U.S. Aims Anthropic's Mythos at…

1. Key Themes

AI Is Becoming Critical Government Infrastructure

The U.S. government is actively deploying frontier AI models for national security applications, signaling a major shift in how previously politically sensitive AI companies are viewed at the federal level.

"The U.S. Cybersecurity and Infrastructure Security Agency is reportedly using Anthropic's Mythos model to scan government code for security flaws, with audits already finding numerous vulnerabilities."

Humanoid Robotics Is in a Full Capital Frenzy — But Operators Are More Cautious Than Investors

Billions are pouring into humanoid robotics at increasingly aggressive valuations, yet at least one key CEO is deliberately tempering expectations, suggesting a gap between investor enthusiasm and commercial reality.

"AI2 Robotics raised roughly $735 million at a nearly $3 billion valuation... Apptronik closed a $935 million funding round valuing the company at more than $5.5 billion... Figure AI closed on $1 billion in Series C funding at an eye-popping $39 billion valuation." Contrasting that: "Peggy Johnson, CEO of Agility Robotics, is surprisingly measured."

AI Is Now Both the Attack Vector and the Defense

Autonomous AI agents are being weaponized for cyberattacks end-to-end, creating an arms race that validates investment in AI-native cybersecurity tooling.

"A newly documented ransomware operation called JadePuffer reportedly used an autonomous AI agent to carry out an entire attack chain, from reconnaissance and credential theft to lateral movement, persistence, privilege escalation, and data encryption."

State-Level AI Regulation Is Filling the Federal Vacuum

With no federal AI standards in place, states are moving aggressively to regulate frontier models, creating compliance complexity for AI developers and a potential new market for regulatory tooling.

"Illinois Gov. JB Pritzker signed a bill requiring large AI developers to publish safety frameworks, measure catastrophic risks, report incidents, and use third-party auditors, as states move to regulate frontier models in the absence of federal standards."

Manufacturing Tech and Physical AI Are Attracting a New Wave of Capital

A convergence of AI, robotics, reshoring, and hardware automation is pulling investors back toward factory floors — including specialized new funds targeting the space at the pre-seed stage.

"Omni Ventures, a new firm founded by former Apple executives Simon Lancaster and Sabrina Paseman, raised a $33 million fund to back pre-seed manufacturing-tech startups as AI, robotics, reshoring, and hardware automation pull investors back toward factory floors."

2. Contrarian Perspectives

The Trump Administration Is Publicly Bullish on AI While Its Own Treasury Analysts Are Quietly Sounding the Alarm

There is a significant disconnect between the administration's public posture and internal risk assessment — a meaningful signal for investors calibrating AI exposure.

"U.S. Treasury analysts have drafted an internal report warning that an AI bubble could ripple through stock markets, private credit, data-center financing, cloud providers, chipmakers, and utilities, even as the Trump administration publicly casts AI as central to U.S. economic growth."

Open-Source AI Dominance Does Not Translate to Business Dominance

Alibaba's Qwen demonstrates that massive open-source adoption — ~1 million downloads per day — does not automatically create a monetization path, and may actually create internal strategic conflict.

"Alibaba's Qwen has become one of the world's most popular open-source AI model families, but the Chinese tech giant is struggling to turn that reach into revenue as infrastructure costs soar and its AI team splits over whether to move toward paid, proprietary models."

The SPAC Route for Robotics May Be a First-Mover Advantage Play, Not a Valuation Dodge

Conventional wisdom dismisses SPACs as a sign of weakness or an attempt to avoid scrutiny. But Agility's rationale reframes it as a strategic timing play to capture public-market demand before competitors.

"Johnson said much of it boils down to the first-mover advantage the company enjoys when it's the first of its ilk to go public. For investors clamoring for shares in a buzzy robotics company, Agility is 'an acceleration story and a timing story.'"

3. Companies Identified

Anthropic | AI safety company | Featured prominently; its Mythos model is being deployed by CISA to scan federal government code for security vulnerabilities, marking a major government endorsement.

"The U.S. Cybersecurity and Infrastructure Security Agency is reportedly using Anthropic's Mythos model to scan government code for security flaws, with audits already finding numerous vulnerabilities."

Agility Robotics | Salem, OR-based bipedal humanoid robotics company, founded 2015 as an Oregon State University spinoff | Going public via SPAC merger with Churchill Capital Corp XI at ~$2.5B valuation; would be the first pure-play humanoid robotics company on public markets.

