The Twenty Minute VCCognition vs Factory | Anthropic Under Threat | ElevenLabs Doubles Its Valuation to $22B
CHAPTERS
- 0:00 – 1:33
Panel setup and preview of the week's AI dealflow
The episode opens with a rotating panel featuring Dev Ittycheria joining the regular hosts. They preview a packed agenda spanning OpenAI vs. Anthropic metrics, a high-profile hiring/ethics controversy, open-model geopolitics, and multiple M&A and valuation debates.
- •Dev Ittycheria joins as special guest (MongoDB CEO)
- •Agenda: OpenAI/Anthropic revenue, Factory vs Cognition controversy, Salesforce acquisition, Reflection open model, more
- •Tone set: fast-moving news cycle and strong opinions
- 1:33 – 3:10
OpenAI’s reported $70B run rate and sudden enterprise momentum swing
The panel reacts to reports that OpenAI has re-accelerated dramatically and is nearing a $70B run rate. They discuss how quickly model competitiveness and pricing can flip developer and enterprise usage—especially given OpenAI’s distribution via ChatGPT.
- •OpenAI’s pace of iteration at scale is described as "crazy" and decisive
- •Developers and startups are actively switching models based on quality/price
- •ChatGPT’s massive usage is a distribution advantage
- •Enterprise embedding (contracts, workflows, training) may create switching friction
- 3:10 – 8:27
If Anthropic’s Q3 isn’t a blowout: TAM vs rotating share
A central question emerges: is OpenAI’s resurgence coming directly at Anthropic’s expense, or are both growing within an expanding market? They unpack prior deceleration signals, the narrative impact of Q/Q re-acceleration, and why Anthropic’s Q3 print matters with an IPO looming.
- •OpenAI previously showed deceleration (18% Q/Q) vs Anthropic’s rapid growth
- •A reported 60–70% Q/Q OpenAI re-acceleration changes sentiment dramatically
- •Key uncertainty: share rotation vs both companies expanding overall usage
- •Anthropic’s IPO timing implies Q3 numbers will be visible in the prospectus
- 8:27 – 9:58
Valuation reality check: OpenAI at $1.4T and the liquidity vs illiquidity trade
They shift from revenue to pricing: a $30B raise at a $1.4T valuation is framed as unprecedented private-market territory. Rory emphasizes that, in uncertain information environments, liquidity can be a decisive advantage versus multi-year private lockups.
- •$1.4T would be a first-of-its-kind trillion+ private venture round
- •Discussion separates company value from liquidity risk
- •Rory leans toward public-market liquidity when information is sparse
- •Private-market holding periods can easily extend 3–4+ years despite expectations
- 9:58 – 18:26
Factory vs Cognition hiring controversy: advisor access, conflicts, and trust
The panel dissects the allegation that a close Factory board observer/advisor interviewed and then became CRO at competitor Cognition. They debate the ethical line: when advisory roles include sensitive roadmap/strategy access, switching sides can look like a breach even if not legally barred.
- •Context: Chris Degnan allegedly advised Factory closely then joined Cognition
- •Dev: if an advisor had confidential insight, moving to a competitor is deeply problematic
- •Jason: many CROs see job moves as pragmatic, especially if not a fiduciary board member
- •Core nuance: light advising vs "inside the tent" strategic access
- 18:26 – 22:36
Has “loyalty” changed in the AI era? Talent raids, reputational damage, and the “rule of two”
The discussion broadens into whether the AI arms race is changing norms around executive moves and team lift-outs. Jason argues competitive movement is accelerating; Dev argues reputation and relationship dynamics still matter and can create long-term career costs.
- •Jason: AI speed increases executive mobility; norms are shifting
- •Dev: people/reputation fundamentals persist; reputational damage accumulates
- •Debate over employees vs advisors: scope of access and obligations differ
- •Jason describes the old “rule of two” (limited team follow-on) as eroding
- 22:36 – 25:13
Vinod Khosla’s tweet: venture etiquette, competitive investments, and public chastising
The panel reacts strongly to a Khosla tweet perceived as attacking Factory despite Khosla leading its round. They frame it as a self-inflicted reputational wound for the firm and an example of how fragile trust becomes when VCs back direct competitors.
- •Tweet seen as giving rival firms a weapon in fundraising and recruiting
- •Theme: praise in public, criticize in private
- •Competitive investments require extreme message discipline
- •General consensus: the tweet was unhelpful for Khosla and both companies
- 25:13 – 31:48
Reflection’s Beam and the enterprise push for US-based open models
They evaluate Reflection’s Beam as a strategic attempt to deliver near-frontier US open weights as an alternative to Chinese open models. The panel argues cost, sovereignty, and procurement constraints may push enterprises toward a “best model per workload” portfolio approach.
- •US open models could satisfy regulated/sensitive enterprise needs vs Chinese options
- •Cost efficiency (3–4x cheaper) becomes compelling as token budgets scale
- •Most workloads don’t require absolute frontier intelligence
- •Dreamforce anecdote: executives overloaded by internal token demand
- 31:48 – 34:49
Open-source token share forecast: could US open models reach ~50%? Safety and benchmarks
The hosts speculate about how open-source token usage may split between US and Chinese models within a year. They emphasize sovereignty concerns, the practicality of model swapping, and the rising need for credible safety evaluation frameworks despite benchmark gaming.
