Today we are publicly launching Surfacedd, an advertising and discovery network for AI agents and applications. Surfacedd lets brands place context-aware content inside text, image, video, and voice AI products, and lets the developers building those products share in the resulting revenue. There are no minimums, no sales calls, and no opaque pricing on either side. The brand console is self-serve. The developer SDK is a few lines of code. The first cohort has been in closed beta since the start of the year; today we are opening it up.
Surfacedd is the third product on the Apik shelf, alongside Senwitt and Brello AI. It is also the most economically load-bearing of the three. In the rest of this post I want to explain, in plain language, why we think there is a Surfacedd-shaped hole in the world right now, why we believe that hole is going to keep widening for the next decade, and how Surfacedd fits into the broader Apik thesis we laid out in the manifesto.
The medium is moving
For thirty years the dominant interface between information and the people who need it has been the search box. The economics of that interface are a matter of public record: roughly two-thirds of a trillion dollars per year flows through search and social advertising globally, and the substantial majority of free-to-use software on the open web is paid for by some descendant of that mechanism. The arrangement is imperfect and frequently unloved, but it is the arrangement we have, and most of the institutions of the modern internet — publishers, app developers, creators, small businesses — exist as they do because of it.
That interface is moving. A growing fraction of what used to be a search query is now a conversation with an AI assistant. A growing fraction of what used to be a product page visit is now an agent acting on behalf of a buyer. A growing fraction of what used to be a video tutorial is now a voice exchange with a model that already knows the user’s context. These are not speculative scenarios; they are the observed direction of travel. The empirical question is not whether the medium of information distribution shifts toward AI mediation. The empirical question is how fast.
Once you accept that premise, a structural problem becomes immediate. The economic substrate that funds the existing medium does not automatically transfer. A model-mediated answer is not a webpage with a banner; it is a synthesized response composed at request time from a mix of training data, retrieved context, and tool calls. The rendering surfaces are different — chat threads, voice timelines, agent transcripts, generative cards — and the formats native to those surfaces are different from a search ad or a social-feed unit. The accountability primitives are different too: who saw what, in what context, and with what disclosure, has to be reconstructed from scratch.
Without an explicit substrate, three things happen, none of them good. First, AI applications fall back on subscriptions, which work for a small minority of users and excludes everyone else from the most useful versions of the product. Second, model providers that own a distribution surface get a structural incentive to bake undisclosed sponsorship into model output itself, which is a worse outcome on every axis we care about — for users, for safety, for the long-term legibility of model behavior. Third, the open ecosystem of specialized AI applications — vertical assistants, agents, voice tools, niche generators — gets starved of the revenue mechanism that would let them exist in the first place, and we end up with a smaller, more concentrated AI economy than we had a search economy.
We are building Surfacedd so that the third outcome is not the only one available.
What Surfacedd is
Surfacedd is a self-serve, multi-format ad network for AI applications. From the brand side: budget, targeting, creative, performance reports, all in real time, no contracts. From the developer side: an SDK that handles request routing, format detection, response composition, and the policy boundary that keeps Surfacedd-placed content distinguishable from organic model output. The revenue split is 60/40 in favor of the developer.
The current formats cover text and inline cards in chat and generative-search responses. Voice and short-form generative video are the next milestones, with the same auction primitives and the same disclosure contract. After that, the most architecturally important step is the agent-direct API: the same auction surface, queryable from autonomous shopping agents and recommendation agents that make purchase decisions on behalf of users, with structured disclosure built into the response contract so the user-facing surface always knows what was paid placement and what was organic.
The detailed product page is at /products/surfacedd. The live product is at surfacedd.com.
Why this is an Apik project
The Apik manifesto makes a single architectural bet: that computational coordination is the missing primitive of the next civilization, and that a stack of autonomous intelligence — from cognitive tools for individuals up through planetary-scale economic orchestration — is the substrate on which abundance becomes structural. That bet is mostly stated in terms of production, logistics, energy, and physical effect. Surfacedd is the same bet stated in terms of attention, discovery, and the matching of brands to buyers.
