Building One with Tomer Cohen

Some builders don't just create breakthrough products.

They create entirely new ways of thinking about how innovation happens.

In this episode of Building One, Tomer Cohen sits down with Astro Teller, Co-founder and Captain of Moonshots at X, Google's legendary moonshot factory.

Astro has helped build one of the world's most remarkable innovation engines—an organization responsible for projects like Waymo, Google Brain, Wing, Taara, Google Glass, Loon, and many more.

But this conversation isn't just about ambitious technology.

It's about building a system that makes ambitious technology more likely.

At X, innovation isn't treated as inspiration or creative genius. It's treated as a discipline. Teams are rewarded for disproving their own ideas, attacking the hardest assumptions first, and learning faster than everyone else. Success isn't measured by how long a project survives—but by how quickly you discover whether it deserves to.

In this episode, Tomer and Astro discuss:

Why X calls itself a "moonshot factory"—and what it takes to build innovation as a repeatable system

Why the goal isn't to prove your ideas right—but to discover when they're wrong

Why 10x thinking can actually be easier than incremental improvement

The famous "teach the monkey before you build the pedestal" framework for attacking risk in the right order

Lessons from Waymo, Google Brain, Taara, Google Glass, and Loon—and why some of X's biggest "failures" produced its greatest insights

How culture, incentives, and organizational design determine whether breakthrough ideas survive

If you've ever wondered why some organizations consistently produce world-changing products while others struggle to innovate, this conversation offers one of the clearest frameworks you'll hear.

It's a masterclass on building not just products—but the systems that build products.

What happens when AI enters a world where being wrong isn’t an inconvenience — it’s a liability?

In this episode of Building One, Tomer Cohen sits down with Gabe Pereyra, co-founder of Harvey, to explore what it actually takes to build AI for one of the most complex and high-stakes industries in the world: legal.

Harvey works with leading law firms and enterprises to draft, analyze, and reason through complex legal work — contracts, filings, cases, and internal workflows where precision, accountability, and trust are non-negotiable.

On paper, legal is a perfect domain for AI.It’s language-heavy. Logic-heavy. High value.

In reality, it’s one of the hardest.

Every word matters.Every output has consequences.And “almost right” doesn’t count.

In this conversation, Tomer and Gabe discuss:

Why the hardest problem in AI today isn’t intelligence — it’s coordination

What makes vertical AI companies like Harvey durable as foundation models improve

Why legal systems must be auditable, permissioned, and accountable

How conflicts and data isolation create unique infrastructure challenges in legal AI

Why the future of work may look less like individuals using tools — and more like teams of humans and AI agents working together

And why the next bottleneck in AI may not be generation — but human review and trust

Gabe also shares his journey from aspiring professional soccer player to finance, AI research, and eventually co-founding Harvey — along with the contrarian thinking that led him to bet early on the future of AI.

This episode is about what it takes to move AI from impressive demos into real-world systems — where the stakes are high, trust is fragile, and the tolerance for error is near zero.

Every builder’s journey is non-linear.

What looks like a straight path in hindsight is often shaped by setbacks, pivots, and unexpected opportunities along the way.

In this episode of Building One, host Tomer Cohen sits down with Guy Raz — creator and host of How I Built This — to explore what he’s learned from interviewing hundreds of founders, and from building his own media company.

Before becoming one of the most recognizable voices in podcasting, Guy was a journalist and foreign correspondent — covering wars, filing stories on deadline, and learning how to operate with urgency and resourcefulness.

That experience shaped his approach to building: prioritize progress over perfection, and improve through repetition.

In this conversation, Tomer and Guy discuss:

The patterns Guy has seen after interviewing thousands of founders

Why relationships — not intelligence — are often the deciding factor in success

The craft behind great storytelling — and what it takes to earn a listener’s time

How a career setback became the turning point that led him into podcasting

Why the “safe” path can sometimes be the riskiest decision

Guy also shares how he thinks about value creation — and why the best builders aim to create an experience where the user feels they received more than they gave.

This episode is about craft, resilience, and the human side of building.

Because in the end, the goal isn’t just to build products.

It’s to build belief.

There’s a moment every builder remembers.

You type a few lines of code.The computer responds.And suddenly you realize: I can build things.

For decades, that moment was reserved for a small group of people who knew how to code. Turning an idea into working software required technical expertise, time, and often a full engineering team.

AI is changing that equation.

In this episode of Building One, host Tomer Cohen speaks with Anton Osika, co-founder and CEO of Lovable — a company building tools designed to dramatically reduce the friction between having an idea and turning it into working software.

Lovable allows people to describe what they want to build and generate functional applications far faster than before. But Anton’s ambition goes beyond helping developers move faster. His vision is to expand who gets to build in the first place.

From founders launching companies without technical co-founders to teams inside enterprises building their own internal tools, Anton believes AI is transforming software from a specialized craft into a much more accessible economic tool.

Tomer and Anton discuss:

Why the next wave of software creation is about enabling the 99% who don’t code

The philosophy behind Lovable — and why simplicity is often the hardest product decision to defend

The real tradeoffs behind AI-driven development, especially the gap between prototype and production

Why Anton believes the most underrated moat in AI is trust and brand love

How AI tools could unlock an entirely new generation of founders and builders

This conversation explores what happens when the barriers to building start to fall — and what it means for the future of entrepreneurship and product creation.

Because when building becomes easier, something bigger happens:

More builders.

What if the biggest problem in healthcare isn’t diagnosis — it’s capacity?

On this episode of Building One, Tomer Cohen sits down with Dr. Thomas Kelly, co-founder and CEO of Heidi, to unpack what it actually takes to build AI for one of the most complex, regulated, and human industries in the world.

Before starting Heidi, Tom was a vascular surgeon. He saw firsthand how some of the most highly trained people on the planet were spending their days on low-value administrative work.

Heidi began by listening to real patient visits and drafting clinical notes. Today, it’s expanding into the vast — and invisible — work around care: follow-ups, calls, scheduling, and coordination.

In this conversation, we explore:

What “doubling capacity” in healthcare really means

Why personalization must be nearly perfect — measured almost like a clinical SLA

What it takes to build AI that doctors actually trust

How GPT-4 didn’t kill Heidi’s moat — it forced a radical pivot

And how Heidi rewrote the healthcare go-to-market playbook by winning clinicians one by one

Everyone talks about AI’s potential.

This episode is about delivering it — in the real world, where trust is fragile, stakes are high, and a 5% edit can break the magic.