Don't build your organization around a model provider
Your agents' identities, knowledge, and working relationships should outlast your choice of models.
Artificial Smarts
A weekly letter about how to make artificial intelligence work for us. Each one is an experiment, an explanation, or an argument about AI and the organizations we build around it. I distinguish what I have observed from what I expect, and return to the things I get wrong.
Three themes
Agents in practice
Context, memory, communication, and what it takes to complete real work.
Organizations and knowledge
Delegation, review, shared knowledge, and who retains the authority to decide.
AI and society
Work, skills, institutions, and who can access, control, and benefit from knowledge.
Essays
Spend your compute on correctness
How I think about allocating agent effort to checking the work we produce.
Pair your coding agent with a reviewer
A working arrangement for two agents, and why giving them different responsibilities matters.
BeadHub: coordination for AI programming teams
An earlier experiment in coordinating AI programming teams, since superseded by aweb.
With the rise of AI-assisted programming we'll want more programmers, not fewer
An argument about productivity, demand, and the continuing need for human judgment.
Data cleaning and anonymizing with GPT-3.5
An early experiment in using language models to clean and anonymize data.