Jens Neuse is the co-founder and CEO of WunderGraph, the company behind Cosmo — the open-source GraphQL federation stack running inside enterprises like Mercedes, Rivian and eBay. This conversation went two ways. Jens walked me through what GraphQL federation actually buys an organization, and it is not what the versus debates say it is: fragments that let a UI evolve without leaving dead fields behind in a REST API forever, and entities that let separate teams agree on a key so their data can be joined deterministically rather than hopefully. Then he made the sharpest claim I have had on the show — that once you go past about five MCP tool calls you are gluing them together with inference, and inference is slow, expensive and unpredictable, so the architecture is simply wrong. Models are good at writing code; let them write a program and run it. We also talked about who owns the megaphone in every technology wave, why he does not care whether GraphQL beats REST as long as everyone agrees on a standard, and how it is more than a little strange that we invented Web MCP while the user agent has been sitting in the HTTP header the entire time.

Lukasz Gornicki spent years as executive director of AsyncAPI, which means he has already run the experiment everyone else is still theorizing about — what a foundation actually does for a specification once the announcement is over. He is now part-time, and hoping to go full-time, at Open Source Europe, a Belgian association built as a European alternative to the US-based Open Source Collective. I brought him the question I have been carrying since I forked Spectral: where should a rule-set specification live. What I got back was not a recommendation, it was a set of distinctions. The Linux Foundation gave AsyncAPI intellectual property and trademark protection and very little else — protection that turns out to mean a free letter from a lawyer, after which you need a litigation budget you do not have. It funds no engineering work, which is why Open Source Collective had to exist alongside it. And one sponsor declined to renew specifically because AsyncAPI was part of the Linux Foundation. Lukasz walks the other options too — why Apache is the only foundation that will actually enforce a governance model on you, and the price it charges in your project's name; why Eclipse left the US for Canada; and what Open Source Europe can and cannot do today. Underneath all of it is the question he says has to come first, and that nobody in this conversation has answered yet: ask the people who would fund this what they actually care about, because open governance needs no foundation at all.

Correction, added after recording: in this conversation I say the German government reached out and that Germany is committed. That overstates it — Germany has made no commitment. The Netherlands is the only government with a published, documented dependency on this format, in the developer.overheid.nl post of 31 July 2026. Conversations with Italy are ongoing and nothing is committed there either.

Danny Preussler is an engineer on the API team at SoundCloud in Berlin, and he came up entirely on the client side — BlackBerry apps, then Android, then eBay, Groupon, and Viacom before landing at SoundCloud seven years ago. This is a conversation about one of the great API origin stories and one of the great API closures. SoundCloud shipped a public API in 2008, one year after the company started, and literally built its own product on the same API everyone else could use — the Android lead got hired because he wrote a better SoundCloud app on the public API than SoundCloud had. Then the major label content arrived, the scrapers and downloaders arrived, a hundred thousand new apps registered in 2016 alone, and in 2017 the program closed for what was supposed to be a temporary pause. Danny walks through the soft reopening of the last two years, the Tyk gateway that made it possible to reopen at all, and the moment that changed everything: they handed an agent the task of creating a SoundCloud app, it could not find a token, so it scraped the website for one and quietly rewired itself from the public API to the internal API. We also get into why deprecated endpoints started spiking after five years of silence, why their ticket volume went up because of a stale doc page, and where Danny lands on MCP when he turns the questions back on me.

Emma Kriskinans is VP Global Marketing at Tyk, and since March she has done the bulk of her working day inside Claude Code and VS Code — not because anyone asked her to, but because she wanted to learn it at the harder level. This one is a two-way comparison of notes on using AI for go-to-market, between a marketing leader running a team inside a company with a brand to protect and a one-man band who answers to nobody. Emma walks through why she chose the IDE over the browser, using the UK driving test as the analogy — learn on a manual and you can drive an automatic, but not the other way around — the folder structure she keeps for Tyk work, advisory work, and personal work, and the trust hierarchy document that tells the model which sources to weight and when to come back and tell her it does not have one. We get into what she deliberately does not use it for, because she considers herself a writer and the generated prose lands in the uncanny valley, and into the etiquette forming around AI disclosure, the shaming that runs in both directions, and why neither of us thinks the dividing up of people is an accident. It ends somewhere neither of us planned, on novels, on the backstory you write and never publish, and on whether the humanities graduate's moment has finally arrived.