"It would make Agility the first pure-play humanoid robotics company to trade on public markets, giving retail investors direct exposure to a sector that has so far been available primarily to deep-pocketed VC funds."

Figure AI | San Jose-based general-purpose humanoid robotics startup | Referenced as a valuation benchmark in the humanoid robotics funding frenzy.

"Figure AI self-reported that it closed on $1 billion in Series C funding at an eye-popping $39 billion valuation."

Apptronik | Austin-based humanoid robotics company for manufacturing and logistics | Cited as evidence of the capital boom in the sector.

"Apptronik closed a $935 million funding round valuing the company at more than $5.5 billion."

AI2 Robotics | Shenzhen-based maker of wheeled humanoid robots | Raised $735M at ~$3B valuation; illustrates the global scope of humanoid robotics investment.

"AI2 Robotics raised roughly $735 million at a nearly $3 billion valuation."

Even Realities | Shenzhen-based smart glasses startup, 3 years old | Raised $150M pre-Series B at $1B valuation; notable for camera-free design projecting info via optical displays, backed by Meituan and Tencent.

"Even Realities raised a $150 million pre-Series B round at a $1 billion post-money valuation."

Bespoke Labs | Mountain View, CA-based AI agent evaluation startup, 2 years old | Raised $40M in seed + Series A; builds realistic enterprise-like environments to train and evaluate AI agents for complex business tasks.

"Bespoke Labs builds realistic company-like environments where AI labs and enterprises train, evaluate, and improve agents for longer business tasks."

Skello | Paris-based workforce scheduling platform for hospitality and retail, 10 years old | Raised $228.4M led by Bridgepoint; addresses compliance-heavy frontline worker scheduling across Europe.

"Skello helps restaurants, hotels, and retailers schedule frontline workers under local labor rules, manage shift changes, and catch payroll issues."

Worldmodeldata | Cambridge, UK-based AI training data startup, 1 year old | Raised $9.3M seed; converts licensed video-game gameplay into structured physical-world training datasets for AI and robotics companies.

"Worldmodeldata turns licensed video-game gameplay into structured training datasets for AI labs and robotics companies building models that predict physical-world outcomes."

Scribe Therapeutics | Alameda, CA-based gene-editing biotech, 8 years old, backed by a16z | Filed for IPO; developing a PCSK9-targeting gene silencer requiring boosters only once every decade or two.

"Scribe Therapeutics is developing genetic medicines for common diseases... including a PCSK9-targeting gene silencer that could require booster shots only once every decade or two."

SK Hynix | South Korean memory chipmaker | Planning a U.S. IPO that could raise ~$28B, the largest potential raise mentioned in the issue.

"SK Hynix is planning to sell nearly 17.8 million shares in a U.S. IPO... the company could raise around $28 billion."

Syntiant | Irvine, CA-based edge AI chip company, 9 years old | Filed for Nasdaq IPO; makes low-power chips that run ML models directly on-device.

"Syntiant's low-power chips run machine-learning models directly on devices."

Alibaba / Qwen | Chinese tech giant / open-source AI model family | Case study in open-source monetization challenges at scale, with ~1M downloads/day but no clear revenue path.

"Alibaba's Qwen has become one of the world's most popular open-source AI model families, but the Chinese tech giant is struggling to turn that reach into revenue."

Threads (Meta) | Meta's social platform | Reached 500M monthly users by pivoting toward community-based content rather than direct X competition.

"Threads has grown to 500 million monthly users by evolving from Meta's would-be X killer into a Reddit-like home for communities around K-pop, sports, dating, books, and TV."

Katalyze AI | San Francisco-based pharma manufacturing AI, 4 years old | Raised $10.5M seed; connects lab and factory data for AI agents to investigate manufacturing issues and document regulated quality work.

"Katalyze AI connects pharmaceutical lab and factory data so AI agents can investigate manufacturing issues and document regulated quality work."

CurifyLabs | Helsinki-based pharmacy automation startup, 5 years old | Raised $14M Series A; automates compounding for patients who can't use standard drugs, including children and cancer patients.

"CurifyLabs automates pharmacy compounding for pharmacists producing custom doses for children, cancer patients, and others unable to use standard drugs."

4. People Identified

Peggy Johnson | CEO of Agility Robotics; former EVP of Business Development at Microsoft (engineered the $26B LinkedIn acquisition); former CEO of Magic Leap | Profiled as the operator navigating Agility's SPAC-based path to being the first public humanoid robotics company; notable for measured, cautious stance relative to sector hype.

