9 Videos From CVAI London: Cohere's Aidan Gomez, Index Ventures' Danny Rimer, Recursive's Josh Tobin & More
1. Key Themes
European AI Sovereignty Is Becoming an Investment Thesis
Geopolitical anxiety is driving a distinct European AI investment wave. Sequoia's Luciana Lixandru said "she's found ample opportunities to invest due to European leaders' realization that they can no longer rely on the good will and protection of the United States of America." Cohere's Aidan Gomez similarly "cautioned against the risk of the Western world relying too heavily on American foundation models for their AI capacity." Defense tech is a direct beneficiary: "Sequoia just co-led a €500 million round in Stark Defense, a startup making kamikaze drones and other weapons."
Trust-First GTM Is the Enterprise AI Playbook
Enterprise AI adoption hinges on credibility before capability. Eleanor Lightbody (Luminance) and Neil Zeghidour (Gradium) "both agreed that earning customers' trust that your AI tools are accurate and consistent is key to earning top-quality deals. Once you build that trust and gain a foothold, it then becomes possible to train models off of a company's internal data." The sequencing is deliberate: trust unlocks data access, which deepens moat.
Prosumer-to-Enterprise Is a Recurring Upgrade Path
Multiple AI companies are discovering that consumer-adjacent launches are a Trojan horse into enterprise budgets. Alex Mashrabov (Higgsfield) and Grant Lee (Gamma) "both started out aiming at a 'prosumer' market. But they quickly saw the enterprise demand for AI creative tools, and are both pushing heavily into enterprise solutions." They are "optimistic that a lot of content creation for go-to-market teams can be fully automated in the near future."
Compute Demand Has No Ceiling in Sight
Recursive's Josh Tobin pushed back on any near-term AI infrastructure cooldown: "Tobin doesn't think there's going to be a demand crash for AI compute anytime soon; if anything, he said, it's only going to get worse." The driver is recursive self-improvement research — training models to train themselves — which implies compounding, not plateauing, resource requirements.
Forward-Deployed Engineers as Enterprise AI Transformation Agents
OpenAI is running a Palantir-style embedding model to accelerate legacy enterprise transformation. Arnaud Fournier's team of "forward deployed engineers at OpenAI have been embedding in legacy businesses to teach existing leaders what AI can do to transform their business practices." A concrete example: his "work with Thrive Holdings to integrate AI into accounting firm rollups."
2. Contrarian Perspectives
AGI Is Already Here Against the prevailing debate about when AGI will arrive, Cohere's Aidan Gomez — a co-author of the foundational "Attention Is All You Need" paper — stated that "effectively we've already reached AGI." This is a notably bold claim from someone building enterprise AI infrastructure, not a frontier lab racing for the title.
Europe Is Faster at AI Adoption Than Silicon Valley Assumes The conventional wisdom is that European enterprises are slower and more risk-averse around AI. Decagon's Jesse Zhang directly contradicts this: "contrary to the prevailing wisdom in Silicon Valley, many established business leaders in Europe have been eager to adopt AI customer-support tools rather quickly" — a signal for founders considering international expansion.
Backing Multiple Competing AI Labs Is Not Contradictory Most investors feel pressure to pick sides in the foundation model wars. Index Ventures' Danny Rimer pushed back: "We're in the early innings," he said, "and there's room for more than one winner." This justifies a portfolio-of-labs strategy rather than concentrated bets, at least at this stage of the market.
