20VC: “Anthropic Wipes Billions Off Markets: Citrini Research — The Ultimate Breakdown: Agents, Ghost GDP, Consumer Spend”

TL;DR

Three investors wrestle with AI’s impact on software valuations in real-time. The Fortnite circle metaphor wins: AI shrinks every software company’s territory incrementally, and the only defense is owning the dominant agent in your category. Ghost GDP — productivity that accrues to equity holders, not consumers — is the right concern on the wrong timeline. Momentum has been the only working investment strategy, but Atlassian at -75% with accelerating revenue is the best value dislocation on the board. The PE-backed SaaS graveyard is heading for forced Frankenstein roll-ups. And Jack Altman giving up solo GP independence for Benchmark tells you something about how hard 2026 is going to be for everyone who isn’t at the top of the food chain.

Executive Summary

The weekly 20VC roundtable delivers its sharpest episode in recent memory. Three seasoned investors — Rory O’Driscoll (Scale Venture Partners), Jason Lemkin (SaaStr), and Harry Stebbings — take on the Anthropic security release that vaporized $20 billion from cybersecurity stocks, the Citrini research paper predicting a “Ghost GDP” crisis by 2028, Figma’s strong earnings in the teeth of AI disruption, and whether momentum or value investing wins in the current panic. The episode is strongest when the three disagree — Rory’s disciplined macro skepticism against Jason’s bleeding-edge anecdotal evidence, with Harry pressing both. It’s weakest when it retreats to “everything is disrupted” hand-waving without specificity. But the central tension — AI is creating immense productivity that accrues to fewer and fewer humans — is the right question for 2026, and they wrestle with it more honestly than most.

Pod Details

• Podcast: The Twenty Minute VC (20VC)
• Host: Harry Stebbings
• Guests: Rory O’Driscoll (Scale Venture Partners), Jason Lemkin (SaaStr)
• Date: February 26, 2026
• Format: Weekly roundtable — three investors debating the week’s biggest tech and market news
• Dynamic: This trio has genuine chemistry. Rory plays the value-oriented, macro-skeptical institutional investor who demands logical rigor. Jason plays the operator-turned-investor living on the bleeding edge, citing real-time product experiences. Harry plays the sharp moderator who knows when to press and when to let the argument breathe. The disagreements are productive, not performative.

Theme 1: The Fortnite Circle — AI Shrinking Every Software Company’s Island

Claude is like Fortnite. Your territory keeps shrinking. Claude keeps consuming more and more of you, and you’re stuck in this smaller and smaller island.”

Jason Lemkin’s Fortnite metaphor is the episode’s most useful framework. Every B2B software company’s addressable surface area is shrinking as foundation models consume more of what they do. This isn’t binary disruption — nobody gets “killed” overnight. It’s incremental maiming. CrowdStrike doesn’t disappear because Claude can run security audits; it just loses the part of its value proposition that was static code review. DocuSign doesn’t vanish; but an agent that auto-contracts for a business takes enough value that growth decelerates from 31% to 20%. And when you’re trading at 16x revenue, the difference between 31% and 20% growth is catastrophic.

The strongest evidence is Jason’s own experience: he ran an enterprise-grade security audit on a plane last week using Claude Code inside Replit, and the Replit team didn’t even know their own product had gotten that good yet. When the builders can’t keep up with what their tools can do, the incumbents have no chance of keeping pace.

Rory’s counterpoint is important: intelligence will diffuse into the enterprise through existing software companies, not around them. The four channels are direct from foundation models, enterprise-built, incumbent-integrated, or startup-built. He bets on channels three and four. But Jason’s response cuts deeper: if you look at every publicly traded B2B company, only Palantir has a competitive agent that’s actually moved revenue. Everyone else is sprinkling AI dust on dashboards.

So what: The investable question isn’t “will AI disrupt software?” — it’s settled. The question is: does your company own the agent in its category? If yes, you’re Palantir. If no, you’re watching the circle shrink.

Theme 2: Ghost GDP — Productivity That Buys Nothing

Our 10 agents at SaaStr generate millions of revenue but they buy nothing. Nothing, nothing except tokens.”

The Citrini research paper positing a “2028 Global Intelligence Crisis” gets properly stress-tested here. Rory correctly frames the analysis: first decide whether the micro disruption claims are credible (will AI replace DoorDash? Amex? coding?), then — and only then — graduate to the macro implications. If you don’t believe the adoption timeline, ignore the macro entirely.

Jason’s Ghost GDP example is the episode’s most concrete data point. His team went from 12 people to 2, generating eight figures of revenue. That productivity gain is real and measurable. But those 10 former employees aren’t spending money on handbags, Netflix, or t-shirts. The GDP shows up in corporate profits and market caps. It doesn’t show up in consumer wallets.

Rory pushes back with 200 years of economic history: productivity gains have always been good, always freed people for higher-value work, always created more wealth than they destroyed. 80% of Americans used to farm; now 4% do, and we’re so productive we’re fat. The onus, he argues, is on the doomsayers to explain why this time is different.

The honest answer — which nobody quite states — is speed. Previous productivity revolutions took decades to displace workers. If AI displaces millions of knowledge workers in 2-3 years, the labor market can’t absorb them fast enough. Jason’s Claude analysis: 50% tech headcount reduction would mean 600-900 billion in GDP impact, 4-5 million jobs lost including multiplier effects, and local economic devastation in 5-6 cities. That’s not an abstraction. That’s San Francisco, Austin, Seattle, and New York.

