Citrini Research Breakdown: Agents, "Ghost GDP", Consumer Spend | Figma Earnings Beat

Anthropic's Security Release and Market Reaction [0:07]

• The release of Anthropic's security review feature led to a significant market drop, wiping out billions from cybersecurity stocks like Cloudflare and CrowdStrike.
• There's a debate whether this was an overreaction, as similar code auditing and penetration testing capabilities already exist on platforms like Replit.
• The speed of AI advancement is so rapid that weekly podcasts can become dated quickly.
• Despite existing capabilities, the market panics, indicating a sensitivity to AI's disruptive potential, even if some of the "news" is months old.

"When you are priced for perfection, anything less than perfection will be a kick in the nuts."

AI's Disruption in B2B Software and Enterprise Adoption [0:33]

• Most existing B2B software is perceived as "terrible" due to the advancement of AI software.
• The adoption of AI capabilities like code scanning is expected to diffuse into the enterprise, likely through incumbents like CrowdStrike.
• Companies in the security sector were trading at prices assuming perfection, making them susceptible to negative AI-driven news.
• The core argument is that AI capabilities will be integrated into existing platforms, rather than completely replacing them, but the pace of change is the primary concern.

"The greatest dislocation. If I look at the public stocks, right, ready to go."

The Value Accrual to Agents and Terminal Decline [0:46]

• The value is increasingly accreting to the "agentic layer" of AI, which can automate complex workflows.
• Companies that do not own their agents are at risk of "terminal decline" as their value proposition shrinks.
• Even established companies like DocuSign, while not fully destroyed by AI, could be "maimed" by agents that automate contract processes.
• The comparison is made to a shrinking island (Fortnite game analogy) where AI (Claude) consumes more territory, forcing companies into smaller market share.

"Claude keeps consuming more and more of you and um and you're stuck in this smaller and smaller island that you got to own more and more market share of our ser almost everyone that's public your surface area is shrinking."

Paths for AI Intelligence in the Enterprise [10:36]

• Intelligence will infuse all software over the next decade, generated by foundation models.
• Four ways AI intelligence can reach the enterprise: direct purchase from models, internal development, buying from incumbents integrating AI, or buying from numerous new companies leveraging foundation models.
• The consensus is that enterprises will likely buy integrated solutions (path 3) or solutions from new companies (path 4), rather than building entirely on foundation models directly or having foundation models build all niche systems.
• This creates an opportunity for a "software mediation layer" between foundation models and the enterprise.

"Intelligence will infuse all software over the next decade. That intelligence is generated by foundation models like Claude and like OpenAI."

The Rise of Agentic Companies and Disruption of Incumbents [11:53]

• Publicly traded B2B companies have not seen significant revenue acceleration from their AI agents, unlike companies like Anthropic that have built impactful agents.
• Privately held, recently founded companies are experiencing explosive revenue growth by leveraging foundation models.
• The core question is whether this value will accrue solely to model companies or to these new specialized companies.
• Platforms like Replit and Lovable are potentially vulnerable to disruption as foundational models like Cloud Code launch similar features, enabling users to build and visualize apps directly within their environment.

"It is not sprinkling AI dust on top of their analytics software. There's no there's not enough value there."

Challenges in Deploying AI Agents for Enterprises [14:47]

• Deploying effective AI agents requires significant work: custom training, onboarding, and data cleansing.
• This is a substantial effort for companies already feeling overworked, creating institutional momentum to overcome.
• A shortage of forward-deployed engineers and technically skilled workers capable of training and tuning agents exacerbates the challenge.
• For companies like Shopify or Monday.com with diverse verticals, building universally effective agents is extremely difficult, making them targets for niche startups.

"Every agent is essentially custom. Every agent needs to be trained. Every agent needs to be onboarded if you want it to be great."

The "Ghost GDP" Concept and Economic Impact [35:33]

• "Ghost GDP" refers to productivity gains not translating to widespread consumer spending because wealth concentrates among fewer people (e.g., a team reducing from 12 to 2 people generating the same revenue).
• This concentration of wealth means fewer people have disposable income to spend on goods and services, potentially leading to an economic slowdown.
• While historically productivity gains have been beneficial, the current AI-driven scenario raises concerns about short-term economic dislocation and a softening of consumer spend due to wealth concentration.
• Japan in the 1990s is cited as a precedent where massive productivity increases did not broadly benefit the population.