3. Companies Identified
| Company | Description | Why Mentioned | Quote |
|---|---|---|---|
| Cohere | Enterprise-focused large language model provider | CEO Aidan Gomez keynoted; discussed AGI claims and European AI sovereignty risks | "Effectively we've already reached AGI" |
| Luminance | Legal tech AI platform | CEO Eleanor Lightbody spoke on building enterprise trust as a GTM strategy | "Earning customers' trust that your AI tools are accurate and consistent is key to earning top-quality deals" |
| Gradium | Voice model developer | Founder Neil Zeghidour co-paneled on trust-first enterprise sales | Same trust-building framework as Luminance |
| Higgsfield | AI video generation company | Founder discussed prosumer-to-enterprise pivot; optimistic on content automation | "Both started out aiming at a 'prosumer' market. But they quickly saw the enterprise demand" |
| Gamma | AI-generated slide/presentation maker | Founder Grant Lee co-paneled on same prosumer-to-enterprise shift | "A lot of content creation for go-to-market teams can be fully automated in the near future" |
| Decagon | AI customer-support tooling | Founder Jesse Zhang discussed faster-than-expected European enterprise adoption | "Contrary to the prevailing wisdom in Silicon Valley, many established business leaders in Europe have been eager to adopt AI customer-support tools rather quickly" |
| Recursive | AI self-improvement research company (ex-OpenAI/Google/Meta founders) | Josh Tobin explained the recursive self-improvement thesis and compute demand trajectory | "There's not going to be a demand crash for AI compute anytime soon; if anything, it's only going to get worse" |
| Stark Defense | Defense startup making kamikaze drones and weapons | Case study for European defense AI investment | "Sequoia just co-led a €500 million round in Stark Defense" |
| Thrive Holdings | Accounting firm rollup | Named as a live enterprise AI integration use case via OpenAI's forward-deployed team | "Work with Thrive Holdings to integrate AI into accounting firm rollups" |
4. People Identified
| Person | Description | Why Mentioned | Quote |
|---|---|---|---|
| Aidan Gomez | CEO of Cohere; co-author of "Attention Is All You Need" | Keynote speaker; made AGI-already-here claim and European sovereignty warning | "Effectively we've already reached AGI" |
| Luciana Lixandru | Partner at Sequoia Capital | Articulated European defense/sovereignty as an investment opportunity | "European leaders' realization that they can no longer rely on the good will and protection of the United States of America" |
| Arnaud Fournier | Forward-deployed engineer lead at OpenAI; Palantir veteran | Described OpenAI's enterprise embedding model for legacy business transformation | "Embedding in legacy businesses to teach existing leaders what AI can do" |
| Eleanor Lightbody | CEO of Luminance | Spoke on trust as the key GTM unlock for enterprise AI deals | "Earning customers' trust that your AI tools are accurate and consistent is key to earning top-quality deals" |
| Neil Zeghidour | Founder of Gradium (voice models) | Co-paneled on trust-first enterprise sales approach | Same panel thesis on trust as prerequisite to internal data training |
| Alex Mashrabov | Founder of Higgsfield | Discussed prosumer-to-enterprise arc in AI creative tools | "Quickly saw the enterprise demand for AI creative tools" |
| Grant Lee | Founder of Gamma | Co-paneled on the same prosumer-to-enterprise pivot | "Content creation for go-to-market teams can be fully automated in the near future" |
| Danny Rimer | Partner at Index Ventures | Offered hiring framework and multi-lab investment thesis | "We're in the early innings… there's room for more than one winner" |
| Jesse Zhang | Founder of Decagon | Challenged Silicon Valley assumptions about European enterprise AI adoption | "Contrary to the prevailing wisdom in Silicon Valley, many established business leaders in Europe have been eager to adopt AI customer-support tools" |
| Josh Tobin | Founder of Recursive; ex-OpenAI/Google/Meta | Explained recursive self-improvement thesis and compute demand outlook | "It's only going to get worse" (re: compute demand) |
| R. Martin Chavez | Sixth Street Chairman; Alphabet board member; former Goldman Sachs CIO | Drew analogy from algorithmic trading to AI agents; shared pain-point sales framework | "Customers buy a product when they have unbearable pain and you have convinced them that only your software product can put an end to their pain" |
5. Operating Insights
Hire for Capability, Taste, and Agency Danny Rimer distilled his hiring framework for founders simply: "evaluate hires on capability, taste, and agency." In an AI environment where generalists can be augmented by tools, agency — the ability to act without being directed — may be the most differentiated filter.
Pain-Centric Sales Is the Only Reliable Path to a Sale R. Martin Chavez relayed a durable sales principle from Goldman's head of sales: "Customers buy a product when they have unbearable pain and you have convinced them that only your software product can put an end to their pain. Everything else is just getting lucky." For AI founders navigating a crowded market, this is a positioning imperative — own a specific, acute pain point, not a broad capability.
Trust Unlocks Data, Data Deepens Moat For enterprise AI operators, the sequencing matters: establish trust in accuracy and consistency first, then use that access to train on proprietary internal data. As the article notes, "once you build that trust and gain a foothold, it then becomes possible to train models off of a company's internal data" — creating a compounding competitive advantage that generic models can't replicate.
6. Overlooked Insights
Accounting Firm Rollups as an AI Integration Beachhead Arnaud Fournier's work at OpenAI specifically targets "accounting firm rollups" through Thrive Holdings as a venue for AI integration. This is a non-obvious vertical — professional services rollups are structurally fragmented, data-rich, and ripe for workflow automation. It signals that PE-backed professional services consolidation plays may be a sleeper AI deployment opportunity.
Recursive Self-Improvement as a Research Category Spun Out of Big Labs Josh Tobin and his co-founders left "big AI labs like OpenAI, Google, and Meta" specifically to pursue recursive self-improvement research — models that learn how to train themselves. This framing positions the next frontier of AI not as better base models, but as self-optimizing systems, a category largely unexplored by incumbents and potentially foundational to the next compute demand wave.