saved
Greg Isenberg: 21 observations from 48 hours in SF
Hraness wrote this summary from a saved copy of the source. Quotations are taken word for word from the source.
gist
After 48 hours with San Francisco AI founders, investors, and executives, Greg Isenberg lists 21 observations about where startup value is moving. He sees harnesses and agent building blocks rising, software becoming easy to clone, data and memory becoming the moats, and distribution getting harder to win.
ideas
- Ask whether a new model helps you. A dinner question about whether the next model makes a company better or worse silenced half the table.
- The computer is being reinvented. He frames models as CPUs, harnesses as operating systems, and skills and plugins as apps; good vertical harnesses treat the model as swappable.
- Cloning shifts value to data. Agents can rebuild software quickly, so years of customer data and workflows may be what legacy SaaS is really worth.
- Memory and deployment are the next fights. Founders told him the model that remembers most wins, and that the hardest AI problems are now deployment, not research.
- New markets and old problems. He expects agent insurance, post-breach security, and data cleanup work, while distribution feels as hard and expensive as the dotcom era.
quotes
“If a new model drops next month, does your company get better or worse?”
“The model is the CPU, the harness is the operating system, and skills and plugins are the apps.”
“Some SaaS companies are secretly data companies.”
“the hardest problems in AI are deployment problems now, not research problems.”