Why almost no company should train its own foundation model — and what to spend that money on instead
Almost nobody should brew their own beer.
The big breweries are better, cheaper, and consistent. Same for foundation models. Buy the frontier. Build the edge. Your moat is your data, workflows, and judgment — not a behind-the-frontier model you spent $20M to ship.
Claude settings
shipped by Anthropic. Public docs cover 40. The configuration surface is the product.
The Sam Adams of foundation models is a phone call away. Use it.
Three pieces from the last 30 days that reinforce this analogy. Forward to the executive who's still calling it hype.
You don't need to brew it. You also don't get to skip learning how to pour. Configuration depth IS the product.
Read the original →Default thinking effort silently shifted high → medium. Two-line fix. "Claude got worse" is almost always a configuration problem, not a model problem.
Read the original →Pre-built, audited, on the frontier. Your moat is the workflow you wrap around them. Don't build the brewery.
Read the original →The catalog is free. The live session is where Don't Brew Your Own Beer gets applied to your specific stuck conversation — built for boards, ELTs, and founder-CEO 1:1s. Half-day, $7,500.