- •Jason floats the possibility of US open models capturing up to half of open-source tokens
- •Even 10–20% would be meaningful due to enterprise-heavy token volumes
- •Dev highlights safety as a key enterprise requirement; defining safety is hard
- •Benchmarks/evals are increasingly distrusted and can be gamed
- 34:49 – 38:36
Investment Committee: ElevenLabs at $22B—moat, margins, and voice as agent infrastructure
The panel roleplays an IC decision on investing at ElevenLabs’ $22B valuation. Jason argues for a concentrated bet, citing widening lead, strong margins, low-latency voice performance, and non-fungibility of quality in production voice workflows.
- •Proposal: invest heavily (up to 10% of a fund) despite the high valuation
- •Voice is positioned as core infrastructure for agentic applications
- •Quality and latency are critical; failures break the product experience
- •Debate over model substitution risk vs ElevenLabs’ multi-tier model coverage
- 38:36 – 44:46
Salesforce buys Listen Labs for $2B: AI market research as a top app-layer use case
They analyze Salesforce’s acquisition of Listen Labs, framing market research as a large, fragmented category ripe for LLM disruption. Rory argues conversational/adaptive surveying and synthesis is a natural LLM strength; Jason questions whether a small revenue asset moves the needle for Salesforce.
- •LLMs improve survey adaptiveness vs rigid legacy questionnaires
- •Market research is a big spend category with prior large outcomes (Qualtrics/Medallia/etc.)
- •Jason questions strategic impact at Salesforce scale (~$50–60B revenue)
- •Dev frames many AI apps as “feature vs franchise,” making early exits rational
- 44:46 – 49:10
Feature vs franchise—and the founder dilemma: take the $2B exit or build a generational company?
The panel debates whether founders should accept large early acquisition offers. They explore the personal/time cost of long journeys, NPV logic, and Dev’s experience receiving repeated acquisition interest before ultimately selling BladeLogic after going public.
- •Early exits can be dramatically easier than multi-decade company-building
- •Founders must assess whether they’re truly building a generational company
- •Dev recounts multiple acquisition conversations (including with Cisco) before eventual sale
- •Key tradeoff: expected trajectory and personal appetite for “running the tape”
- 49:10 – 59:04
Vercel at $600M ARR: agents driving growth and the new ‘agent recommendation’ economy
They examine Vercel’s metrics—agents becoming ~50% of new business—as evidence that “agents pick the winners.” The panel reframes go-to-market strategy around being discoverable, trusted, and integrable for agents rather than solely optimizing for Google-era human search.
- •Agents as buyers: shifting from page-one Google optimization to agent visibility and credibility
- •Agent-driven expansion can be a “force of nature” once embedded
- •Startups face a disadvantage vs incumbents due to smaller online corpus
- •Emerging category: tools that measure and shape how brands/products appear to LLMs/agents
- 59:04 – 1:05:33
Nvidia licensing deals lawsuit (Groq): de facto acquisitions, cap table fairness, and regulatory workarounds
They discuss former engineers suing over an Nvidia licensing/hire deal that allegedly hollowed out Groq while disadvantaging certain common holders. The panel frames it as a collision between longstanding Delaware shareholder-equality principles and modern deal structures used to bypass slow M&A approval.
- •Claim: licensing + talent transfer functioned as an acquisition without equal treatment
- •Delaware principle: same-class shareholders should be treated the same
- •Dev: employees may start demanding protections for “IP/talent carve-out” exits
- •Jason: if courts validate the challenge, these structures may become too risky to use
- 1:05:33 – 1:10:46
Meta’s Muse vs OpenAI Dots: shipping cures everything, but product clarity matters
The panel praises Meta’s Muse as a strong execution and product launch, shifting prior skepticism into credit for delivery. In contrast, they find OpenAI’s Dots unclear and underwhelming in messaging, though Jason argues it may win if it becomes a persistent, developer-centric agent tightly integrated with Codex.
- •Muse is framed as frontier-quality and a meaningful Meta execution win
- •Dots is criticized for unclear value proposition and a tech-centric demo narrative
- •Possible Dots wedge: persistent always-on coding agent integrated with Codex
- •Broader theme: hard to be consumer, developer, and enterprise-forward simultaneously
- 1:10:46 – 1:15:51
Oura pulls its IPO: pricing expectations, banker dynamics, and what it signals for venture liquidity
They close with concern about Oura withdrawing its IPO despite strong growth, interpreting it as a pricing mismatch and a negative signal for venture liquidity. The panel discusses how banker optimism can lead to failed processes and argues that sometimes companies should “hit the bid” to secure the window.
- •IPO pull viewed as bad news for venture liquidity and late-stage exit pathways
- •Likely cause: price mismatch between expectations and investor demand
- •Banker incentives can overpromise on pricing during pitches
- •Lesson: when markets are fragile, securing the deal may beat holding out for perfection
