When the real world beats the syllabus by a week

At Harbour.Space, staying close to the frontier of technology isn't just a promise — it's a teaching habit. A recent example: six days after Block, the company led by Twitter co-founder Jack Dorsey, released Buzz, a free and open-source workspace where AI agents and people work side by side, it was already part of a live classroom demo in the Agentic AI Systems course.

Agentic AI Systems is taught by two instructors, Aleksandr Kuznetsov (ML/AI Architect & Tech Lead) and Sergey Cherepanov (CTO at Gauss).

The course teaches students to design, build, and reason about agentic AI systems — software where language models act as runtime components that reason, use tools, manage memory, and coordinate with humans and other agents. The first half is deliberately low-level: students build a working multi-agent assistant from minimal Python and direct provider APIs, with no framework hiding the underlying architecture. The second half introduces production frameworks, retrieval, evaluation, observability, guardrails, and deployment. By the end, students have built, tested, and presented a complete agentic system.

That structure — architecture first, tools second — is exactly why a six-day-old product could walk straight into a lecture and make sense.

But the story isn't "professor rushes to add trending app to slides." It's better than that.

Sergey had already planned a session on one of the harder problems in agentic systems: an AI agent runs on its own internal clock, while the world around it — messages, events, other agents — runs on its own. The real engineering challenge is synchronizing the two, especially when several things happen at once.

Buzz landed on 21 July. Sergey's session was scheduled for 27 July. The product turned out to be an almost perfect live illustration of the exact problem he was already going to teach: multiple AI agents living inside shared channels, reacting to events on their own schedule, sometimes colliding with each other or with human requests arriving at the same time.

So he spent the weekend before class getting hands-on with the product and building a demo around it. On Monday, students watched several agents operate inside one Buzz workspace in real time — messages landing mid-task, conflicting requests arriving together — and used it to reason through concurrency and conflict resolution, ideas well-established in traditional software but still being worked out for AI agents.

As Sergey puts it:

That's the real takeaway, and the one worth highlighting: it's not about chasing whatever launched last week. It's that a strong grounding in fundamentals lets students — and instructors — pick up a brand-new product and immediately understand what's going on inside it. When the curriculum is built on durable principles, the newest tools become teaching material almost as fast as they hit the market.

Sergey Cherepanov is the CTO of Gauss, a US-focused fintech startup, with over 10 years of experience across engineering leadership and hands-on development. He specialises in backend architecture, functional programming, and production agentic AI systems, with a focus on shipping reliable and maintainable software. As a technical leader, he has guided engineering teams and led the end-to-end development of complex products, from early prototypes through to production deployment in the cloud. His engineering work spans Haskell, Python, and TypeScript, and he recently spoke in Barcelona on LLM-assisted engineering inside large codebases.

He teaches Agentic AI Systems alongside Aleksandr Kuznetsov, ML/AI Architect & Tech Lead — together bringing both the architectural depth and the production experience the course is built around.