Oops!

What if Every Fruit were Red?

In any technology monoculture, we lose more than just color.

When a single optimization criterion shapes all output, the system converges on sameness. That sameness is a debt, a fragility disguised as short-term efficiency. This edition examines software monoculture and what happens when systems are built around a narrow set of capabilities.

Breeding Faster Horses Does Not Produce Automobiles

Scaling LLMs does not make them AGI.

Claims like scaling language models could lead to AGI are as unfounded as breeding horses could lead to motorvehicles. Scaling expands the space language models draw inference from, but there is no evidece to claim its underlying capacity to infer has improved. This edition examines the assumption that larger AI models, would become a Swiss-knife for all reasoning & inference tasks.

If We Built Houses like Software, We'd Bury the Earth in Waste

If what we build today can't support what comes next, no value accumulates.

Imagine if houses could be built as cheap as software. We might build around whatever people want today, demolish what no longer sells, rebuild for every new requirement, and give little thought to what could support what comes next. The result wouldn't be abundance but wasteful accumulation of structures with little worth building upon. This edition examines what happens when we optimize for products and immediate requirements instead of durable foundations and reusable capabilities.

If You Write for the Market, Who is Writing?

Thinking backwards from what can be sold is not how enduring value is created.

Imagine a painter who starts at “What painting has the highest conversion rate?”, or an architect who starts with ‘What kind of building is selling best this year?’.” The market is a useful test for value. It is a poor substitute for having something valuable to create or express. This edition examines how the popular template of building products optimizes the wrong creative outcomes.