How Lindy 5X'd revenue by building for the next generation of AI models

Executive overview

Most AI startups chase the current model generation. Flo Crivello built Lindy by designing for where models are going, not where they are — accepting a broken product today in exchange for a strong one tomorrow.

The result: 5X revenue growth in six months with a team that only doubled. The core disciplines are prospect before you mine, anchor on unchanging user needs, and lead like a gardener setting direction without doing the growing.

Build for the next model generation, not the current one — a product that barely works today will be excellent when the models catch up.

Build for future models, not current ones

• If your first version works, you're not thinking big enough.
• Current frontier models should "barely support" your vision — that's the signal you're aiming far enough ahead.
• Lindy's earliest product was a plain prompt; it failed regularly because models weren't yet good enough.
• They kept building anyway, treating it as AI intern — useful for step-by-step tasks, future full AI employee.
• "Skate to where the puck is going to be" — design for the next model generation.

Prospect before you mine

• Two distinct startup phases: prospecting (finding the right market) and mining (scaling into it).
• Building a mine — hiring, raising money — before finding gold locks you in and is costly to undo.
• When Flo's prior company TeamFlow stalled after Covid, he fired two-thirds of the team rather than keep a mine without gold.
• He deliberately "mismanaged" the team during the pivot — letting operations slip to free his time for market exploration.
• When all competitors in a space are struggling, that's signal the market is gone: stop mining, prospect again.

Anchor on what doesn't change

• In fast-moving markets, find the stable pillars: user needs that are immutable.
• AI agent users always want something easy, effective, simple, affordable, and fast — that hasn't changed.
• Unexpected use case: HIPAA-compliant medical charting. Doctors used Lindy on mobile to record patient visits and auto-generate SOAP notes.
• User demand surfaced the use case — Lindy didn't design it top-down.
• Focus on the problem the user is trying to solve, not the technology available to solve it.

Leadership as gardening

• Half the CEO job is recruiting: "The people you hire is the company you build."
• Set direction — that's the second half. You can't grow for the plant, only guide it.
• Micromanagement is not optional: your authority exists because someone trusted your taste; exercise it.
• Write explicit standards. Flo spent a Sunday writing an 18-page "standard of excellence" doc for one new hire.
• The CEO's job is to win. Use that as the epistemological filter for every decision.

Definite optimism over vague bullishness

• Peter Thiel's definite optimism: have a crisp, specific view of what the future must look like — then build it.
• Indefinite optimism ("we don't know how, but it'll be fine") leads to drift, not vision.
• AI agents will turn everyone into managers of AI swarms rather than individual contributors.
• Long-term: any individual will be able to have the impact of a large organisation — a levelling of capital, network, and geography.
• The winning variable will be the quality of someone's vision and their ability to leverage agents.