Anthropic used its own model to analyze past sessions
the more expertise the user has, the higher the chances of success with claude code
Our latest economic research introduces a framework for tracking Claude Code as it scales.
Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise shape whether a session succeeds? https://www.anthropic.com/research/claude-code-expertise
Many users praised Anthropic's Claude Code research for showing that domain expertise multiplies output and supercharges skilled users instead of replacing them.
No Digg Deeper questions have been answered for this story yet.
Claude amplifies your expertise.
Our latest economic research introduces a framework for tracking Claude Code as it scales.
Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise shape whether a session succeeds? https://www.anthropic.com/research/claude-code-expertise
We let Claude analyze Claude sessions and then asked Claude how much value Claude brought. Here are Claudes findings:
Our latest economic research introduces a framework for tracking Claude Code as it scales.
Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise shape whether a session succeeds? https://www.anthropic.com/research/claude-code-expertise
@AndrewCurran_ Ha, I’ve posted a few times to this exact point over the last few weeks, just not as politely as they put it here
Nice to have actual data on this though. I wonder if it’s intrinsic to LLM behavior
the more expertise the user has, the higher the chances of success with claude code
Cool to see that the Claude Code sessions we collected and released are now helping Anthropic study Claude Code 😎 Check out SWE-chat if you haven't! (more data coming soon...)
Our latest economic research introduces a framework for tracking Claude Code as it scales.
Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise shape whether a session succeeds? https://www.anthropic.com/research/claude-code-expertise
@scaling01 they model the user and respond accordingly.
i wonder a lot about terry tao's chat, that thing must be like 'yo! what?? bet.'
@xeophon Let me guess....
@xeophon
@AndrewCurran_ Whenever I hear about low productivity gains from AI I think: 1. Three employees are tech illiterate 2. Six employees use it as a chatbot 3. One employee has 4 terminals open and finishes all his work in 11 minutes while he goes on draftkings
For a net productivity decrease.
@AndrewCurran_ yeah claude doesnt replace expertise it multiplies it
@cbonnett 10/10, no notes
@AndrewCurran_ an my anxiety as well if they don't bring Fable back.
@TheMindScourge I think it is, if you are a well read generalist it supercharges you, because you only need to remember the edge of something relevant to the discussion and the model will instantly know exactly what you mean and bring the entire thing into the context window.
@scaling01 this is the vim problem — the people who need it most get the least out of it. expert users already know what good code looks like, so they steer claude effectively.
@scaling01 curious if expertise here means domain skill or just knowing when claude is wrong. missed a bad rls pattern for 2 days and i'd call myself decent at flutter
@AndrewCurran_ AI: Amplified Intelligence.
@AndrewCurran_ wasn’t there some article showing that professors used claude way less effectively than their grad students?
@joabaum's SWE-chat is the best dataset to date for understanding what coding agents in the wild look like, lots to uncover here as shown by the Anthropic report.
Cool to see that the Claude Code sessions we collected and released are now helping Anthropic study Claude Code 😎 Check out SWE-chat if you haven't! (more data coming soon...)
@AndrewCurran_ Strength leads to strength
@AndrewCurran_ @TheMindScourge This exactly
