Anthropic: What Remains for Humans in a Company of Agents? Research Insights Made Simple #40 (#DevEx)
Tomorrow at 10 in the morning, live, I will discuss “How employment is changing at Anthropic”, which I wrote about this morning. In effect, this continues the conversation we began when discussing the “Google Developer Productivity for Humans” series, where we asked how to tell whether growing activity helps people produce useful results.
First, we will examine the article itself: — what the authors learned about hiring, productivity, workflows, and changing roles; — how the research works: internal data, interviews, and comparisons of employees before and after adopting agents; — what the charts of agent hours, code volume, and quality show, and which conclusions are premature; — why organizing agents’ work creates more work, and where human attention goes.
Then I will return to the questions in my margin notes on the article, available in the PDF attached to this morning’s post: — How can we organize the working loop without checking every agent action manually? — Where will future independent engineers gain experience if simple tasks are already automated? — Under what conditions does automating one’s own work benefit an employee: what do they receive in return, and is there another role for them?
This studies one fast-growing company that develops AI itself. Its data is interesting in the details; we should be cautious about extending it to the entire labor market. Bring questions and examples from your teams. I am particularly interested in the work that appeared after introducing agents, even though it was absent from the original promise to “save time.”
Earlier conversations:
- Developer Productivity for Humans — episode #30
- 3 AImigo: “An AI-native Company — What Does That Mean in Practice?”
- 3 AImigo: “A Junior Without Simple Tasks — Entering IT in the AI Era”
#AI #Research #Agents #Engineering #Management