The Twenty Minute VCAnthropic Files to Go Public | Cognition Raises $1BN at $26BN Valuation | The 996 Work Ethic
CHAPTERS
- 0:12 – 6:39
Anthropic’s IPO filing and the psychological reset for founders, employees, and VCs
The group debates whether Anthropic going public is healthy for the ecosystem or whether it creates a demoralizing “nothing else matters” effect. They unpack how trillion‑dollar outcomes compress timelines and distort incentives for talent, founders, and investors.
- •Anthropic’s speed to IPO at massive scale resets expectations for how fast “mega outcomes” can happen
- •VC and employee motivation shifts when the perceived bar becomes $10B–$1T outcomes
- •Rory argues you can’t build a strategy around repeatedly finding once‑a‑decade winners
- •IPO transparency could reduce mystique and help the ecosystem “move on”
- 6:39 – 15:41
The “billion-dollar position” era: concentrated investing and higher bars for meetings
Jason explains his new filter: he’s less interested unless an investment can become a billion‑dollar position, not just a good outcome. Harry and Rory challenge the feasibility of predicting such outcomes, and they discuss how this raises the meeting bar for founders.
- •Jason’s shift: optimize for billion‑dollar positions, not merely billion‑dollar companies
- •How to identify blockers: founder/CTO caliber, ambition, TAM expansion thinking, and non-complaining cultures
- •Harry’s counterpoint: biggest winners are often underestimated, so strict upfront filters can miss upside
- •Rory’s caution: risk feels easy late in booms; the reality of drawdowns changes behavior
- 15:41 – 21:25
Public markets rush: OpenAI vs Anthropic timing and the capital “queue jump”
They argue that elite AI companies are accelerating IPO timelines because the next era is capital intensive. Rory frames it as a mad dash to the front of public markets to secure massive equity capital for AI CapEx.
- •Shift from ‘staying private is cool’ to ‘go public now’ to raise enormous capital
- •Google’s $80B raise as signal: even cash-flow giants are becoming CapEx-heavy
- •AI builds convert formerly CapEx-light businesses into cash consumers—often a headwind for valuations
- •Equity issuance as strategic timing: high stock price makes dilution tolerable
- 21:25 – 30:02
Is the SaaS apocalypse over? Sector rebound, but fundamentals still bifurcate
They review the strong earnings week and SaaS multiple rebound, calling the panic overdone. However, they argue the core pressures remain: seat contraction, budget diversion to AI, and a split between AI-attached winners and “human seat” laggards.
- •SaaS bounced sharply from oversold levels; ETFs show a big snapback but not a full leadership return
- •AI spending up sharply implies cuts elsewhere, often in traditional per-seat software
- •Winners are those that re-accelerate and/or directly benefit from agents (Twilio, Datadog, Okta examples)
- •Classic per-seat SaaS growth is structurally pressured even if stocks recover from the ‘hard deck’
- 30:02 – 32:47
Cognition raises $1B at $26B: why autonomous AI engineers could be the killer category
The conversation shifts to Cognition/Devin and why autonomous engineering agents may be more consequential than code assistants. They acknowledge rapid leadership changes in the space and the risk that trillion‑dollar incumbents will attack every winner.
- •Autonomous ‘AI engineer’ vision: delegate work, return commits—more impactful than incremental copilot gains
- •Early adopters among top CTOs validate the direction even if early models were mediocre
- •Valuation debates: comparing Devin/Cognition economics to Cursor-style multiples
- •Competitive intensity: category leaders can change quickly as incumbents ‘eat your lunch’
- 32:47 – 42:09
Token budgeting panic: CFOs discover the bill and companies start capping usage
They analyze the sudden corporate realization that token spend exploded after permissive Q1 adoption. The group argues this is validating for model providers (a new spend category is locked in), but budgeting discipline and caps will reshape usage patterns.
- •Corporate ‘unlimited card’ moment: token bills show up in accruals and shock finance teams
- •Uber-style caps (per user/month) as an early control mechanism
- •Despite panic, companies aren’t saying ‘stop’—they’re reallocating budgets to fund AI
- •Frontier models aren’t getting cheaper, but today’s frontier becomes tomorrow’s cheaper tier
- 42:09 – 1:00:21
Tokens vs humans: budget trade-offs, layoffs-by-tokens, and the crucial percentage mix
Jason predicts a coming budgeting framework where leaders choose between headcount and token allocation—driving cuts in “bubble roles” like QA and parts of CS. Rory pushes for a more quantitative, realistic view: what share of engineering spend becomes tokens will determine long-term TAM and labor impact.
