Anthropic Unleashes a Version of Mythos

1. Key Themes

AI Model Deployment Is Entering a Controlled, Safety-Gated Phase

Anthropic's release of Claude Fable 5 signals that frontier AI labs are moving toward tiered public releases with built-in guardrails rather than open launches. The company is "blocking or deferring answers in high-risk areas like cybersecurity, biology, chemistry, and model distillation" and instituting a "mandatory 30-day data-retention policy for all Fable 5 and Mythos 5 traffic, which Anthropic says is needed to detect jailbreaks and other novel attacks." This sets a precedent for how powerful models will be commercialized going forward.

SPV-Based Access Is Disrupting Traditional Late-Stage Venture

A new model of VC is emerging that bypasses the traditional fund structure entirely. Justin Ernest of Sabertooth VC "used his network to secure allocations of stock in high-profile, later-stage companies" and offers "these individual deals to a group of about 30 smaller institutional investors using special purpose vehicles (SPVs), which act as single-deal funds." Over 12 months, the firm "invested nearly $400 million into 10 companies, including Anthropic, Anduril, Databricks, PsiQuantum, and SpaceX" — without launching a formal fund.

Defense and Dual-Use Tech Is Attracting Serious Capital

From AI-powered air defense (Alta Ares) to radar satellites (Iceye at an $11.6B valuation), defense-adjacent hardware is commanding large rounds from institutional investors. Iceye "raised a $520.5 million round at an $11.6 billion valuation" with participation from sovereign wealth (Qatar Investment Authority) and national infrastructure players (Nokia, Solidium), signaling a sustained institutional shift toward defense and dual-use technology.

AI Infrastructure Costs Are Spawning Radical Physical Innovation

The pressure of AI-driven data center demand is pushing infrastructure innovation to extremes. China "has begun operating what it says is the world's first wind-powered underwater data center, a 24-megawatt facility submerged off the coast of Shanghai that uses seawater for cooling and power from a nearby offshore wind farm, reducing electricity and freshwater demands as AI drives a surge in data-center construction." Meanwhile, SpaceX is "telling IPO investors that the company aims to begin testing orbital AI computing systems by late 2027," pointing to space as the next frontier for compute infrastructure.

Regulatory Pressure on Big Tech Is Intensifying Globally

The EU is actively reshaping how U.S. tech companies operate in Europe. Regulators rejected Apple's claim that DMA rules forced it to delay Siri AI, "saying Apple alone chose not to roll out the upgraded assistant." Separately, European regulators "ordered Meta to reverse a WhatsApp policy that effectively barred rival AI chatbots from using its business messaging API, saying the interim measure is needed to prevent competition from being lost before the bloc finishes its antitrust probe."

2. Contrarian Perspectives

"AI Layoffs" Are a Convenient Excuse for Poor Management — Not a Real Productivity Story

While the mainstream narrative frames current tech layoffs as AI-driven efficiency gains, Palantir co-founder Joe Lonsdale argues the opposite: "Many CEOs are using 'AI productivity' as cover for layoffs that really stem from overhiring, lowered standards, or missed growth targets during the 2021-to-2023 boom." Marc Andreessen has echoed this view. Companies including Block, Atlassian, and Coinbase are citing AI in workforce cuts, but the underlying cause may be balance-sheet correction from the prior bubble — not genuine AI-driven productivity transformation.

Skipping the Traditional Fund Structure Is a Feature, Not a Bug

Conventional wisdom holds that institutional credibility requires a formal fund vehicle. Sabertooth VC's Justin Ernest challenges this: rather than spending "12 to 18 months" launching a formal VC fund, he went straight to SPVs. In just 12 months, he deployed nearly $400M into 10 top-tier companies. His model works because "in the sometimes-shady world of small allocations and SPVs targeting family offices, he's earned a solid reputation" — suggesting trust and access, not fund structure, are the real currency.

Perplexity Is Deliberately Avoiding the AI IPO Herd

Rather than capitalizing on current AI IPO momentum, Perplexity CEO Aravind Srinivas is holding firm to a 2028 timeline regardless of market conditions. He is "still planning to go public in 2028, regardless of how the market receives the upcoming SpaceX, Anthropic, and OpenAI IPOs, though he acknowledged those offerings will shape investor appetite for AI listings." This is a contrarian bet that their business fundamentals — not market sentiment — should dictate timing.

