Recent and related:<p><i>Claude 2</i> - <a href="https://news.ycombinator.com/item?id=36680755">https://news.ycombinator.com/item?id=36680755</a> - July 2023 (255 comments)<p><i>Model card and evaluations for Claude models [pdf]</i> - <a href="https://news.ycombinator.com/item?id=36681982">https://news.ycombinator.com/item?id=36681982</a> - July 2023 (25 comments)
I've been on the API waitlist for months. I'd like to integrate Claude in my open source AI coding assistant tool.<p>I've had numerous users request this over the past couple of months:<p><a href="https://github.com/paul-gauthier/aider/issues/7">https://github.com/paul-gauthier/aider/issues/7</a><p>Feels like it would be unwise to build atop something unofficial like this?
Note that Claude 2 scores 71.2% zero-shot on the python coding benchmark HumanEval, which is better than GPT-4, which scores 67.0%. Is there already real-world experience with its programming performance?