The Frontier

VR investment's primary payoff was in developing foundational technologies for robotics, like SLAM positioning and depth sensing, rather than mass consumer adoption.

AR glasses using waveguides and micro LEDs are ahead of their time because manufacturing yields and costs are not yet ready for mass production.

Hardware's development cycle involves only 4 or 5 total 'compiles' - full design and manufacturing builds - with no ability to ship over-the-air updates after final production.

A key reason for the current AI hardware boom is the anticipation that digital AI capabilities will saturate, making the physical world - robotics, manufacturing, and sensing - the next frontier.

Current humanoid robots are advanced prototypes, not yet ready for scale. Safety concerns require designs with lighter, softer actuators to reduce impact energy.

Scalable robotics depends on a complex, globally outsourced supply chain. A safe supply chain requires independence in foundational layers like raw magnets, actuator production, and batteries.

Kalinowski argues for significant U.S. reindustrialization to ensure military safety, advocating for greater investment in drones over traditional assets like aircraft carriers.

Apple's hardware excellence stems from a first-principles approach where every design decision, even internal ones, is considered to force clarity on core outcomes.

Key hardware development principles: define immutable KPIs like cost or weight early, design the hardest parts first, iterate most on components users touch most, and execute known tasks immediately.

Memory prices are a major supply chain risk for consumer hardware and robotics, driven by AI demand. Kalinowski advises companies to pre-buy memory to ride out price spikes.

AI is beginning to assist in hardware engineering via high-level planning and tools like Excel, but true transformation requires models that understand physical properties like friction and contact.

Humanoid robots face a hype cycle; dedicated, task-specific robots are more suited for scalable manufacturing and logistics than a single generalist humanoid shape.

Creating robots that feel human and connected requires non-threatening, soft designs and clear intent signaling, drawing from research in nonverbal cues and animation principles.

Sam Altman's leadership lesson was pushing for 100x or 10,000x scale ambition, while Steve Jobs set an unwavering bar for technical excellence that motivated teams.

Building a zero-to-one hardware team requires a mix of strong generalists, scaling experts, and AI-native young engineers who approach problem-solving fundamentally differently.

The Lean Startup's 'minimum viable product' and hypothesis-driven development principles are now the standard operating model for leading AI companies like Anthropic, which launch research previews to gauge interest before full commitment.

A pervasive 'force' drags successful organizations toward mediocrity, where their very success becomes a liability as the temptation to extract value grows with their financial worth. Eric Ries calls this 'financial gravity'.

Ries argues standard corporate governance is fundamentally flawed. Harvard Law School data shows only 20% of venture-backed founders remain CEO three years after a company's IPO, making founder ousting a statistical likelihood, not an anomaly.

Ries cites Philip Morris's 2021 acquisition of the inhaler company Vectura as a real-world case of 'fiduciary duty' forcing a board to accept a value-destroying bid that outraged the public and medical community.

The principle 'harder is easier' states that upfront commitment to principles like quality or safety builds trust, which lowers long-term costs like customer acquisition and increases organizational velocity, despite creating short-term difficulty.

Cloudflare embodied 'harder is easier' when it made SSL encryption free for all users, sacrificing short-term revenue to fulfill its 'make a better internet' mission. This built massive trust and contributed to its $70 billion market cap.

In contrast, Groupon eroded trust by incrementally increasing email frequency from one to eight per day based on ROI data, which destroyed its core value proposition and contributed to its decline.

Ries claims traditional 'shareholder primacy' governance is a recent, 40-year-old dogma. For most of corporate history, companies were required to have a specific beneficial public purpose, and deviating from it could result in a corporate death penalty.

The simplest structural defense is filing as a Public Benefit Corporation (PBC), which legally codifies a specific company purpose beyond shareholder returns. All major AI labs, including Anthropic, are PBCs.

For durable protection, companies need a 'mission guardian' like a Perpetual Purpose Trust or nonprofit foundation. Novo Nordisk, created in 1923, is governed by a nonprofit foundation that has protected its scientific integrity for over 100 years.

Academic research shows companies with fortified governance structures like Novo Nordisk's are six times more likely to survive to their 50th anniversary and deliver superior returns on invested capital compared to conventionally governed peers.

Anthropic's Long-Term Benefit Trust appoints AI safety experts without equity in the company to its for-profit board. This structure allows it to refuse releasing models or lucrative contracts, like a $200M Pentagon deal, on safety grounds.

Eric Ries was consulted by Anthropic's founders early on, advising on governance structures to protect its AI safety mission before the generative AI boom and ChatGPT's release.

The 'Todd Park rule' for culture is to only make deposits (sacrifices for values) and never intentionally make withdrawals (acting against values for short-term gain), as withdrawals happen accidentally enough.

Max Schoening says AI makes the first 10% of any project free, drastically lowering the effort to build a startup's first version.