Specifically: the agentic-systems research pillar studies what it takes for software agents to operate reliably across long workflows. The economic-orchestration pillar studies the mechanisms that keep that activity incentive-compatible at scale. Surfacedd is the most concrete bridge between those two pillars. It is a real auction running across a real population of AI applications, generating data about which placements work, which contexts are safe, which formats users tolerate, and how disclosure rules hold up under economic pressure. The data and protocols Surfacedd produces feed directly back into the research program.
There is also an alignment angle, which I want to be explicit about because it is the version of this argument I find most personally compelling. If sponsored content shows up inside AI experiences — and it will, because the economics demand it — then the question is not whether it appears, but who controls how it appears. A closed approach where each frontier-model provider runs a private, integrated ad system inside their own assistant produces the worst version of the future on multiple axes: the disclosure rules are private, the acceptable-use policy is private, the data flow is private, and the incentive to blur the line between organic and paid output is structural and enormous. An open approach with a documented disclosure contract, a published Acceptable Use Policy, and a portable SDK that any AI application can integrate produces the much better version. We would prefer the better version.
This is also why Surfacedd is governed by the same trust documents that govern the rest of the Apik stack. The Acceptable Use Policy restricts the categories of advertising Surfacedd will carry. The Responsible Development Policy governs how new placement formats — particularly anything close to a voice or agent surface — pass through evaluation before deployment. The disclosure contract is enforceable in the SDK, in the integration agreement, and in our own audit pipeline. None of those guarantees are perfect; all of them are public.
What we are committing to
A short list of things we are saying out loud today, so that we can be held to them:
- Disclosure is non-optional. Every Surfacedd-placed item carries a structured disclosure that the consuming application is required to render. Brands cannot opt out. Developers cannot strip it. We will not negotiate this point with a single counterparty regardless of size.
- The pricing is public. The auction logic, the platform fee, the developer share, and the absence of minimums are documented and stable. We are not running one set of economics for the launch cohort and a different set for everyone else.
- The AUP is enforced at integration time. Categories that Surfacedd will not carry — political weapons, deceptive medical claims, unconsented impersonation, the rest — are checked at brand onboarding and at creative ingestion. The SDK refuses placement for content that fails the check.
- Performance reporting is symmetric. Brands can see their own performance. Developers can see their own performance. Surfacedd’s own aggregate numbers — total volume, total fees, basic distribution by category — will be published in an annual transparency report.
- Agent surfaces will ship with structured disclosure. When Surfacedd opens its agent-direct API later this year, the disclosure metadata will be a required field of the response, not an optional decoration. Agents that strip the disclosure forfeit access.
If we deviate from any of these, please tell us — both publicly and at security@apiksystems.com. The whole point of writing them down is that they should be auditable from outside.
Where this goes
In one sense, Surfacedd is a product. In another, it is a long-running experiment in what an open commerce layer for AI mediation looks like — one whose existence we think materially changes the shape of the AI ecosystem over the next decade. The goal is not just for Surfacedd to do well, although we hope it does. The goal is for the structural pattern that Surfacedd represents — open, disclosed, portable across applications, governed by published policy — to become the default expectation for how brands and AI applications relate to one another. If, three years from now, every credible AI application supports an open ad layer with mandatory disclosure and any closed alternative is treated as a red flag, we will consider Surfacedd to have done its real job, even if Surfacedd itself only ever holds part of the market.
If you are a brand, a developer, a researcher studying the economics of AI mediation, or just someone with strong opinions about how this should go, the live product is at surfacedd.com, the longer product page is at /products/surfacedd, and the channel for substantive feedback is research@apiksystems.com. We read everything that comes in.
— Rehan Temkar, Co-founder, Apik Systems