Introducing Whimsey Labs: Applying AI Forecasting to Building Businesses

Welcome to the Operator Economy

You might have noticed things look a little different around here. We’re rebranding from Pioneering Thoughts to Notes from Whimsey Labs. Same writing, same Substack, same you and me. The “Notes from” framing is deliberate — these are lab notes from the work of building AI companies. Experiments, observations, what’s working and what isn’t. The name change will make a lot more sense by the end of this post.

Last week, SaaS stocks dropped across the board. Not because these companies reported bad earnings. Not because of a macro shock. They dropped because investors are starting to price in something that founders have been whispering about for months: AI is going to eat the software industry from the inside out.

I watched those tickers fall and felt something I didn’t expect. Not dread. Recognition.

Because this is the shift I’ve been watching take shape. When software costs approach zero, the entire economic structure built around the old cost basis doesn’t survive. It just takes a while for everyone to notice. SaaS companies that raised mountains of capital, hired hundreds of people, and built monolithic platforms? That playbook is getting squeezed from both ends.

Two distinct types of companies are emerging on opposite ends of the spectrum. On one end: high-end AI-enabled services with human judgment at the core. Companies like Crosby — an AI-enabled law firm that charges a premium but delivers at scale, maintaining the human relationships and judgment that clients actually pay for. On the other end: ultra-lean teams, sometimes a single person, churning out AI-native tools that do 80% of what a legacy SaaS platform does at a fraction of the price.

And everything in the middle? Your Asana. Your Monday.com. Your LexisNexis. They’re in trouble.

But here’s the part I can’t stop thinking about: what replaces them isn’t just better software. It’s a fundamentally different kind of economy. I want to put a name on it: the operator economy. And I think it’s the most important shift since the assembly line.

The Assembly Line Moment

Before the assembly line, goods were made by skilled artisans. A cobbler made your shoes. A blacksmith forged your tools. Each item was handcrafted, expensive, and slow to produce. Then Henry Ford figured out how to break manufacturing into repeatable steps, and an entire economic model evaporated in a generation.

We’re at that inflection point again. Only this time, the thing being disrupted isn’t manufacturing. It’s knowledge work itself.

Lovable crossed $100M ARR in eight months with fewer than 20 people. That’s not an anomaly. That’s a signal. And when the cost of creation drops by 10x, the constraint stops being capital or engineering headcount. It becomes something else: judgment, taste, domain expertise, and the ability to orchestrate AI agents toward a specific outcome.

Those are human skills. And they’re the skills that will define the next generation of entrepreneurs.

We’re heading toward a world with massive infrastructure providers — your OpenAIs, your Anthropics, your cloud companies — on one end, and a vast ecosystem of nimble 1-to-5-person companies on the other, leveraging that infrastructure to operate at a scale previously reserved for organizations with hundreds of employees. The people who thrive won’t be the deepest specialists. They’ll be the ones who can manage across functions, direct AI tools effectively, and connect dots between domains.

That’s the operator economy.

What Tenki Taught Us (And Why We Evolved)

I need to tell this part honestly, because it’s core to why we’re doing what we’re doing.

I’ve built three AI companies across three very different industries: climate risk models at ZestyAI (now underwriting $4T+ in insured assets), legal research at Paxton AI ($28M Series A), and AI forecasting at Tenki. Each time, I learned something that only became clear in hindsight. At ZestyAI, I learned that AI can transform an entire industry’s infrastructure. At Paxton, I learned when to let go of something that’s working but no longer needs you. At Tenki, I learned something bigger — and it changed the entire direction.

We built a forecasting engine and pointed it at sports and prediction markets. The technology was solid. The research was real. We got paying customers. But here’s the uncomfortable truth we had to confront: sports betting markets are too efficient. Too many smart people, too much data, too many sophisticated models already competing for the same razor-thin edge. The margins were getting thinner every month, not wider.

That forced a question we should have asked earlier: where are the truly inefficient markets? Where is there real information asymmetry, real uncertainty, and real payoff for better decision-making?

The answer was staring us in the face. Running a business. Starting a company. Deciding which market to enter, which product to build, when to pivot, when to double down. These are all forecasting problems — just playing out in a wildly inefficient market where most decisions are made on gut feel, survivorship bias, and whatever framework the founder read about last week.

Think about what a good forecaster does: gather information, assess probabilities, size bets appropriately, track outcomes, update models. Now think about what a good venture builder does. It’s the same skill set. You’re making decisions under uncertainty, sizing investments, tracking what works, and feeding results back into your next decision.

Over the past year, Jay and I examined dozens of potential businesses. We developed a systematic approach to evaluating markets, sizing opportunities, and validating demand at speed. We got obsessive about separating signal from noise — identifying real pain points versus imagined ones, finding existing wallets versus trying to create entirely new budget lines. We learned how to validate fast. Throw together a landing page, point people to a form, and see if anyone cares. No six-month product development cycle. Ship something, iterate from there.

And we realized the skills we’d built were the product. Not the prediction market app.

