Understanding the technology, business, policies, and people of Apis.
Nobody wants to talk about APIs at a dinner party, and they never will. APIs are plumbing, and plumbing is invisible right up until the moment it fails. I have made peace with this after a long time doing this work–the API will never be the cool thing in the room. But I want to say something that sounds like a contradiction and is not: APIs are roughly a hundred times more important today than…
Nathan Booker asked me a good question on LinkedIn a couple of weeks back, and it has been sitting in my notebook since. He wondered whether changes to MCP APIs are really breaking changes at all — that when your API is agent-facing you can evolve much more quickly, because there is reasoning on the other end that can deal with the change, the way a person deals with a redesigned interface.…
The oldest fight in the enterprise is the one between business and IT, and it never really got settled–it just changed costumes. It was business versus IT, then it was product versus engineering, then it was the design team versus the platform team, and every few years someone sells us a new ceremony to paper over the gap. Requirements documents. Two-week sprints. A backlog. A Jira workflow with…
I have spent this whole series banging my head against the same wall. Company after company tells me they are all in on AI, all in on agents, all in on automation, and then they hand me an onboarding flow that assumes a human with a mouse, a marketing email I have to confirm, and a dashboard I have to click through to get a single credential. The contradiction never gets old to me. You cannot…
Deprecation is where most API programs quietly fall apart. Not because teams don’t know how to sunset an operation, but because they do it by hand, one operation at a time, across a dozen specs, on whatever day someone finally gets to it. Somebody sets deprecated: true in one place, forgets the Sunset header in another, writes “please migrate soon” in a description here and “this endpoint is going…
I have spent a good chunk of this year authoring Arazzo across my network — more than 5,000 workflow files spanning 543 API providers, plus another 255 that deliberately cross provider boundaries, because the interesting workflows almost never live inside one company’s API. Every one of them validates. Every step points at a real operationId in a real OpenAPI document. And somewhere in the middle…
As I continue to profile the portfolio companies of venture capital firms, as well as the services and tooling those companies use, I am left thinking about the reasons why more of these companies aren’t programmable — not just for their core offerings, but for their operational surface area. I know if I was a VC who was all in on the agentic hustle, I’d be like, “MAKE IT ALL PROGRAMMABLE!”
This is the ninth and final post in my series on Germany’s federal API authorization blueprint. I have spent eight posts pulling the architecture apart layer by layer and, last time, mapping it onto the US. I want to close where the work has the most immediate leverage of all: the rest of Europe. Because for a European member state, this is not a distant example to admire — it is a neighbor’s…
Editorial note: this is how the API Evangelist papers and reports are produced. I have published it as a permanent page at About API Evangelist Papers, alongside About API Evangelist, About Kin Lane, and my stance on how I use AI. It is linked from the navigation across the network, and from the top of every paper on the storefront. When the process changes, I will change the page and say so.
I keep coming back to the same wall. We are all supposedly “all in on AI,” wiring up agents that are meant to go off and do real work on our behalf, and yet the very first step of almost every integration still requires a human to log into a dashboard, squint at a settings page, and click a button. I have been banging my head against this for long enough that I now treat it as a test. When a…
Ten days ago I spent an afternoon with curl figuring out how Time serves ads to AI agents, because the trade press covered the business arrangement and openly said they could not tell you how it worked. I came out of that with a mechanism, a set of measurements, and a strong opinion about the parts of it nobody should copy. Then I went and looked at my own properties, and found something I should…
I have spent enough years watching people love GraphQL for the wrong reasons and hate it for the wrong reasons that I want to step around both of those piles entirely and talk about the one place where I think it is genuinely, quietly excellent. Not GraphQL as your public edge. Not GraphQL as a REST replacement. Not GraphQL because a frontend team wanted to stop waiting on backend tickets. I mean…