Anthropic Accidentally Exposes Claude Code's Code

---

1. Key Themes

AI Seed Valuations Have Entered a New Regime

The benchmark for seed rounds has shifted dramatically upward, driven almost entirely by AI hype and real early traction from breakout companies. The new normal is no longer modest — it's compressed timelines and inflated expectations.

"Today, 'it's pretty typical' to see a $10 million seed round at a $40 million to $45 million post-money valuation, especially if you are an AI company. Actually, that type of thing happens only if you are an AI company, as investors are showing little interest in anything else."

Outlier AI Traction Benchmarks Are Distorting the Entire Market

A handful of breakout AI companies have set new velocity standards that are now being applied to the broader category, even though these are statistically exceptional cases.

"Shanea Leven, founder of the enterprise AI application platform Empromptu, blames Cursor, which, in early 2025, hit $100 million in revenue in just 12 months. It was one of the first high-profile AI companies to raise the bar for how fast these startups could gain traction... Though these are outliers, it's hard for some not to feel the reverberated heat."

Defense Tech Is Attracting Serious Capital at Extraordinary Multiples

Autonomous military systems are commanding some of the largest and fastest-growing valuations in the private markets, rivaling top-tier AI infrastructure companies.

"Saronic, a four-year-old Austin startup that makes autonomous surface vessels for the U.S. military, raised a $1.75 billion round at a $9.25 billion valuation, more than doubling its valuation from a round it raised last year."

Supply Chain Vulnerabilities in Open-Source Software Are a Growing Attack Surface

Two separate incidents involving npm — one accidental, one malicious — highlight that widely trusted open-source infrastructure is increasingly a high-value target for both espionage and sabotage.

"Anthropic accidentally exposed the inner workings of Claude Code by publishing an internal source map file to npm... rivals and researchers now have a detailed look at how Claude Code's coding agent handles memory, orchestration, and background tasks."

"Hackers linked to North Korea hijacked the widely used Axios JavaScript library on npm by compromising a maintainer's account and slipping in malware that could give remote access to any systems that installed the poisoned update."

Quantum Computing Is Accelerating Crypto's Existential Risk Timeline

The threat to blockchain cryptographic security is materializing faster than anticipated, with implications for Bitcoin holders, crypto infrastructure investors, and enterprises running legacy encryption.

"Google researchers warned that future quantum computers may be able to crack the elliptic-curve cryptography used across crypto with about 20 times less hardware than previously estimated, adding pressure on Bitcoin and other blockchain communities to move faster toward post-quantum defenses."

---

2. Contrarian Perspectives

Large VC Firms Moving Upstream Are Destroying Seed-Stage Economics for Smaller Funds

The conventional narrative is that more capital flowing into AI is good for the ecosystem. The reality is that big-fund incursion at the seed stage is crowding out specialized, smaller investors — suppressing deal count even as headline valuations rise.

"The big venture firms, flush with cash, are also moving into rounds earlier, driving up startup prices and valuations... Smaller VC firms have an insatiable appetite for AI companies, too. As an investor focused on AI infrastructure, Smith said she can easily find herself priced out of a round, especially when a larger firm moves in. That's one reason why seed deal count is down but valuations are up."

Investors Are Pricing AI Startups Years Ahead of Actual Traction

The market is operating on narrative and optionality, not fundamentals. Even investors who acknowledge this dynamic are participating — a classic signal of late-cycle behavior in a specific sector.

"Even with those early revenue numbers, Smith said investors in this market are pricing rounds 'years ahead of traction.'"

Microsoft's AI Buildout Is Stalling, Not Scaling — and Markets Are Starting to Notice

The consensus view has been that hyperscalers will be the reliable winners of the AI era. Microsoft's results challenge that thesis, pointing to the possibility that enterprise AI adoption is slower and more expensive than expected.

"Microsoft just logged its worst quarter on Wall Street since 2008, with the stock down 23% as investors fret over the payoff from its AI buildout, weak Copilot adoption, and the growing cost of supporting both Azure demand and its own AI products."

---

3. Companies Identified

---

4. People Identified

---

5. Operating Insights

Early Revenue Is Table Stakes, Not a Differentiator — Velocity Is the New Moat

At the current YC Demo Day, eight-week-old companies were already showing six- to seven-figure customer contracts. For founders, the implication is that go-to-market speed and early enterprise sales must be compressed dramatically. For investors, this resets the baseline for what "early traction" means.

"Many startups had already landed six- to seven-figure customer contracts, including a company that was only eight weeks old... there were companies asking for $5 million at a $40 million post money."

The "Human-in-the-Loop" Model Is Becoming a Scalable Business Architecture

Multiple funded companies this issue — Crosby (law), Tenex.ai (cybersecurity), and Jimini Health (mental health) — share the same operating model: AI handles volume, humans handle judgment and liability. This hybrid approach is generating institutional investment across highly regulated verticals.

"Crosby runs a law firm where AI reviews commercial contracts and human lawyers do a final legal check." / "Tenex.ai has built an AI-enabled managed detection and response service designed to find, track, and respond to security threats with human oversight."

Strategic Partnerships at Funding Stage Are Becoming Valuation Accelerants

Also (Rivian spinout EVs) and DoorDash announcing a partnership simultaneously with a funding round — and DoorDash participating as an investor — signals a growing pattern where distribution partnerships are being structured as investment instruments.

"Also raised a $200 million round at a $1 billion post-money valuation... The company also announced a partnership with DoorDash to work together on autonomous deliveries" with DoorDash as an investor.

---

6. Overlooked Insights

Open-Source Package Registries Are Now a Primary Attack Vector for Nation-State Actors

The North Korea/Axios incident is not an isolated event — it represents a pattern of state-level actors targeting trusted developer infrastructure to achieve massive silent distribution of malware. This has implications for any organization with dependencies on public npm packages, and signals a likely sustained campaign.

"Hackers linked to North Korea hijacked the widely used Axios JavaScript library on npm by compromising a maintainer's account and slipping in malware that could give remote access to any systems that installed the poisoned update."

Yupp's Shutdown Suggests "Selling Picks and Shovels to AI Labs" Is Not a Defensible Business If Labs Internalize the Need

Yupp raised $33M to resell user preference data to AI labs — but the model collapsed in under a year. As foundation model companies build their own RLHF and preference infrastructure, third-party data brokering to labs may be structurally unviable without a durable supply-side moat.

"Yupp, a two-year-old San Francisco startup that let users compare outputs from 800 AI models and resold that user preference data back to AI labs, is shutting down less than a year after launch. It raised a $33 million seed led by Chris Dixon at Andreessen Horowitz."