If the model is 1000x better than you need for your use case, then squeezing out 10x more quality doesn't matter!

2025-12-08 · Bits and Bobs 12/8/25

If the model is 1000x better than you need for your use case, then squeezing out 10x more quality doesn't matter!
- In that case, the model quality isn't the bottleneck.

More on this topic

From other episodes

The subscription I hold most dear currently is my Claude Max subscription.2026-03-17
We've left the Generative AI era and entered the Agentic AI era.2026-02-16
Some people think the models have plateaued in quality and some people think that's crazy because what we can achieve with them keeps growing without bound.2026-01-19
If the system assumes an LLM as the main loop then there's a floor of how cheap and performant it can be.2026-01-06
The quality of LLMs is model + harness.2025-11-24
Model quality no longer feels like the bottleneck with LLMs.2025-11-10
You want a system where the model quality is not the ceiling but the floor of possibility.2025-10-13
My friend Varun Godbole: The AI That Feels Good Wins.2025-09-15