Sebastian Loch runs business development at fatsecret, where he has spent thirteen years building out the largest verified nutrition database in the world — country by country, across sixty-two markets and twenty-six languages. This is a check-in on a conversation we started back in 2024, when he was genuinely worried that the AI rush would strip-mine the proprietary data they had spent twenty years verifying. Three years on, the answer is more interesting than either the doom or the hype: the moat held, and it held because of scale, global reach, and two decades of investment in the API. Sebastian walks through how the arrival of LLMs made ten years of their own computer-vision work obsolete overnight and simultaneously handed them the image recognition engine their developers had been asking for since 2016, why they will use AI to improve every process around the data but never to generate the nutrition data itself, and how vibe coding has driven a hundred- fold increase in daily developer signups at a bootstrapped, profitable, never-funded company. We also go two ways: Sebastian turns the questions back on me about API sprawl, AI slop on the API side, and whether any of it can be rated and ranked.

Phil Leggetter is a developer experience and developer relations leader who heads product-led growth and DevRel at Hookdeck, a small team building an event gateway for event-driven applications. In this conversation we dig into why so much of the event-driven world is over-architected, and how Hookdeck packages durable queueing, filtering, and transformation behind plain HTTP so that everyday events do not require a multi-year Kafka project. Phil walks through the story behind Outpost, their open-source implementation of event destinations, and how an unplanned demand for a hosted version turned into a second product alongside their event gateway. We then turn to AI: how Hookdeck is making its developer platform agent-ready, whether agents will grow the pie of event producers versus consumers, and how a small team scales by mapping job specs to agent skills and standing up a fleet of role-based agents. It is a grounded, builder-to-builder look at where events, APIs, and AI agents are converging.

Lin Sun is the Head of Open Source at Solo.io and one of the most recognizable leaders in the cloud-native world — an Istio co-founder and maintainer, a recent member of the CNCF Technical Oversight Committee, and a KubeCon co-chair. In this conversation we start with who she is and the career leap she took from fifteen-plus years at IBM into a thirty-person startup, then move into the work pouring out of Solo.io's open source efforts: donating Gloo to the CNCF as kgateway, launching kagent, and building Agent Gateway from the ground up for the dynamic protocols of agentic APIs like MCP and A2A. Lin makes the case that AI workloads are "microservices plus," why gateways remain a critical control point for policy, security, and observability, and how a gateway can smooth breaking spec changes for users. We close on her hard-won advice for taking a project through the CNCF's sandbox, incubation, and graduation stages — and on what keeps her going after twenty-five years in tech.

Sameh Jarour is a senior product marketing manager and AI evangelist at Apify who made a deliberate jump from marketing and business into the deep technical end of the AI world. In this conversation we start with his journey and Apify's bet on the MCP protocol — a single token fronting 30,000 tools, dynamic tool discovery, and a developer marketplace built on Dockerized actors. From there we widen out to what the last two years of AI have done to the developer market, how Apify acts as an abstraction layer that turns anything into an API, and what enterprises in the GCC actually want from agent work. We get into the forward-deployment consulting model, the death of feature-centric SaaS, and why people with a commercial background who went technical are positioned to win. We close on the shift from features to outcomes that Sameh sees reshaping product marketing.

Tony Tam is the creator of Swagger — the specification that became the OpenAPI Specification — and the founder of Wordnik. In this conversation we go all the way back to the late-night desperation at Wordnik that started it all: a six-person team that couldn't keep up with hand-writing clients and documentation, and an email Tony sent at three in the morning asking what if the API could just describe itself. From there we trace the whole arc — the accidental naming of Swagger, the moment a stranger turned up in an IRC channel to say they were pushing two billion dollars a day through it, the inversion from code-generates-spec to spec-generates-server, and the move into the Linux Foundation as OpenAPI. We talk about the two moments Tony realized this would matter for AI, why an OpenAPI doc is now a ticket into AI infrastructure, whether AI even needs APIs at all, and where the human craft of designing the contract still adds value in a world where machines write most of the code.

Sudeep Goswami is the CEO of Traefik Labs, the company behind the cloud-native, Kubernetes-native gateway that grew out of the widely adopted open source Traefik proxy. In this conversation we go back to first principles and ask what a gateway actually is — and Sudeep reframes it as runtime governance: writing granular policies, enforcing them in real time, and observing what actually happened. From there we trace how that same job has to stretch across three new personalities — the API gateway, the AI gateway, and the MCP gateway — without fragmenting an enterprise into three vendors and three operating models. We get into federated versus centralized platform strategies, why Kubernetes-native discovery makes shadow APIs disappear, how agents experience guardrails when they get blocked, and where Sudeep thinks the real constraint on putting AI into production now lives.