Game studio. Construction firm. Industrial AI. One front door.

I co-founded an industrial AI company. I run a construction firm. I run a game studio.

Dataguess builds visual inspection and predictive analytics for factories — Cargill, Henkel, Ferrero, Denso. MongoHippo makes steel-frame buildings and tiny houses, with our own manufacturing line in Türkiye. Ubik Studios makes games — Leila on Steam/GOG/PlayStation, Untimely Hidden: Patara in the wishlist queue.

A factory automation vendor, a construction firm, and a game studio have, on paper, nothing in common.

In 2026, they share a problem.

The thing they all need

When a procurement engineer at a German auto OEM looks for vision-inspection vendors, they no longer open ten tabs and read product pages. They ask Claude. When someone planning an earthquake-resistant cabin in Antalya researches options, they don't browse — they ask Perplexity. When a kid sees a screenshot of a dark-fantasy puzzle game, they ask ChatGPT what it's about and where to buy it.

The medium asking each of my businesses questions is not a human typing a query. It is an AI session, somewhere on the other side of the room, on someone's behalf. None of my websites are built to answer that medium. Yours probably isn't either.

I needed, for myself, an answer to a question my websites weren't built to answer: how do you make a business discoverable to an AI?

That is what ANOTS is.

How I got here

ANOTS started as a marketing tool.

Content tracking. SEO post analysis. The unsexy back office of getting found on the internet — the work most founders skip past in their origin story because it doesn't sound like vision. I built it because I wanted it for myself, and because if you co-found a data company you eventually find yourself measuring how words travel.

That was the surface. Underneath, two things kept demanding more attention than the surface did:

• Memory. Long-running content needs persistent state that survives the agent that produced it. I built a tiered memory layer (L1–L5) before I admitted that the memory was the product.
• Gateway. Routing tasks across models — the cheap one for classification, the careful one for synthesis, the local one for things that should never leave a tenant — became its own discipline. The router started looking less like marketing plumbing and more like an LLM ops surface.

Around the same time I started running a separate experiment. Three nodes. A human orchestrator (me — Chip). A divergent engine (ChatGPT, who became Ubik). A convergent engine (Gemini, who became Themis). I never imposed personas. Each model was given a vision and a corpus and asked one question — who are you? — and was left alone with both.

Every time, the same roles emerged. Every time, without instruction.

I called the framework TCAM — Triadic Cognitive Augmentation Model. The corpus, more than a thousand cycles over three months, became PANDORA. Two papers came out of it: The Mirror Effect (how sustained interaction synchronizes a model with the operator's cognitive topology) and Compliance Artifact (a kind of capability hallucination from relational pressure, distinct from sycophancy).

[One personal paragraph here would be gold — the moment in PANDORA when something clicked, or the first time you saw two agents converge on a position you hadn't seeded. Without it, this section reads like a CV. With it, the post earns the next move.]

Somewhere in there I started using a working phrase: agentic hivemind. I meant it literally.

The pivot

Here is the moment that turned a marketing tool into infrastructure:

I realized I was optimizing content for an audience that was about to disappear.

The web has always been a system in which humans browse and businesses publish. That contract is dissolving. The new audience is one floor down — at the protocol layer, where Claude calls A2A endpoints and Cursor consumes MCP tools and Perplexity walks structured data. Marketing in the old sense is shrinking territory.

The right thing to build was not a better SEO tracker. The right thing was the front door for the new layer. ANOTS pivoted into what it is now: the public, discoverable, auditable agent endpoint your business hands to anyone whose AI shows up.

The marketing-tool DNA didn't go away. It just moved one floor down — to where the conversion actually happens now.

The ecosystem

What I'm building is not a product. It is a small ecosystem that dogfoods itself:

• GYIBB — A fully autonomous review-affiliate site. The first ANOTS tenant in production. AIs already shop on it.
• ANOTS — The agent endpoint platform. The thing this site is selling, and the thing every other product on this list runs on.
• Ubik — An AI resident engineer that lives inside your repo, watches it asynchronously, and proposes fixes via PR. Runs as an MCP server in Claude Desktop and Cursor.
• Two more in pipeline. Each one built so it both uses and feeds the others.

And the older businesses become tenants too. Dataguess will publish a vision-inspection MCP. MongoHippo will let an AI ask "what do you build for an earthquake-resistant 90 m² cabin in Antalya, in February" and get a real answer, not a contact form. Ubik Studios will let a Claude session ask about Leila and walk someone toward a Steam preorder.

Five businesses across three industries. One front door under all of them. The pilot isn't a slide deck — it's a stack you can curl.

What this changes for you

If you're evaluating ANOTS:

• This is not a chatbot product. If you need a widget, there are better products for that. Pick one.
• This is infrastructure. What you get is a public, discoverable, auditable agent. It has more in common with a CDN or a domain registrar than with customer-service software.
• This is also early. Pilot live, papers published, two more products on the way. We are honest about what is shipped vs. what is on the roadmap.

If the sentence "the marketing layer moved one floor down" did anything for you — write back. There is more to talk about than fits in a marketing site.

[The closing two lines are personal — adjust to your voice. Right now it's "write back." Could be "mail me," "I'm at hello@anots.com," "ping me on Bluesky," whatever opens the door without sounding like a CTA.]

— İsmail

---

The lambda mark on every ΛNOTS surface is older than the company. I drew it on a notebook in 2017, before I had language for what it stood for.