Starting a new specification is exciting. You have a vision. You flesh out the vision. You distill it down into some documentation, schema, blog post, and a social media post. You don’t have the baggage of a specification that has been around for a decade or more. Things are easy. But, does anyone care? It is an interesting place to be and something I find it exciting, but I have been here numerous times before and I’ve learned to be skeptical of when I get excited about something new—-knowing there is always more going on behind why I am getting excited.
The early days of Swagger were exciting. Even the early days of OpenAPI were hopeful. Early days of AsyncAPI were exciting. I can feel the same excitement behind the waves of new specifications emerging to met this AI moment. I caught hints of this stirring inside of me with Agent Skills and JSON Structure—probably for different reasons than many AI champions, engineers, and data people. But, I get why people get excited and feel the energy, and I work hard not to diminish that with the realities of managing a specification over time, let alone the demands put upon you when your specification is successful. I truly want to celebrate people’s enthusiasm for the new.
I am excited about Agent Skills not because of how they are being used with AI, but because I see them as a bridge with business stakeholders. I am excited about JSON Structure, because it provides a fresh opportunity to validate and compose with schema, helping engineers and business stakeholders get on the same page–without all the baggage (or momentum) of JSON Schema. I’ve even managed to find some enthusiasm latterly for MCP when it comes to providing guardrails for developers as part of their chosen stack. I’ve long gotten over there needing to be a single specification to rule them all, and I celebrate whatever specification stack your enterprises uses, and will help you strengthen and expand upon what you have.
Honestly I struggle with what we leave behind in all of this. I was once a big fan of WSDL, and still act as champion for developers who have to maintain integrations on top of SOAP APIs. I am also struggling with moving on from OpenAPI, JSON Schema, and other bedrock specifications of my work and identity. Like SOAP, I don’t think either of them are going away. But I do feel the disruptive sands shifting again under our feet, and I’m seeing spec-driven development combined with AI introduce opportunities for moving things forward in positive and negative ways.
Moving forward I am working to get to know each spec, as I’ve done in the past. I want to understand what it does and how it can be applied. Then I want to find the stories that matter the most to both engineers and business stakeholders. I am also first and foremost a storyteller, so I am looking for stories. Swagger emerged out of Wordnik, a dictionary API, and I believe this was for a reason. I’m looking for the same magic across MCP, Agent Skills, Agent-2-Agent, CLAUDE.md, and other artifacts today. I’m interested in where the story is, but also where the rubber meets the road when it comes to helping find the signal in the noise.
Ultimately I am going to be less loyal to any individual specification. I will support them all and be critical of them all–even mine. I am going to ride the waves of each spec. There is a new level of disruption and volatility occurring right now and while AI is the focus of the conversation, the extraction of value at the specification level is where the real shift occuring. I need to be able surf across the top of all the specs, and this means understanding them all, and being able to apply what matters in any given moment. It is definitely a new storytelling medium that exists at a new scale, which involves surfing at the intersection of determinism and non-determinism.