3. Companies Identified

Anthropic

Description: AI safety company behind the Claude model family
Why mentioned: Released Claude Fable 5, a public version of its powerful Mythos model, with safety guardrails and a mandatory data-retention policy
Quote: "Anthropic released Claude Fable 5, the first public version of its powerful Mythos model, making the technology available through its API and enterprise plans while blocking or deferring answers in high-risk areas like cybersecurity, biology, chemistry, and model distillation."

Sabertooth VC

Description: SPV-based investment firm founded by Justin Ernest
Why mentioned: Case study in a new VC model that bypasses formal fund structures to give family offices access to top late-stage startups
Quote: "Over the last 12 months, his firm, Sabertooth VC, has invested nearly $400 million into 10 companies, including Anthropic, Anduril, Databricks, PsiQuantum, and SpaceX."

Iceye

Description: 12-year-old Finnish company operating a synthetic aperture radar satellite constellation
Why mentioned: Raised $520.5M at an $11.6B valuation; one of the largest defense/dual-use infrastructure rounds in the issue
Quote: "Iceye raised a $520.5 million round at an $11.6 billion valuation. It also sold $636 million in secondary shares."

Kalshi

Description: Prediction markets platform
Why mentioned: Taking proactive steps on market integrity by requiring employer disclosure from users making sensitive bets
Quote: "Kalshi plans to require users in certain prediction markets to disclose their employers, adding new guardrails for bets tied to sensitive information such as company performance and national security after an audit committee recommended stronger protections against insider trading and market manipulation."

Perplexity

Description: AI-powered search company
Why mentioned: CEO publicly committed to a 2028 IPO timeline independent of peer AI company offerings
Quote: "Perplexity CEO Aravind Srinivas says the AI search company is still planning to go public in 2028, regardless of how the market receives the upcoming SpaceX, Anthropic, and OpenAI IPOs."

SpaceX

Description: Aerospace and space technology company
Why mentioned: Pitching orbital AI computing as a long-term vision to IPO investors, dependent on Starship cost reductions
Quote: "SpaceX executives are reportedly telling IPO investors that the company aims to begin testing orbital AI computing systems by late 2027, using demonstrator satellites to validate a long-term plan for space-based data centers that depends heavily on Starship lowering the cost of moving hardware into orbit."

NinjaOne

Description: 13-year-old IT operations platform based in Austin
Why mentioned: Sold $400M in secondary shares at a $12.3B valuation — more than doubling its February valuation of $5B in roughly four months
Quote: "NinjoOne sold $400 million in a secondary share transaction at a $12.3 billion valuation. It raised a round at a $5 billion valuation in February."

Standard Bots

Description: 15-year-old New York company building AI-native industrial robot arms
Why mentioned: Raised $200M Series C at a $1B valuation; robots that learn through demonstration rather than code
Quote: "Standard Bots develops AI-native industrial robot arms that can learn manufacturing tasks through demonstration rather than code."

Beacon

Description: Three-year-old Toronto startup that acquires software businesses and integrates them into a shared operating platform
Why mentioned: Raised $225M Series C; $500M+ raised total — notable rollup model using AI to improve acquired software businesses
Quote: "Beacon acquires essential software businesses and integrates them into a shared operating system that automates engineering, back-office, and go-to-market workflows to improve performance."

Morpho

Description: Five-year-old Paris startup operating blockchain-based lending markets
Why mentioned: Raised $175M at up to $2B valuation with backing from Paradigm, a16z, Apollo, and Circle — blue-chip DeFi validation
Quote: "Morpho raised a $175 million round at a valuation of up to $2 billion. Investors included Paradigm, Ribbit Capital, Andreessen Horowitz, Apollo, Circle, and VanEck."

Meta

Description: Social media and technology conglomerate
Why mentioned: Subject to two separate EU regulatory actions; also expanding data collection practices
Quote: "Meta will start using people's activity on other websites and apps — including purchases and games they play — to personalize Facebook and Instagram feeds and AI responses, expanding a data-sharing system it already uses for ad targeting."

Stepful

Description: Five-year-old New York startup training individuals for entry-level healthcare roles
Why mentioned: Raised $55M Series C; addresses healthcare workforce shortage through online vocational training
Quote: "Stepful provides online training programs that prepare individuals for entry-level healthcare roles such as medical assistants, pharmacy technicians, and patient care technicians."