- •A new budgeting model: fixed EPD budget, leaders allocate between humans and tokens
- •Potential cuts in QA/CS and other marginal roles to fund tokens for top performers
- •Rory’s key question: 10% vs 33% vs 100% token-to-salary ratios radically change outcomes and TAM
- •Productivity bottlenecks may shift to productization, sales enablement, and go-to-market capacity
- 1:00:21 – 1:12:56
Big Law’s AI arms race: Kirkland’s $500M pledge and build vs buy dynamics
They debate whether Kirkland & Ellis building in-house AI is truly disruptive or mostly signaling. The discussion expands to when vertical industries should build proprietary AI to protect “secret sauce” versus buying horizontal tools, and why full-stack AI law firms are unlikely to replace top firms.
- •Kirkland’s spend is small relative to revenue; may be PR + optionality more than existential vendor threat
- •Build-vs-buy logic: build only if AI encodes proprietary differentiation and workflows
- •Caution on vendor conflict: AI providers can’t credibly threaten to become full-stack competitors
- •AI likely expands access to legal services for consumers/SMBs while high-end work still demands humans
- 1:12:56 – 1:21:05
Robinhood’s AI agents: financial planning automation vs creating trading alpha
They distinguish between AI improving financial planning (goal-based allocation, risk profiling) and the harder problem of beating markets. The central promise is that agents could make every user an ‘expert’ in decision-making, even if they can’t reliably generate alpha.
- •Best use case: personalized planning, scenario analysis, and discipline—not stock-picking outperformance
- •Robinhood’s demographic tension: long-term planning vs trading-as-entertainment
- •Agent autonomy vs education: should the AI teach users or execute on their behalf?
- •Product principle: AI should make every customer an expert in the domain (example: Klaviyo-style guidance)
- 1:21:05 – 1:25:00
Apollo warns PE software returns will be ugly: leverage math and long holds
Rory explains why PE equity outcomes get crushed when software multiples compress and leverage sits above equity in the stack. They discuss how overpaying for slower-growth SaaS leaves few levers besides time, cost cuts, and bolt-on acquisitions.
- •If debt is stressed, equity is worse—multiple compression plus leverage is punishing
- •Mature SaaS lacks ‘accelerants’ to grow out of an overpaid entry price
- •Likely outcomes: long holding periods, bolt-ons, grinding to mediocre multiples (1.2x–1.3x)
- •LP incentives vs GP incentives: firms with real skin-in-the-game are more likely to grind for better exits
- 1:25:00 – 1:30:27
Mega distributions and VC firm dynamics: retention, retirement, and motivation after huge wins
They explore what massive carry payouts (Anthropic, SpaceX) do to venture firm structures and individual motivation. The consensus: big liquidity events reveal preferences—some will retire, others keep playing—and the key issue is whether future funds can feel meaningful after a generational win.
- •Huge distributions can cause partner turnover; humans react to life-changing money differently
- •Some firms may rationally wind down after big wins rather than compete in a new AI era
- •Jason’s view: it must be ‘10x intellectually’ to justify another multi-decade cycle
- •Rory’s counter: reinvest personally (be an LP too) to restore meaningful upside—Peter Thiel model
- 1:30:27 – 1:38:59
The 996 work ethic: startup intensity, performative culture, and the required quid pro quo
They debate whether 996 is real, new, or mostly performative, agreeing intensity has always existed in startups and Big Law. The key nuance is fairness: extreme expectations need a credible path to outsized outcomes, and leaders must avoid burnout-driven bad judgment.
- •Startup and Big Law intensity isn’t new; the question is how deep it goes across the org and for how long
- •Quid pro quo: seven-day weeks only justify if early employees get real upside (potentially ‘eight figures’)
- •Risk of performative overwork: loss of judgment and degraded decision quality
- •Irony of the era: people claim AI will automate work soon, yet top teams are working harder than ever
