Week 2 of Summer Launchpad gave the founders a sharper lens on one of the hardest parts of building: figuring out what's actually worth building, and for whom.
Workshop: The Discovery Ladder Level 1
The first major session of the week introduced founders to the Discovery Ladder: a framework for structuring customer research as a sequence of levels, each de-risking the next. Level 1 surfaces whether a problem is real. Level 2 maps the friction inside that problem. Level 3 tests prototypes. Level 4 runs pilots. Skip a level, and you compress all your unknowns into a single high-stakes experiment.
The session devoted particular attention to five ways pilots teach you nothing: handing over a login with no expectations set, demoing features instead of listening, leaving meetings without a mapped approval chain, treating usability as a proxy for value, and closing on a verbal yes before money is in the bank. Each one has a fix, and all five are avoidable.
Pricing came up as a consistent blind spot. The right price isn't what feels comfortable to charge, it's the most a customer will pay even while complaining about it. Investors, the cohort was reminded, consistently push founders to charge more.
On the tooling side, coaches walked founders through an upgraded DRM workflow: transcripts uploaded to Notion, tagged by customer segment and discovery level, then fed into Claude to score each interview against the team's specific problem statement. The system flags completeness gaps, surfaces key quotes, and recommends follow-up questions, updating automatically when the problem statement evolves. The output is a dashboard: primary ideal customer profile (ICP), pain-level interview counts, disqualified segments, and hygiene flags for the interviews themselves.
Workshop: Using AI to Test with Customers with Dan Maccarone
With roughly 25 years in tech and six AI startups launched in the last three years alone, Dan Maccarone has seen the cycle play out more times than most. His diagnosis of the current moment was blunt: building has never been easier. That's exactly the problem.
Maccarone opened with a confession: he's launched six AI startups in the last three years, integrated Claude and ChatGPT into his design process, and still thinks most founders are using AI wrong. The tools are great, but they're usually solving the wrong problem. Shipping faster means nothing if you haven't validated what you're shipping. "Just because I can," he asked the room, "should I?"
The bulk of the session was a practitioner's masterclass in customer discovery. Maccarone pushed back on the assumption that more interviews equals better data, patterns emerge after five or six conversations, and 15 to 20 is his ceiling. What matters more is who you're talking to and how you structure the conversation. His preferred format is an inverted triangle: start broad (daily life, general habits), and only slowly narrow toward the relevant domain. Never lead with what you're building. Avoid focus groups, where one personality tends to hijack the room.
On using AI in research: Maccarone now uses Claude to synthesize interview transcripts, write screeners and discussion guides (tasks that used to take a full day, now taking minutes), and rapidly prototype lo-fi flows for usability testing. The prototype pipeline he described, experience brief, Claude-coded lo-fi, five pilot users, feedback loop, real development, compresses a sprint from weeks to days. Black-and-white prototypes on purpose: remove the design distraction, focus the feedback on function.
What AI can't do, he was equally clear about: conduct the interview, find the right participants, or craft the product strategy. "AI can help write the document," he said, "but it cannot craft the strategy itself."
We leave you with this question: Just because I can, should I?
Stay tuned for more updates from Summer Launchpad and if you want to be in the room where it happens, come find us at the Leslie eLab.