Alta Ares

Description: Two-year-old French AI air defense startup
Why mentioned: Raised $57.7M Series A for AI-powered drone/aircraft interception systems; signals European defense tech momentum
Quote: "Alta Ares develops AI-powered air defense systems to detect, identify, and intercept enemy aircraft and drones."

4. People Identified

Justin Ernest

Description: Founder of Sabertooth VC; former investor at Playground Global
Why mentioned: Built a $400M+ deployment track record in 12 months using SPVs instead of a formal fund
Quote: "Having spent over five years at Playground Global investing in deep tech and helping lead fundraising, Ernest was confident his connections to both investors and founders would allow him to bridge that gap."

Aravind Srinivas

Description: CEO of Perplexity
Why mentioned: Publicly committed to a 2028 IPO on Perplexity's own terms, independent of the AI IPO wave
Quote: "Perplexity CEO Aravind Srinivas says the AI search company is still planning to go public in 2028, regardless of how the market receives the upcoming SpaceX, Anthropic, and OpenAI IPOs."

Joe Lonsdale

Description: Co-founder of Palantir
Why mentioned: Publicly challenged the "AI productivity" layoff narrative, calling it cover for prior management failures
Quote: "Palantir co-founder Joe Lonsdale says many CEOs are using 'AI productivity' as cover for layoffs that really stem from overhiring, lowered standards, or missed growth targets during the 2021-to-2023 boom."

Marc Andreessen

Description: Co-founder of Andreessen Horowitz
Why mentioned: Named as echoing Lonsdale's view that AI-labeled layoffs often mask prior operational failures
Quote: "A view echoed by Marc Andreessen and others as companies from Block to Atlassian to Coinbase cite AI in workforce cuts."

5. Operating Insights

Network Access, Not Fund Structure, Is the Real Competitive Moat in Late-Stage VC

For emerging managers, the Sabertooth model suggests that proprietary access and trust are more valuable than formal fund mechanics. Ernest bypassed 12–18 months of fund formation by leveraging relationships built at Playground Global. The key differentiator: "In the sometimes-shady world of small allocations and SPVs targeting family offices, he's earned a solid reputation." Operators building investment vehicles should invest in relationship capital before institutional infrastructure.

Safety Architecture Is Becoming a Go-to-Market Differentiator for AI Companies

Anthropic's decision to implement domain-specific blocks (cybersecurity, biology, chemistry) and a mandatory 30-day data retention policy is not just risk management — it is a selling point for enterprise buyers in regulated industries. The pattern: "blocking or deferring answers in high-risk areas" while maintaining API and enterprise plan availability suggests that responsible AI deployment frameworks will increasingly be a competitive advantage, not a constraint.

AI Degree Proliferation Signals a Credential Gap Worth Exploiting

The rapid explosion of AI academic programs — "at least 74 majors and 89 minors now available on U.S. campuses, up from just five AI majors in 2021" — is outpacing employer validation. For operators hiring AI talent, this means credentials alone are an unreliable signal. Building internal assessments or partnering with bootcamp-style programs (like Stepful in healthcare) may yield better talent outcomes than traditional degree-based recruiting.

6. Overlooked Insights

Secondary Markets Are Becoming a Primary Liquidity Mechanism at Scale

NinjaOne's $400M secondary transaction — at a valuation 2.5x higher than its primary round just four months earlier — signals that secondary share sales are no longer a niche exit tool. They are becoming a core capital markets instrument for late-stage private companies managing cap table pressure and early-investor liquidity without triggering a full IPO process. The $636M in secondary shares sold alongside Iceye's primary round reinforces the same pattern. Investors and founders should be modeling secondary transactions as a deliberate part of their liquidity strategy.

Blockchain Lending Is Attracting Traditional Finance at Scale

Morpho's $175M round at up to a $2B valuation drew investment from Apollo and Circle alongside crypto-native firms Paradigm and a16z. The inclusion of Apollo — a $650B+ AUM traditional asset manager — alongside VanEck signals that institutional finance is beginning to treat on-chain credit infrastructure as a legitimate asset class, not a speculative bet. This is an early signal of a deeper convergence between DeFi infrastructure and traditional credit markets that warrants closer attention from fintech investors.