Build Collective

Bet on fundamentally different deep tech early, not on hot markets. Build Collective backs deep technology that unseats incumbents by being fundamentally different, not just better on features. Fadell's rule is that by the time something is hot, it is too late to get in.

Examples include Simbe Robotics (retail inventory robots, around 7 to 8 years in), Recursion for drug design (10 years) and early bets on Groq and Cerebras. Fadell argues software-only companies are being questioned, while companies combining atoms and software have staying power. "When it's already hot, it's already too late to be in it."

Lenny's Podcast▶ 1:20:50

The fund helps founders reach a workable product by version one or two. Build Collective has held over 200 companies at times and helps with product, operations, financing and communications. Many of its bets, such as AI plus robotics, took years of work on product-market fit, marketing and version 3.0 before taking off.

Lenny's Podcast▶ 1:20:50

Start from a long-standing pain that new technology can finally solve. Heating and cooling was about 50% of the energy bill, yet people hated thermostat interfaces and programmable ones were rarely used. Nest used AI to learn schedules, so a $249 thermostat could save $800 to $1,200 a year.

Fadell says old pains persist because products evolved but never revolutionized, and people stop noticing a pain when a product relieves a bigger one. AI was a founding piece of Nest in 2010, but in 2011 it was called a learning thermostat because "AI thermostat" would have scared people.

Lenny's Podcast

Launch when the enabling technologies ripen together. The iPod arrived with portable mass storage, MP3s, lithium polymer cells and low-power ARM processors. The iPhone arrived with multi-touch, fast enough processors, Wi-Fi everywhere, digital cameras and YouTube.

Lenny's Podcast

Reinvent the whole system around the product, not just the device. Nest sold through installers instead of Best Buy and cost five to six times more than typical thermostats. The iPod came with iTunes and the music store; the iPhone came with the App Store.

Lenny's Podcast

A first version has little data, so a small team's opinion must decide. In the iPhone keyboard debate, only 1% or 2% of phone users had a BlackBerry, so the question was what the other 98% needed. Steve Jobs chose the virtual keyboard and told dissenters to leave the project.

Months of timed typing tests measured speed, errors and ease of correction, with scores climbing a little each round. Fadell's verdict: "Were we as good as a hardware keyboard? No. But were we good enough? Yes."

Lenny's Podcast

Back the gut with prototypes and experts, not consultant studies. Fadell says decisions made only from data produce a product that isn't different, and leaders who commission studies are covering themselves. A great product leader says "I might be wrong," then corrects later and takes the heat.

He favors a "benevolent dictatorship" that explains what it decided and why, warning that repeated kind lies build dysfunctional systems.

Lenny's Podcast

Micromanage only the few decisions that matter. Fadell limits close control to customer-facing details, manufacturing or cost, long-term vision, crises and system-level changes where many layers move at once.

Early in his career he thought everything mattered, and it drove everyone nuts.

Lenny's Podcast

No product works the first time; plan for three generations. The first two iPods sold only to Mac loyalists, under 1% of the market, and took off in the third generation with Windows support and the iTunes store.

Fadell's sequence: make the product, fix it with customer feedback, then fix the business and its margins.

Lenny's Podcast

Write the press release first to force three or four key features. Customers only see a product through its marketing, so Fadell treats storytelling as part of the product. Steve Jobs honed the iPhone story daily, telling it to friends until it was perfect.

The same iPod marketing failed in Europe until it was rewritten for slower adopters.

Lenny's Podcast

Prototype with AI, then have humans architect the product. Fadell compares fast AI-built software to fast fashion and says AI-generated code creates brittle products. He recommends prototyping with AI, then architecting it and working on narrow sub-parts.

He cites engineers calling Claude's leaked main loop brittle and unreadable, a claim made without evidence shown. Lenny Rachitsky: "Because it's so easy to build, the things that stand out are the things that are really well thought through."

Lenny's Podcast