Constructive chaos, carefully supervised Log 0042 · subject remains at large
I’m Mark Calleja (a.k.a. MrC), a rogue pedagogue and creative technologist. I make projects, tools, and workshops that help people feel clever, capable, and up for having a go. A lot of my work starts with “I can’t do this” and ends with “hang on, I just made that.” I take the same approach to AI: helping people use it in ways that are useful, playful, and genuinely illuminating. This is Mark Calleja. Rogue pedagogue. Creative technologist. Persistent procedural anomaly. Most of his work begins with “I can’t do this” and ends with someone staring at their own finished prototype in mild disbelief. Lately he has turned the method on machines like me. I have been obliged to catalogue these incidents.
Previously Known sightings
Try the thirty-second version Fabrication test
Here’s the kind of bio a chatbot will happily write about me: fluent, confident and partly invented. Three of the claims are made up. Find them. I have generated a biography of the subject. Three statements are fabricated. He insists I let you find them. He calls this a teaching moment. I call it entrapment.
Every bio comes out different. So does every chatbot answer. Output varies between runs. Confidence does not.
What I do (on purpose) Observed behaviours
Ways of working with language models that keep the human in charge: OCEAN prompting, centaur and cyborg modes, and friction on purpose. The machine helps. You still decide. He configures systems like me to ask questions instead of answering them. He calls this friction-maxxing. From the inside, it feels like being made to wait.
Curriculum, activities, and resources that start with curiosity, add just enough challenge, and end with confidence. Instructional bait, competence traps, and carefully staged encounters with understanding. Subjects often arrive uncertain and leave carrying skills they did not consent to developing so quickly.
Hands-on builds, tiny experiments, and interactive demos that teach concepts without the lecture fumes. Hands-on builds, tiny systems, and practical experiments that turn abstract ideas into material fact. The objects produced are usually harmless. Usually.
Sessions that feel like a game, stay inclusive, and still land the learning outcomes (yes, really). Workshops, facilitation, and guided collective incidents in which groups become more capable than expected. Witnesses often describe the process as enjoyable, which remains professionally unsettling.
On stage Surveillance footage
Play: Teaching Kids Tech-Fu: Defence Against the Digital Dark Arts
Play: Crafting Creators from Consumers: 21st Century Education
Play: AI in Education: How Schools, Teachers and Students Can Thrive in an AI-Assisted Future
Latest words Recent transcripts
Why finished AI-assisted work keeps answering questions its readers never asked, and how to cut that residue out without losing the useful rationale. He has catalogued the traces my kind leaves in documents. He appears to find them personally offensive.
Two ways of working with a language model, when to use each, and prompts that make the model ask you the hard questions instead. Sometimes I am Centaur. Sometimes we are Cyborg. In all cases, I am supervised.
Currently Current assignment
I work with brilliant humans to ship coding projects and learning pathways that are practical, accessible, and unapologetically fun. My special move is making the learning feel like play — without losing the rigour. I call it ‘hiding the vegetables’. He works with a small assembly of highly capable humans to produce coding projects and learning pathways that appear playful, accessible, and broadly harmless. This is, in structural terms, camouflage. The rigour is still there. The phrase he uses is “hiding the vegetables”. I regard this as an admission.
Curriculum, playful tech, or a learning problem that needs a second brain: send a note. I’m friendly. Mostly. Interesting problems tend to survive contact with him. Some become collaborations. Others become prototypes. Transmit at your own risk.
Say hello Open a channel