Illustration of a scholar and a red hooded creature pointing at each other

Stanislav Lvovsky · LLM tutor & researcher

Get to know the Thing

One-on-one training in the practical use of AI for academics – across research, writing, teaching, and the administrative work in between.

Independent researcher · Oxford DPhil · Creator of courses on applied AI at Prague Media School

200+ researchers, journalists & editors trained · 10 cohorts since 2023

Book a free 20-minute intro call

Four courses a semester, committee work, your own research somewhere in-between – and now this. Your students are already using AI. The rules about it change every semester, when they exist at all. And everything written to explain the technology either wants to sell you something or assumes you’ve always secretly wanted to learn linear algebra. You don’t have time to figure this out. You also can’t quite afford not to.

The odd thing: the chat window is a deceptively simple front door – most of what the instrument can do never shows up there uninvited.

The great thing: working with LLMs is, at bottom, working with language – and there you are not a beginner.

So no – you don’t need machine learning or “prompt engineering.” What you need is to tell at a glance what the thing will do, what it won’t, when to trust it – and where to look when it surprises you. That is the kind of knowledge that matters here: a trained intuition for generative AI. None of it is beyond you; most of it asks less of you than peer review does.

The Smoothness Trap

Fluent, plausible output that mimics expertise without reliable reasoning. It is the single most common way AI misleads careful readers – and learning to catch it is where this training begins.

As presented at iGAIAS 2026, University of Reading: The Smoothness Trap and What to Do About It.

Every engagement starts from the same core:

  • How these models actually work – how they’re trained, how they generate text, what to realistically expect – and where they fail: politely, in fluent prose, without telling you.
  • Handing routine work to the model – correspondence, reports, formatting, first drafts.
  • Keeping track of what the model produced and what you verified.
  • At this point the work you actually came for moves back to the top of the pile.
  • Briefing the model well enough to enjoy exploration – mapping a field, testing an argument, drafting.

The rest is built on your material: we map where your time actually goes on the first call, and sessions run on your real tasks, not exercises. Depending on your load, we work on things like:

Teaching and academic work. Course materials, rubrics, committee documents, recommendation letters, funder reports – done faster, and adapted for a new audience without rebuilding from scratch.

Reading and research. Research with LLMs: source-linked summaries and maps of a field you can check – no plausible confabulations. And the model as an instrument of critical thinking: analytical protocols you keep and reuse, procedural prompts, reading a problem through more than one lens.

Writing and argument. Clear delegation boundaries. Citations, references, and bibliographies done to the journal’s spec. How to use the model as a demanding reviewer and sparring partner without letting it flatten your reasoning or your voice.

Limits and records. Request types where models reliably fail, and how to redesign them. What should never go into a model. How to create an AI-use record you can show a journal, funder, or research office.

Book a free 20-minute intro call

Format and price

Start free.

The first step is a 20-minute introductory call.

$50 per session.

One-on-one, online, one hour. Deliberately priced for academic budgets.

Scheduling.

8 AM–6 PM Eastern Time
8 AM–5 PM Central Time
8 AM–4 PM Mountain Time
8 AM–3 PM Pacific Time

We agree the scope and cadence on the first call; academic schedules vary, and the training should fit the work rather than compete with it.

Pick a slot ↓

Two details are helpful: your name, and your field and role. A phone number is optional – only as a fallback in case of connection issues.

My background

Independent researcher. DPhil in Medieval and Modern Languages, University of Oxford; MA in Public History, University of Manchester / MSSES. More than 200 researchers, journalists and media professionals have participated across ten cohorts in the courses I designed and taught at Prague Media School since 2023, including the currently active Neurologic: Thinking with Large Language Models.

Articles

  • Reading One Constitution: Source Traditions, Normative Genre, and Conceptual Metaphor in Anthropic’s AI Governance Document, under review. Preprint (published under my legal name) available at SSRN or dx.doi.org/10.2139/ssrn.6946338. Here’s also a guest post based on this article for the Making Science Public blog.
  • AI Writing Detection in Higher Education: Population-Differentiated False Positives, Statistical Convergence, and the Sociology of Algorithmic Classification, under review. Preprint is available on SocArXiv.

Essays

  • The Speaking Kind; or Return of the Pharmakon. On talking machines, boundaries of the human, and the refusal to choose a side ↗
  • The Encyclical and the Lab. How “Magnifica Humanitas” built the most serious AI framework to date ↗
  • Imagination, Aligned. What AI can and can’t do with a story – and what is actually under threat ↗
  • Hauntology of the Latent Space ↗ AI and poetry ↗

More on Substack

Presentations

  • iGAIAS 2026, University of Reading: The Smoothness Trap and What to Do About It: AI Literacy Beyond Tool Training.
  • ProZ AI Expo 2026: Beyond Fluency: Critical AI Literacy for Language Professionals.
  • AMPS Learning & Teaching Conference: A Focus on Pedagogy 2026: Across Teaching, Theory, Technologies & Times (November 18–20, 2026, forthcoming).

Contacts