jcarterwil

In 1920, Alfred Sloan invented the modern corporation to solve an information problem. Companies were too big and too complex for any one person to coordinate. His divisional management model routed knowledge from the edges to the center and back again. It worked for a century. Enterprise software is the digital version of Sloan's org chart. And the trillion-dollar selloff in software stocks last week is the market recognizing that this architecture has an expiration date. AI doesn't replace bad software with slightly less bad software. It dissolves the coordination layer that made the software monopoly possible. When an AI agent can bridge the gap between a customer signal and an engineering response directly, you don't need five layers of management or a monolithic ERP system in between. Jensen Huang says replacing software with AI is like reinventing the screwdriver. He's wrong. Nobody is reinventing the screwdriver. They're eliminating the guy who stands between you and the toolbox and charges rent. New post on what this means for GDP growth, whether copycat economies can close the gap, and the three things to watch as the Sloan model unwinds.

We've wasted 30 years arguing who pays for healthcare: Dems want universal coverage, GOP wants markets.

Both miss the point. The "healthcare" we're buying is junk. It's designed to manage chronic diseases we create with bad food. ACA subsidies? Just funding the epidemic.

Healthcare administrators grew 3,200% from 1975-2010. Physicians grew 150%. This isn't a funding problem. It's a signal problem. When bureaucrats allocate capital, we get more bureaucrats. When consumers allocate capital, we get solutions to their actual problems. Today's piece makes the case for expanded HSAs and consumer-directed spending as the only viable path to ending chronic disease. The argument draws on Hayek's distributed knowledge problem: no central planner can aggregate the preferences and needs of millions of individuals fast enough to match market discovery. Healthcare destroyed the price mechanism. We need to rebuild it. 100M+ pre-chronic Americans represent $50-60T in future healthcare spending. The current system ignores them until they're sick. Give them capital for prevention and watch how fast markets reorganize.

Consumer apps are giving shoppers instant nutrition scores and ingredient transparency. Yuka reports that 94% of users put back red‑rated items; Intermarché reformulated 900 products and removed 142 additives to improve scores.

Upstream, companies like Edacious are measuring nutrient density and making the data usable across the supply chain. The hypothesis: verified nutrition unlocks demand and sends a reformulation signal that rewards value‑added agriculture.

Questions for HN: What data standards, APIs, or open ontologies should this stack build on? Where do these tools break (gaming, model drift, labeling‑law conflicts)? What would you want retailers, payers, or EHRs to expose so that nutrition data can flow?

Disclosure: we’re investors in Edacious.

Most attempts to improve processed food stop at swapping ingredients: less sugar, fewer calories, lower salt. The evidence suggests that doesn’t solve the real problem.

A more useful way to approach reformulation is to focus on the metabolic systems most affected by ultra-processed foods. Think about protecting the liver by capping added sugar and fructose and reducing seed oils, feeding the gut by adding real fiber and removing emulsifiers that disrupt the microbiome, and supporting the brain by rebalancing fats and including essential nutrients. This way of thinking, sometimes called the metabolic matrix, has already been applied at scale in the Middle East, where a major food company reformulated hundreds of products.

The other missing piece is verification. Labels and nutrient panels don’t tell you whether a food reduces metabolic stress. Continuous glucose monitors do. A simple A/B test comparing the old and reformulated product shows immediately whether you flattened the glucose spike. That gives you real outcome data within days rather than waiting for epidemiology years later.

This frames reformulation as an engineering problem: change the inputs, define the constraints, then measure the outputs with real human data. It shifts the conversation from “does the label look better” to “does this food reduce glycemic pressure in practice.”

I’m curious whether others have tried using CGMs or other biometrics in product design. Could this become a standard tool for validating food and nutrition claims?

We recently published a FutureCast piece imagining Heartland Mart, 2036 — a discount retailer that evolves into the frontline of healthcare.

The scenario is fiction, but it’s based on real trends:

Continuous glucose monitoring & nutrient density scoring becoming cheap

Rural retail chains reaching more households than clinics

Chronic disease costs ($1.9T annually in the U.S.) creating an incentive for food/health convergence

Data + supply chain innovations aligning farmers, retailers, and insurers around prevention

In the story, Heartland Mart starts as a dollar store, then layers food scoring, metabolic coaching, and insurance alignment. Over a decade, this collapses healthcare costs by nearly half.

245 million acres of US public land could sequester more carbon than all of California's emissions while producing the world's highest quality beef. The tech exists. The economics work. #RegenerativeAg #systemsthinking