Back to Building what's next (Interview).
Why Do Agents Auto-retry Standard Rate Limits?
Zeno Rocha says agents know how to auto-retry when a service follows the rate-limiting RFC. A proprietary header can report that a caller has used six of 10 requests. Many good services might use a language they invented for themselves. The RFC defines three headers: one for the count, one for the unit of time, and one for when to try again. Follow the RFC and agents know how to auto-retry.
The basics he went back to are the command line, the OAuth layer, the right headers, and the right verbs, and they matter more than ever. When someone builds a proprietary thing because they believe they have a better opinion than others, the agent will know what to do.
A YAML OpenAPI spec was already available. Two of the agent experience core services he checks still said no OpenAPI spec was there, because they search for the .json version, not the .yaml version. He has to build a converter from YAML to JSON and deploy it in the right place so the spec is discoverable. The primitives are the only thing that matters.
Building what's next (Interview) ·
From the transcript
So true. And, man, we've learned that lesson the hard way in the sense that the primitives are the most important thing that matters right now, like getting the primitives right. So what does that mean to you? Write that down. For example, the rate limits, you can just have a proprietary header to just like respond with the– like where you are in terms of your rate limit.
Oh, you've used six of your 10 requests so far. Great. So this is one thing. Like you go to many of these really like good services. They might use like a language that they have for themselves that they invented, right? Well, there's actually an RFC for rate limiting with three different headers, one for the count, one for the unit of time, and then one for when you should try again.