Kontur engineer on Habr compares D&D wizards and warlocks to the LLM era

Alexander Ovchinnikov, a senior QA specialist at Kontur, has published an essay on Habr titled "From Wizard to Warlock: magic in D&D as a metaphor for the LLM era." In it, he uses the switch between character classes in the tabletop game Dungeons & Dragons to describe what is happening to people's work now that large language models have arrived.

What do D&D wizards and warlocks have to do with it?

In the tabletop role-playing game Dungeons & Dragons, a wizard spends years studying spells from books and learns the mechanics of each one inside out. A warlock, by contrast, draws magical power from an external source — a patron — and can wield powerful effects without understanding every detail of how they work. Judging by the title, the author uses this split between "magic from knowledge" and "magic from an external source" as a metaphor for the LLM era, although the source does not include the essay's text beyond its introduction.

• Author: Alexander Ovchinnikov, senior QA specialist at Kontur
• The piece was published on Habr under the title "From Wizard to Warlock: magic in D&D as a metaphor for the LLM era"
• By his own account, the author has been actively using LLMs for personal tasks for several years

Kontur (SKB Kontur) is a major Russian IT company based in Yekaterinburg that builds services for businesses: electronic reporting, electronic document management, and accounting systems. The fact that a piece about LLMs comes out of a QA specialist's practice at a company like this is telling in itself. This isn't about developers of AI products, but about an adjacent technical discipline where large language models are becoming a working tool alongside familiar testing and automation software.

Why does the author consider showing LLMs to others more important

than using them himself?

Ovchinnikov writes that the most interesting experience for him wasn't using LLMs in his own projects, but starting to show these tools to other people. The phrasing highlights a gap that often opens up around new technologies: some people already work with LLMs fluently as a working tool, while others haven't yet seen what these models can do in practice. They need someone to show them a concrete example on their own task, not an abstract demo.

Taken through this metaphor, the issue is less about the technology itself than about how people relate to the results it produces. Some want to understand how it works under the hood before they start using it; others are willing to use the tool as-is if it solves the problem here and now.

What this means

The essay captures a theme that comes up more and more in professional communities: LLMs are changing not only what specialists actually do, but also how they explain what's happening to their colleagues — through metaphors, analogies, and live demonstrations rather than dry instructions. For technical specialists rolling out these tools within their teams, leading by example and showing things hands-on often proves more convincing than any formal training.

It's also telling that pieces like this come not from the developers of AI models themselves but from adjacent technical professions such as testing. That indirectly shows how widely LLMs have already spread into workflows beyond the narrow AI community.