20VC

Harry Stebbings × Parag Agrawal

Discover’s fate changed when it moved one screen left beside friends’ Stories, showing that interface placement can outweigh a year of design work.Parag Agrawal argues that zero-to-one products require instinct before data, while separate revenue and innovation teams can prevent short-term incentives from blocking experimentation.At Match, success may mean never seeing a user again, making attention against TikTok and games the broader contest and Tinder’s AI photo selection a potential onboarding catalyst.

Harry Stebbings × Matteo Franceschetti

AI is letting Eight Sleep scale output: engineers direct hundreds of AI engineers, while email marketing generates close to $100M with a team of zero and paid media produces $200–300M with two people.Watch whether this productivity supports the three-year target of 250 people making a billion, amid talent retention, Anthropic’s cost curve, regulation, and model self-learning risks.

Harry Stebbings × Andrew MacDonald

Uber’s AI efficiency story is less about line-item ROI than tighter headcount constraints: a 30-person team cut pricing allocation from 15 hours to 2 and marketing QA from two weeks to two days.MacDonald calls autonomy existential and Uber One a compounding consumer lever, while cheaper modes, the pending Delivery Hero deal, and AV dollar-share concentration shape the next growth and competitive tests.

Harry Stebbings × Mike Mignano

With OpenAI, Anthropic, and xAI building the infrastructure layer, Mike Mignano sees the application layer entering its value-creation window, though the buildout is not finished.Labs may lose the app layer to regulatory clearance, organizational context, and focused distribution, while startups can maximize token spend and exploit the 70% of markets left beyond the winner.The key variables are recursive self-improvement versus model commoditization, agent trust, routing economics, energy demand, and whether fund size matches repriced ownership and cash-on-cash math.

Harry Stebbings × Josh Browder

Josh Browder’s one-man accelerator combines sub-$5M pre-seed entry, founder cohabitation and hands-on screening, producing 33 deals in fund four and a Micro 1 return well over a thousand X.His broader signals are that framing can flip funding outcomes, he thinks Anthropic will get to a trillion in revenue, and Nevada land serves as his hedge against both post-economic abundance and tech going to zero.

Harry Stebbings × Cliff Weitzman

Speechify expects token spending to exceed engineering salaries next year, with great companies reaching that point in 3 years.It tests 1,000 AI-generated ads daily plus 8,000 human creatives monthly on proprietary tooling and is among 200 companies provisioned for OpenAI ads.Single-digit inference costs versus Eleven Labs’ $70–100 support profitability, but undisclosed valuation and the OpenAI-Anthropic contest remain key watchpoints.

Harry Stebbings × Jake Paul × Geoffrey Wu

Anti-Fund’s core bet is that attention, cultural fluency, and response-generating access will appreciate as AI commoditizes coding and financial analysis.Its $30 million vehicle targets roughly ten exceptional late-stage companies, with an ambition to manage $10–$20 billion and extend the distribution advantage into public markets.

Harry Stebbings × Andrew Dudum

Hims & Hers is down 66% in six months to a $4.35B market cap on over $2.3B of revenue, while Andrew Dudum says weight loss remains one of a dozen clinical categories, not the core business.Its consumer-platform model is reshaping pharmaceutical distribution, using lower-priced GLP-1 access, AI-enabled productivity, and loss-leading diagnostics to build a broader healthcare front door; retention decay remains the key risk to monitor.

Harry Stebbings

AI-native firms below $100M ARR averaged -126% FCF margins versus -56% for non-AI, making burn multiple useful but vulnerable to churn, CapEx and runway distortions.Public software’s valuation reset remains unresolved: 30%-growth companies trade around 15–20× revenue versus a historical 6–7× NTM benchmark.Fivetran’s dbt acquisition, at roughly $400M and $100M ARR, signals consolidation’s potential to turn a crowded unicorn backlog into IPO-scale businesses.

Harry Stebbings

GPT-5’s underwhelming launch shifts attention from AGI speculation to software economics, with pricing 8–10x below Anthropic’s high-end token loads pressuring coding revenue and benefiting Cursor.AI CapEx is rewarding co-attached suppliers as Datadog posted a record $260M net-new ARR quarter yet fell 10%, while Palantir’s 45% growth and $843M US commercial bookings confront a roughly 120x-revenue valuation.