What I learned this Week: October 12, 2025
Capitalism’s R&D Cycle
Lessons of the Week
Bubbles are a feature, not a bug. Schumpeter told us so.
Every boom is a form of economic discovery. Planned economies try to allocate capital rationally, but knowledge, as Hayek argued, is local, tacit, and constantly shifting. In free markets, we don’t plan the future; we overfund it, then let failure compress the noise. The bubble is how our economy learns. It amplifies the weak signals of innovation until the real signal breaks through.
In the late 1990s, capital flooded into fiber optics, routers, and dot-coms. Most went to zero, but the infrastructure they left behind carried the bandwidth that powered Web2 and cloud computing. Today’s AI and clean-tech bubbles rhyme with that history: the same exuberance, but now building hard assets and profitable layers of compute. It
The pattern is consistent. First come dreamers and excess capital. Then collapse, consolidation, and durable monopolies. The destruction is necessary; it converts Shannon noise into Schumpeterian order. Amazon was nearly lost in the last washout, down to single digits before compounding for two decades.
So don’t fear the froth. Bubbles are capitalism’s research and development cycle at planetary scale. The only real question: can you hear the signal before the noise clears?
What I (re)learned this week….
Gen Z social usage peaked in 2022. 57% set limits on social media, homeschooling doubled, 66% buy sustainable food. “Detox” isn’t just apps: bad schools, junk diets, toxic brands, all getting cut. Ignore Gen Z and you miss the future they’re already building.
The business model that integrates food, health, pharma, and wellness seamlessly to optimize human outcomes will trigger a wave of Creative Destruction. See Aggregation Theory
The breakup of Corteva and the FBN divestiture of chemicals may signal that companies want to firewall themselves from glyphosate and PFAS lawsuits. Pure play biologic distribution will likely benefit.
As food companies move to fiber for GLP/PYY like attributes, they are unlikely to recognize that the forms of fiber and how they interact in humans are as complex as titrating the right medications for a patient.
Food formulation may start looking more like drug trials.
With AI, change can come from anyone in the value chain. A farmer, a co-op manager, a product developer, a nurse. They can draft their vision in ChatGPT, assemble a development team with agents, trigger Perplexity Comet to identify key partners on LinkedIn, Vibe-engineer the supply chain, reengineer a food product for nutrition, model the human metabolic impact, and publish the marketing strategy all within the same morning.
The process of Creative Destruction is the essential fact about capitalism.
— Joseph Schumpeter, Capitalism, Socialism, and Democracy (1942)
Food is Health
This is what the inside of a B737 wheel well looks like. In aviation maintenance, it’s called the “thinking bay.”
Since its first flight in 1967, the B737 has flown more than 6.5 billion passengers, on more than 650 million flights, with fewer than 6,000 deaths. And it is doing everything it can every day to improve safety.
Meanwhile, 261 people die in the US every day from Type 2 Diabetes, the pandemic that drives chronic disease. It is completely preventable.
Healthcare is much more confusing than the wheel well of a B737. Boeing will clean sheet this design and introduce a new single-aisle aircraft, further improving safety. Will healthcare?
AI Applications
I’m a heavy AI user. It’s supercharged my productivity and expanded my reach into distant corners of our network. For over four years, we’ve woven AI into our venture operations, from early OpenAI models to today’s pro/ultra usage of ChatGPT, Claude, Perplexity, Gemini, etc. As our data capture improves and context windows grow, every layer of our stack compounds.
Some of the more powerful use cases we lean on:
Conferences: stream sessions, get transcripts in real time, then feed them to ChatGPT with your own questions or annotations. At day’s end: ask for a distilled summary and a target list of people to meet (while the conference is still live).
Portfolio Agents: give each company its own agent, hooked to its diligence files, web presence, email threads, and meeting logs. Then schedule weekly scans that surface intros, insights, or red flags for the CEO. It’s like letting every CEO “sit in” on everything without being there. (You could replicate that across functions, marketing, finance, strategy, via agents tailored to each domain.)
Strategic Integration: after calls or reviews, dump the transcript and meeting notes into our “Food Is Health” knowledge base + ChatGPT. Ask it to restate the strategy, suggest pivots, and map cross-company synergies. Then send a polished version to the CEO, ensure alignment, and show what connections we’re tracking across the portfolio. We now link into HubSpot so it automatically gathers the emails, mobile, etc for contacts.
Voice Mode: I talk to ChatGPT in cars, airports, and hotel rooms. With earbuds, it passes for a phone call. I speak through half-formed ideas, revisit them over days, and by the end get a composed, workable output.
Some Data and Links
→ Less Climate Change? Interest in the climate cause is fading. Polling shows that fewer Americans now insist the climate is human-driven (down to ~51 %). Energy numbers anchor the realism: fossil fuels still supply 82 % of global primary energy. Renewables and low-carbon sources now contribute only a sliver of total consumption, despite accelerating deployment. Climate proposals that clash with incentives, job markets, trade balances, and household budgets are losing momentum. Voters see that mismatch. For the innovator, the right way forward is to make cheap, plentiful, secure, and clean energy the default choice, not a moral demand or subsidy, but an economic edge. The final answer will be a mix of nuclear, solar, batteries, and DC power.
→ AI Usage: A majority of Americans now say they interact with AI at least several times per week. 31 % claim “almost constantly/several times a day”. Those with more education use AI more. Perhaps because they are already knowledge workers. But the fact that AI has already seeped into everyday tasks means its leverage lies beyond novelty. What will it look like in a year? Usage is pretty even across demographics, suggesting there is little adoption friction in the system.
This Week’s Charts and Social
Design for Manufacturing: SpaceX Raptor engine evolution
Occam’s razor: Late-night hosts are getting fired because of fascism, or maybe people are not watching late-night, and fascism is a good thing to blame when canceling a contract.
Easier said than Done: There are many reasons not to circumcise baby boys. Some reasons are a bit weaker than others.