So what: Ghost GDP is the right framing for the wrong timeline. The productivity concentration is real and accelerating. The question isn’t whether it happens but whether the transition takes 3 years (crisis) or 15 years (manageable restructuring). Rory is right about the long arc of history. Jason is right about the pace of this particular moment.

Theme 3: Momentum vs. Value — The Investor’s Existential Crisis

Here’s the greatest dislocation if I look at the public stocks, right? Klaviyo versus Shopify.”

The most practically useful segment is the raw investing debate. Jason has shifted from bargain-hunting to pure momentum: buy only stocks that are up over 12 months (Palantir, Figma, MongoDB, Cloudflare, Shopify). Rory, the institutional value investor, acknowledges that momentum has been the only strategy that’s worked in both public and private markets for the last 18 months — but warns that trees don’t grow to the sky.

The Klaviyo-Shopify dislocation is the episode’s sharpest specific insight. Klaviyo is essentially a derivative of Shopify — nearly 100% revenue-attached. Yet Shopify is up 2.63% over the year while Klaviyo is down 58%. The bargain hunter buys Klaviyo. But Rory’s counterpoint is devastating: if Shopify is going to survive, Toby has to build agents that cannibalize what Klaviyo does. The cheap stock is cheap because the adjacent competitor has to kill it to survive.

Then Atlassian: down 74.85% in a year despite accelerating revenue growth (20% to 23% at $6.3 billion). Against Palantir, up 41%. Rory calls it “the greatest dislocation in the market” — the most beaten-down stock with the most revenue acceleration. By any value metric, Atlassian is the best buy on the board.

The PE-backed SaaS graveyard is the quiet horror story underneath: companies growing 6% with massive debt leverage that can’t be serviced. Jason predicts 5-8 startups at $50-200M revenue will get mashed together at 1-2x revenue into Frankenstein B2B conglomerates. It’s the best idea anyone has, and it’s still a miserable outcome.

So what: In a momentum market, value investors look stupid until the inflection point, and nobody rings a bell at the top. Atlassian at -75% with accelerating revenue is the episode’s best actionable insight for anyone willing to be early and patient.

Theme 4: The Benchmark Play — What Jack Altman’s Move Says About 2026

He gave up the dream of 95% of folks stuck in venture.”

Jack Altman raised $275M for AltCap, then shut it down to join Benchmark as a partner. Jason’s read is the most interesting: this is a man who had what 95% of venture capitalists dream of — a solo fund with complete autonomy — and gave it back. Even the management fees on $400M are a perfectly comfortable life. Something about 2026 made the solo path less attractive than joining an institution.

Rory reads it as Benchmark’s long-running playbook: they don’t grow talent, they pick proven talent, offering equal partnership with autonomy. It’s worked for 20 years.

The unspoken implication: the solo GP model is getting harder as AI raises the cost of entry and compresses the time horizon on which investments can go sideways. Having Benchmark’s brand, deal flow network, and institutional support is increasingly valuable when the pace of change means last month’s investment thesis might already be obsolete.

So what: When a GP with proven alpha voluntarily gives up independence for institutional backing, it’s a signal about market conditions, not just personal preference.

Action Plan

1. Audit your portfolio for Fortnite circle exposure (This week). For every B2B software investment or holding: does this company own the dominant agent in its category? If yes, hold. If no, quantify how much surface area AI has consumed in the last 12 months. If the answer is “I don’t know,” that’s your answer.

2. Run the Rory framework on any macro panic piece (Always). Micro first, macro second. Before graduating to “civilization is ending,” establish whether you actually believe the adoption timeline. If you don’t believe DoorDash gets disrupted in 18 months, you can ignore the GDP doom scenario that depends on it.

3. If you’re investing in public SaaS, study the Klaviyo-Shopify dislocation pattern (This month). Look for stocks that are cheap-because-derivative, where the parent company has existential incentive to cannibalize the derivative’s value. That’s a value trap, not a bargain. Conversely, look for the Atlassian pattern: beaten down with accelerating revenue and no direct cannibalizer.

4. Watch the PE-backed SaaS space for forced consolidation (Next 6-12 months). The Frankenstein roll-up thesis — 5-8 companies at $50-200M mashed together at 1-2x revenue — is Jason’s best non-obvious call. Where there’s forced selling, there’s asymmetric opportunity for buyers with patience and operational skill.

5. Ask the Ghost GDP question about your own business (Today). Jason went from 12 to 2 people generating eight figures. Where is the productivity gain going? If it’s accruing to equity holders and not creating downstream consumer spending, you’re part of the problem and the opportunity simultaneously.

80/20 rule: The Fortnite circle framework. If you internalize one thing: every software company’s territory is shrinking. The only question is how fast, and whether your company owns the agent.

Follow-Up Questions I Wish Were Asked

1. Jason says every B2B software product is “terrible now” compared to AI-native alternatives. But most enterprise software buying is driven by procurement cycles, security reviews, and integration requirements — not product quality. When has “the better product” actually won in enterprise software? Is there a scenario where the incumbents win precisely because they’re entrenched, not despite it?

2. The Ghost GDP argument assumes displaced workers don’t find new employment. But the episode itself describes an explosion of AI-native startups generating massive revenue with tiny teams. If 10 people leave Jason’s team and 3 of them start AI-native companies, is the GDP actually “ghost” or just redistributed with a lag?

3. Rory says momentum investing works for 6-18 months, then value takes over. Given that AI disruption is accelerating the clock on everything else, is it possible that the momentum-to-value transition also happens faster — meaning the window for momentum plays is shorter than historical patterns suggest?