We didn’t abandon the thesis. We sharpened it. The forecasting DNA, the probabilistic thinking, the obsession with sizing bets correctly — that’s still the foundation. We just stopped pointing it at sports markets and started pointing it at the question that matters more to us: which businesses should exist in the operator economy, and how do you build them intelligently?

Whimsey Labs: An AI Lab for Business Building

Today I’m announcing the next evolution. We’re now Whimsey Labs — an AI business lab that applies the same forecasting and probabilistic thinking we built at Tenki, but to a wildly different (and far less efficient) market: starting and operating businesses.

Here’s where we’re heading.

We’re building an AI-native business lab. Not a venture fund. Not an incubator in the traditional sense. Something closer to a factory that launches, operates, and scales small software businesses using AI agents, with a lean team of human operators running each one.

The model: identify acute pain points where current solutions are overpriced or underperforming. Build AI-native tools that solve them at a fraction of the cost. Operate each business with a small team augmented by AI agents that handle everything from content creation to customer research to operational workflows.

Every decision we make — which market to enter, which product to build, how to price it, when to double down or walk away — runs through the same probabilistic framework we developed at Tenki. We’re not guessing. We’re sizing bets.

And here’s what makes this compound: each venture we launch makes the next one faster and smarter. The agent infrastructure compounds. The operational playbooks improve. The data from running real businesses feeds back into better decisions about which businesses to run next. This is the flywheel — the same kind of compounding intelligence loop we originally set out to build at Tenki, just applied to a market where the inefficiencies are massive and the upside is real.

This publication — now Notes from Whimsey Labs — is the front door. Not content marketing dressed up as thought leadership. An honest, in-the-trenches account of what it actually looks like to build and operate businesses when AI changes the rules. Next week I’ll be publishing a deep future-state analysis called “When Software Costs Hit Zero,” mapping where the economy is headed across three time horizons. That’s the kind of thinking we’ll keep doing here. But we’ll also share the tactical stuff. What’s working. What isn’t. What we got wrong. The real numbers when we can share them.

Why LinkedIn Content Is Our First Bet

If you’re an AI business lab, you can build anything. So how do you choose where to start?

We looked at this the same way we’d size any bet. Where is the pain acute? Where are people already spending money on bad solutions? Where can we validate fastest?

The answer kept coming back: LinkedIn content for founders and professionals.

Here’s the insight. The founder is the most effective voice for an early-stage company. Everyone knows this. And it doesn’t scale. Until now. Founders are paying $1,000 to $5,000 per month to agencies to ghostwrite 3-5 LinkedIn posts per week. Blazel AI just announced a $7.3M seed going after exactly this problem, charging $3,000 per month. The pain is acute and universal — every founder we’ve talked to either has this problem or is paying someone to solve it badly.

Our first venture generates LinkedIn content that actually sounds like the person posting it. Not templates. Not generic thought leadership. Content that captures your voice, your audience, and your goals. Price point: $99/month.

I know. It doesn’t sound as sexy as “AI lab for the operator economy.” But that’s exactly why it’s the right first move. It has clear demand, an existing budget line we can capture, and it lets us build and test the entire lab infrastructure — the AI voice engine, the content pipeline, the operational playbook — on a real product with real customers. When Blazel is charging $3,000 a month and just raised $7.3M, and we can deliver comparable quality at $99 — that’s not a race to the bottom. That’s what happens when AI collapses the cost of production and you price accordingly.

We already have a waitlist with strong signal and beta customers lined up as design partners. People are marking that they’d be “very disappointed” if they couldn’t use this. More on our first venture in a separate post soon.

Why This Matters Beyond Us

I’m not the only one seeing the operator economy take shape. Dan Shipper at Every has reorganized his company around it — fewer than 20 people shipping multiple AI products and writing virtually zero code themselves. Andon Labs partnered with Anthropic on Project Vend, giving Claude an actual vending machine business to run. Feltsense just raised $5.1M to build fully autonomous “agentic founders.”

Multiple smart groups converging on variations of the same thesis makes me more confident, not less, that the shift is real.

But here’s what I keep coming back to. As AI displaces traditional knowledge work, there’s an enormous population of talented operators, marketers, analysts, and builders who need somewhere to go. The operator economy creates a path for them: small teams running real businesses, powered by AI infrastructure they couldn’t build on their own. Economic opportunity doesn’t have to shrink as AI grows. It can widen — if someone builds the system that makes it possible.

Imagine a world where starting a software business is as accessible as starting a blog. Where the barrier isn’t capital or credentials or connections — it’s just having a good idea and the willingness to run at it. Where AI agents handle 80% of the operational load, and human operators focus on the 20% that requires judgment, creativity, and relationships.

That’s what Whimsey Labs is building toward. Not one product. The system that launches many.

The biggest risk right now isn’t building something that fails. It’s clinging to a model that already has.

Follow along here and at whimsey.ai. We’re building in public, and we’ll share what’s working and what isn’t as we go.