"Johnson...was careful throughout our conversation. She declined to offer forward-looking financial guidance, declined to disclose the bill of materials for Agility's flagship robot Digit, and pushed back politely whenever questions veered toward speculation."

Masayoshi Son | Founder of SoftBank | Cited as a concentrated risk factor; his all-in AI bets across Arm, OpenAI, data centers, chips, and robotics are raising investor concern about SoftBank's over-dependence on one person's instincts.

"SoftBank founder Masayoshi Son has put the company at the center of the AI boom through Arm, OpenAI, data centers, chips, and robotics, raising investor concern that the sprawling Japanese group is again overly dependent on its founder's all-or-nothing instincts."

Elon Musk | Entrepreneur, owner of X | Lost appeal in a fraud case over Twitter acquisition; potential damages could reach $2.6 billion.

"Elon Musk lost his bid to overturn a jury verdict that found he defrauded Twitter investors by falsely casting doubt on the company's bot numbers...Damages could potentially reach $2.6 billion."

Asha Sharma | New CEO of Microsoft Xbox | Inheriting a "not healthy" gaming business; cutting 3,200 jobs (20% of staff) and divesting or spinning out five game studios.

"New CEO Asha Sharma tries to reset a gaming business she told employees is 'not healthy' after years of heavy spending and weak margins."

Simon Lancaster & Sabrina Paseman | Former Apple executives; co-founders of Omni Ventures | Launched a $33M fund targeting pre-seed manufacturing-tech startups at the intersection of AI, robotics, reshoring, and hardware automation.

"Omni Ventures, a new firm founded by former Apple executives Simon Lancaster and Sabrina Paseman, raised a $33 million fund to back pre-seed manufacturing-tech startups."

Gokul Rajaram | Prominent angel investor | Participated in Katalyze AI's $10.5M seed round, lending credibility to AI in pharmaceutical manufacturing.

"Katalyze AI raised a $10.5 million seed round led by Bonfire Ventures, with...Gokul Rajaram also chipping in."

Bryan Johnson | Anti-aging entrepreneur | Disclosed an incurable autoimmune condition, providing a note of irony given his public profile as a longevity optimization figure.

"Anti-aging guru Bryan Johnson posted on X that he has an incurable autoimmune condition that is leading his stomach to eat itself."

5. Operating Insights

Use SPAC Structures Strategically for First-Mover Public Market Positioning

In sectors where public market comps don't yet exist, being first to list — even via a non-traditional route — can be a genuine strategic advantage that justifies avoiding the traditional IPO process. Agility's rationale is a replicable playbook for category-defining companies in pre-IPO markets.

"Johnson said much of it boils down to the first-mover advantage the company enjoys when it's the first of its ilk to go public...The proceeds will also help Agility ramp up production at its 70,000-square-foot manufacturing facility in Salem, Oregon, and fulfill an existing pipeline of customer orders."

Enterprise AI Agent Evaluation Is Becoming Its Own Infrastructure Layer

As agents are deployed on complex, multi-step business tasks, the tooling to train, evaluate, and improve those agents in realistic simulated environments is emerging as a distinct and fundable infrastructure category.

"Bespoke Labs builds realistic company-like environments where AI labs and enterprises train, evaluate, and improve agents for longer business tasks."

Top-Tier Talent Is Bypassing Traditional Career Pipelines for Startup Bets

Operators hiring and competing for elite talent should expect accelerating attrition toward founding roles — AI has structurally changed the cost-to-build calculus for the best early-career candidates.

"Elite college students are skipping Wall Street, Big Tech, and even senior year to spend the summer in San Francisco hacker houses and incubators, betting that the AI boom has made startup-building a better career path than traditional internships."

6. Overlooked Insights

Video Game Gameplay as a Physical-World AI Training Dataset

Worldmodeldata is a one-year-old company that has identified a non-obvious data sourcing strategy: licensing video-game gameplay to generate structured training data for models that need to understand physical-world physics and outcomes. As robotics and embodied AI scale, proprietary training data sourcing strategies like this could become significant moats.

"Worldmodeldata turns licensed video-game gameplay into structured training datasets for AI labs and robotics companies building models that predict physical-world outcomes."

Idle Cash as a Consumer Product

Stoa, a London-based fintech raising at pre-seed, is threading together two typically separate consumer behaviors — earning yield on idle cash and unlocking brand rewards — into a single product. It's a quiet but potentially interesting model at a moment when high interest rates have made consumers more aware of cash opportunity costs.

"Stoa lets consumers and businesses place idle cash into fixed-term deposits that provide upfront perks from partner brands."