Intelligence Is Free. Good Luck.

gist

Evis Drenova argues that cheap intelligence will push startups away from thin software layers and toward research-driven products. She sees today's AI investment as expensive but unlike the dot-com bubble because established companies have cash flow and customers, while demand remains compute-constrained. As models become widely available, proprietary data, custom models, harnesses, and tools become differentiators. Developers who learn to fine-tune, evaluate, and serve open models can redirect time saved from coding into deeper technical work.

ideas

• AI investment differs from the dot-com bubble. Drenova points to established companies with proven demand, cash flow, and low leverage, while noting that valuations are high and compute demand still exceeds supply.
• Cheap intelligence makes the app layer harder to defend. Cloud compute enabled SaaS, while broadly available intelligence makes proprietary data, custom models, harnesses, and tools more important differentiators.
• Research can be both product strategy and recruiting strategy. Drenova points to Ramp, Cursor, Stripe, and Thomson Reuters as companies using research to build capabilities beyond a generic model interface.
• Developer work is moving toward model operations. Fine-tuning open-source models, evaluating them against existing systems, and serving them cheaply become skills alongside shipping applications.

quotes