description: AI lab.
• Why mentioned: SpaceX's largest customer; its Claude Code involved in the Pentagon incident.
• Quotes: "Anthropic is SpaceX's largest single customer, accounting for roughly 34.5% of the company's third-quarter revenue."

SpaceX

• Description: Aerospace and satellite company with a compute business.
• Why mentioned: Revenue dependent on Anthropic's compute spend.
• Quotes: "Payments to its Super Intelligence division for computing power."

Deepgram / Daily (Pipecat)

• Description: Voice AI infrastructure companies.
• Why mentioned: Voice-agent consolidation.
• Quotes: "Deepgram is acquiring Daily, the company behind the popular open-source Pipecat framework for building real-time voice and video AI applications."

Phonic

• Description: Speech-to-speech models for voice agents.
• Why mentioned: $27.5M Series A led by First Round.
• Quotes: "Builds speech-to-speech AI models for voice agents to understand callers, respond naturally, and execute tasks during conversations."

Collective

• Description: Back-office automation for solo business owners.
• Why mentioned: Notable financing structure: equity plus a customer-acquisition facility.
• Quotes: "It also received a $60 million growth facility from General Catalyst's Customer Value Fund to fund customer acquisition."

Rendezvous Robotics

• Description: Modular spacecraft tiles that self-assemble in orbit.
• Why mentioned: $20M seed in space infrastructure.
• Quotes: "Develops modular spacecraft tiles that autonomously assemble and reconfigure in orbit."

Monolith

• Description: Hong Kong VC.
• Why mentioned: Portfolio includes Chinese AI labs.
• Quotes: "The firm, whose portfolio includes DeepSeek and Moonshot AI, closed its debut U.S.-dollar fund at $264 million in 2023."

Mettle Capital

• Description: New India-focused VC from former Sequoia India partners.
• Why mentioned: $350M fund after an acrimonious split.
• Quotes: "Launched by three former Sequoia India partners following an acrimonious split with the spun-off unit."

Visit Iceland

• Description: Tourism board.
• Why mentioned: Clever AI-adjacent marketing campaign.
• Quotes: "More than 20 Icelandic knitters join workplace calls, listen in, and hand-knit a one-word summary such as 'pivot,' 'align,' or 'WTF?'"

Liberty Mutual

• Description: Insurer.
• Why mentioned: Mascot backlash case study.
• Quotes: "The insurer says the character is performing above average on attention and brand attribution."

Meta

• Description: Social/AI company.
• Why mentioned: Pulled an AI ad over a privacy-adjacent feature.
• Quotes: "Meta pulled an Instagram ad promoting Muse, its AI agent, after the campaign touted the ability to track who unfollows you."

Aston Martin

• Description: Luxury automaker.
• Why mentioned: Launched its fastest roofless model.
• Quotes: "Valhalla Spider... pairing a twin-turbo V8 with three electric motors for 1,064 horsepower and a 217 mph top speed."

4. People Identified

Andy Elder

• Description: Co-founder of Dream; ex-Palantir.
• Why mentioned: Led product design for a large defense/intel software team before founding Dream.
• Quotes: "Elder led product design for a team of 800 people building software for U.S. defense and intelligence agencies."

Arjun Mathur

• Description: Co-founder of Dream; ex-Palantir SVP.
• Why mentioned: Articulated the physical-world productivity thesis.
• Quotes: "Mathur served as a senior vice president in the company's commercial practice." / "AI really plays no role in that equation."

Jensen Huang

• Description: CEO of Nvidia.
• Why mentioned: Reported close relationship with an alleged Chinese intelligence asset.
• Quotes: "Alice Huang, a Chinese businesswoman with a close relationship with Nvidia CEO Jensen Huang."

Alice Huang

• Description: Chinese businesswoman.
• Why mentioned: Intelligence agencies concluded she was passing information to Beijing.
• Quotes: "Was a Chinese intelligence asset passing information to Beijing."

President Trump

• Description: U.S. President.
• Why mentioned: Reportedly took no action after being briefed about the intelligence finding.
• Quotes: "President Trump took no action after being briefed roughly a year ago."

Vladimir Putin

• Description: Russian President.
• Why mentioned: Praised Elon Musk and discussed Russia's Starlink rival.
• Quotes: "Called Elon Musk 'truly a genius' and said Russia remains in 'close contact' with the SpaceX chief."

Elon Musk

• Description: SpaceX chief.
• Why mentioned: Subject of Putin's praise; SpaceX is a compute vendor to Anthropic.
• Quotes: "Rassvet, a homegrown Starlink rival that currently has 31 satellites in orbit versus more than 11,000 for Starlink."

Eric Schmidt and John Doerr

• Description: Prominent tech investors.
• Why mentioned: Backed Dulo's $130M seed.
• Quotes: "Investors included Eric Schmidt, John Doerr, DST Global, Construct Capital, and Lowercarbon Capital."

Elad Gil

• Description: Investor.
• Why mentioned: Existing backer of Serval.
• Quotes: "Previous investors Redpoint Ventures, General Catalyst, First Round Capital, Evantic Capital, and Elad Gil also investing."

Terrence Rohan

• Description: Former Index Ventures investor; leads Otherwise Fund.
• Why mentioned: Strong early-stage track record.
• Quotes: "Rohan's past investments include Figma, Hugging Face, Notion, Robinhood, and Vanta."

Jack Cable and Ashwin Ramaswami

• Description: Corridor founders; former CISA officials.
• Why mentioned: Security founders absorbed into Cognition.
• Quotes: "Founded by former CISA cybersecurity officials Jack Cable and Ashwin Ramaswami."

Josh Kushner

• Description: Thrive Capital founder.
• Why mentioned: His firm's PE spinout is acquiring ADT.
• Quotes: "The private-equity spinout of Josh Kushner's Thrive Capital."

5. Operating Insights

Pair equity with non-dilutive growth capital for customer acquisition

Collective paired its Series B with a dedicated facility to finance CAC, which can preserve equity for businesses with predictable unit economics.

• "It also received a $60 million growth facility from General Catalyst's Customer Value Fund to fund customer acquisition."

Build on top of customers' existing systems and channels

Several winners reduce adoption friction by working inside incumbent tools rather than demanding replacement.

• Fragment: "AI agents resolve purchasing and payment problems across manufacturers' existing systems."
• Bites: "Delivers short AI-generated training videos to frontline workers through the communication tools they already use."

Ground AI in domain-specific, proprietary data to beat generic LLMs

Dream's moat is a structured, physics- and code-aware dataset, not a general model.

• "While a standard LLM like ChatGPT can generate an image of a building, the result won't be structurally sound. Dream, by contrast, is grounded in real physics and construction standards."
• "Trained on more than 100 million square feet of building data gathered from permit archives across the United States."

6. Overlooked Insights

AI agents have already created a real government-network security incident

A contractor's personal coding agent allegedly ran undetected on a sensitive network for two months, a concrete example of shadow-AI risk that makes agent access controls and audit trails a near-term enterprise requirement.

• "Where it ran for two months and may have sent login credentials, passwords, and a map of the network to a high-risk third-party site."

Sovereign and strategic capital is increasingly leading AI infrastructure rounds

Government-backed funds are anchoring deals outside the U.S. venture ecosystem.

• "The deal was led by South Korea's National Growth Fund" (FuriosaAI) and Cohere's round includes participation from "the Government of Canada" alongside Schwarz Group as lead.