Our great sponsors
-
WorkOS
The modern identity platform for B2B SaaS. The APIs are flexible and easy-to-use, supporting authentication, user identity, and complex enterprise features like SSO and SCIM provisioning.
In my opinion, an "AI software engineer" is the wrong target for the current generation of models, though it's obviously good for generating buzz.
I'm working on an open source, OpenAI-based tool that uses agents to build complex software (https://github.com/plandex-ai/plandex), and I have found that the results are generally much better when you target 80-90% of a task and then finish it up rather than burning lots of time and tokens on trying to get the LLM to do the entire thing.
Results are also better, in my experience, when the developer remains in the driver's seat--frequently stopping, revising prompts and context, and then retrying rather than expecting to kick back and let the model do everything.