About Inoku

Inoku is an independent research lab working on one question: how do we make AI agents tell the truth about their own work?

An agent that says “done” when the task isn’t done wastes the time of the person who trusted it, and sometimes much more than time. We build ways to measure how often that happens and ways to reduce it, so that an agent’s report matches what really took place, including an honest “I couldn’t” when a task can’t be done.

What we’re building

Inoku Verified™, in development. An independent evaluation of how often an agent’s “done” is true, on a private test suite it can’t have trained on. See Inoku Verified.
Private evaluations, design partners first. Credible third-party numbers for the teams that build models and agents.
An independent record for production agents, at the research stage. What your agents actually did, kept where the agent can’t change it.

Nothing on this list is for sale today. Each page says what exists and what doesn’t.

The mark

The navy frame is the system an agent works in. The blue frame is what comes back out of it: the record. The square between them is the claim. The dotted zero in the name is the programmer’s zero, the one that can’t be mistaken for a letter.

The logo files, colours and usage rules are on the brand page.

Founder

Ivan Popov has worked in web and systems engineering for more than twenty years, across backend architecture, platform performance, technical SEO and migrations, and runs the IvanLabs advisory practice for revenue-critical web platforms. He started Inoku in 2026 after seeing how often agents report success they didn’t achieve.

How we work

Evidence over assertion. Every factual claim we publish links to its source. The site is the first proof of the promise.
We publish methods and results, including the negative ones. A result that didn’t work is still a result.
We say what is built and what isn’t. “In development”, “research stage” and “design partners first” mean exactly that.
Open by default. We plan to publish our tools and benchmark openly. Organisations above 10 people or $1M in revenue pay for private evaluations and support; researchers, educators, open-source projects and smaller teams don’t.
Independent. Payment never changes a score.

How we use AI

Inoku uses AI models across the company, and says so: in research and development, in engineering, in writing and marketing, and in day-to-day operations. Claude (Anthropic), GPT (OpenAI), GLM (Z.ai) and Qwen (Alibaba Cloud), alongside smaller open-weight models, help with literature search, code, analysis, drafting and editing, including the text on this site, the research notes, press releases and posts elsewhere. They are also the kind of systems whose claims Inoku sets out to measure, which is one more reason to keep the measurement independent of any one of them.

Three rules apply. A person decides what is published, and Ivan Popov is responsible for every page. Every factual claim is checked against its source before it goes live, whoever or whatever drafted the sentence. And a model’s output is treated like any agent’s report: a claim to be verified, not a fact.

Where we are

Inoku is based in Larnaka, Cyprus, and was founded in 2026. The company details are in the legal notice.

If you want to work with us, apply to the design-partner programme or get in touch.