"Go GDP is that this productivity is not going to human workers that then spend it. That's the fear that we are creating this productivity, but it is not going to any there's no humans to spend the money."

AI's Impact on Software and Investment Strategies [40:40]

• AI agents are buying millions of tokens but nothing else, unlike past economic drivers.
• Productivity increases are only beneficial if consumer wallets are dispersed and spending power is maintained.
• The pace of AI diffusion is faster than previously anticipated, with AI-driven security audits and better-than-expected capabilities emerging rapidly.
• Most B2B software is now considered "terrible" and dated due to the rapid advancements in AI, putting pressure on incumbents.

"Almost all the B2B software we use today is terrible now. It's terrible."

The Pace of AI Diffusion and Potential for Disruption [44:06]

• AI diffusion, while often underestimated in speed, can also be faster than expected, as seen with AI-driven security audits.
• The rapid improvement and expansion of AI capabilities mean that even products considered safe from disruption may be at risk.
• The current acceleration of AI is making existing B2B software products seem outdated and struggling to keep pace.
• While some macro-level arguments about civilization ending are dismissed as clickbait, the micro-level impact on B2B software companies and potential job displacement is a valid and crucial discussion.

"But God this acceleration it's it's just it's hard to it it's it's so fast."

Figma's Earnings and Defensive AI Integration [01:00:06]

• Figma reported strong Q4 2025 earnings with accelerating growth (40% year-over-year) and high customer retention (97% GRR, 136% NDR).
• The company's CEO is actively integrating AI capabilities to move from design to coding, aiming to defend against AI-native competitors.
• This represents a "winning" strategy in SaaS, particularly in creative and co-generation fields where AI disruption is already present.
• The debate is whether Figma's efforts will be enough to stave off disruption from AI models that can generate equally elegant designs.

"This is what winning looks like in SAS in a sector where it's very credible, very credible, perhaps more credible than most that an AI native product could disrupt you."

Momentum Investing vs. Value Investing in AI's Flux [01:07:00]

• In the current uncertain market, a momentum investing strategy—betting on stocks that are already winning and showing positive price action—is being employed.
• This contrasts with a value investing approach, which relies on fundamental analysis of undervalued assets.
• The challenge lies in identifying when the market transitions from a momentum-driven phase to a value-driven one.
• Companies like Atlassian, despite recent declines, show accelerating growth, presenting a potential "dislocation" where a beaten-down stock is improving.

"In the short term, over kind of 6 months to in a public market, 6 months to 18 months, momentum plays work, value plays don't."

OpenAI's Ambitious Spending and Future Bets [05:14]

• OpenAI is planning to double its spending to $665 billion by 2030, while increasing its revenue forecast to $280 billion based on future products.
• This aggressive expansion into hardware, ads, and other areas relies on significant anticipated future revenue, indicating a strong leaning into the future.
• The "fear-mongering" narrative around AI's power is what enables companies like OpenAI and Anthropic to secure funding, as investors want to back potential market dominators.
• Despite competition, OpenAI is still considered a winner in the consumer space for mind share, though it faces challenges in enterprise adoption and developing new monetization strategies beyond subscriptions.

"As long as these companies are perceived as so powerful that they can destroy everything, they will be able to get money."

Consolidation in Venture Capital and Solo GP Dynamics [01:15:16]

• The move of Jack Altman from Alt Cap to Benchmark signifies a consolidation trend within venture capital, with firms attracting proven talent.
• Altman's decision to return a large fund and join Benchmark suggests a strong belief in the firm's offering and a desire to leverage its established platform.
• This move highlights the ambition of many General Partners (GPs) who dream of managing significant capital with autonomy, a dream Altman gave up to join a larger entity.
• The decision underscores the value placed on brand, team, and the potential for collective success over individual independence in the current venture landscape.

"I think it was a clever move by a very shrewd firm. I think it's been their MMO for 15 20 years which is we have a very compelling offering to make to any GP."