The hottest running app has nothing to do with speed | E2303

What It Covers

Two separate segments: Louis Phillips demos Interval, a territory-claiming gamified running app with 1 million downloads built by a team of five, then Alice Zhang of Verge Labs explains how the company pivoted from developing ALS drugs to licensing AI-powered patient-matching tools to pharma giants like Eli Lilly and AstraZeneca.

Key Questions Answered

• •Gamification over speed: Interval deliberately excludes speed metrics, allowing anyone — including walkers — to compete for territory by encircling geographic areas. This design choice removes the intimidation factor present in apps like Strava, where elite athletes dominate leaderboards. The result is broader participation across fitness levels, making territory size and run frequency the competitive vectors rather than pace.
• •Organic social before paid ads: Interval reached 100,000 Instagram followers and 1 million downloads using zero paid media initially. The content format that drove growth was founder-led talking-head videos explaining the game mechanic with on-screen overlays. Founders should be willing to appear publicly imperfect — Louis Phillips credits willingness to "fail publicly" as the core unlock for organic reach and early user acquisition.
• •Meta ad economics for fitness apps: Interval's current cost per trial start on Meta is $12, with an annual subscription price of $60 USD and an average customer lifetime of 17 months, yielding roughly $85 in lifetime value per paid user. They use a third-party agency called Scale on a spend-scaled fee model to manage campaigns, adding predictability to what was previously an algorithm-dependent, volatile download cycle.
• •Brain tissue as AI training ground truth: Verge Labs built a dataset from over 12,000 human brains across 6,000 deceased patients, sourced from 24 tissue banks globally. This tissue provides molecular-level disease data unavailable from blood or imaging alone. Alice Zhang compares it to LiDAR in self-driving: without this direct sensor, AI models trained only on proxy data like blood biomarkers produce lower-accuracy disease representations.
• •Transformer architecture for incomplete patient data: Verge Labs uses a contrastive multimodal learning architecture — developed within the last 18 months — that retains three data signals: what blood data reveals independently, what brain tissue reveals independently, and the synergy between both. Traditional ML required complete datasets per patient. This approach fills missing data modalities, enabling a virtual brain biopsy reconstructed from a single blood draw.
• •Pharma partnership structure and validation rates: Verge Labs structured deals with Eli Lilly ($706M total milestones) and AstraZeneca Alexion ($700–800M milestones each), with $25–42M upfront payments. Lilly internalized two AI-derived ALS drug targets in 2024 — the first such targets in their pipeline. Critically, 83% of Verge's proposed targets validated in wet lab experiments, against Lilly's pre-partnership expectation that 20% validation would already exceed expectations.