How Wing VC separates tracks from trains to predict which AI companies will endure | Jake Flomenberg

Wing Venture Capital Partner Jake Flomenberg opens every IC meeting at Wing with a question that didn't used to matter: if a frontier model ships something meaningfully better, does this company still have reason to exist? In this episode of BUILDERS, he walks through the trains vs. tracks framework, the three diagnostic tests he runs on every AI company, and what a real flywheel looks like versus a pitch deck concept. Topics Discussed: Why "enduring differentiability" is now Wing's central IC question — and why early ARR no longer answers it

Trains vs. tracks: which you're building determines your entire sales motion and pricing model

"Blast radius" as a diagnostic to stress-test your own switching cost before a buyer does

Three questions that reveal whether an AI company will compound or compress as frontier models improve

Grayswan: a real self-reinforcing red team loop from a Carnegie Mellon AI safety team

Category creation in the AI era — why announcing on day one is usually wrong

GTM Lessons For B2B Founders: Know whether you're building a track or a train — then sell accordingly. Trains sit on existing infrastructure: easy to adopt, easy to remove. Tracks require full re-architecture to replace because workflows and accountability structures get built around them. Jake's diagnostic: what's the blast radius of removing your product? Affects a couple of people's workflows — train. Changes how the whole org makes decisions — track. The mistake isn't building either — it's misrepresenting which you are. AI agent infrastructure sold as "just sign up and try it" understates the gravity of the decision. An AI writing tool locked into a multi-year contract overclaims stickiness that isn't there. Ask whether frontier models are fuel or a threat — before you fundraise. Jake calls it the most underappreciated question in AI: does a better base model make you stronger or more replaceable? Durable companies treat frontier models as fuel — their assets and value appreciate as base models improve. When a new model ships, is your differentiation compressing or compounding? If compressing, rethink before a better-informed investor asks. The flywheel test: answer all three specifically, or don't claim you have one. (1) How does your product get better as it operates? (2) How quickly? (3) How does the customer experience that improvement? Vague answers mean aspirational, not operational. The Grayswan contrast: 15,000+ red teamers generate attack trajectories that train a custom attacker LLM called Shade, which stress-tests a runtime firewall called Signal until nothing gets through. Each layer compounds. That's what operational looks like. Race to be last, not first. The companies that matter in ten years are racing not to be replaced — which means building, through close design partner work, a story that solves the problem today and five years out, not whack-a-moling whatever surfaces this quarter.

// Sponsors: Front Lines — We help B2B tech companies launch, manage, and grow podcasts that drive demand, awareness, and thought leadership. www.FrontLines.io

Don't Miss: New Podcast Series — How I Hire Senior GTM leaders share the tactical hiring frameworks they use to build winning revenue teams. Hosted by Andy Mowat, who scaled 4 unicorns from $10M to $100M+ ARR and launched Whispered to help executives find their next role. Subscribe here: https://open.spotify.com/show/53yCHlPfLSMFimtv0riPyM