How the TelosExpedition works.

It is a six-week critical-thinking and AI-literacy course for ages 11 to 14: twenty-five hands-on missions across five worlds. Here it is week by week and day by day, plus the research behind why it works.

The goal is not to keep kids away from AI. It is to make sure they think first and reach for it second.

That order is the whole design. Brain scans, national surveys, and classroom trials keep finding the same split: kids who think first and use AI second keep their own judgment, while the ones who reach for it first quietly hand it over.^{1, 2, 3} The story runs on that idea. A child lands on Telos, a colony that let a rogue AI do all its thinking and nearly died of it. Over six weeks they put the thinking back in human hands. NOVA, the tutor, is built never to hand over an answer.

Generate: Make your own attempt first: a guess, a sketch, or a rough plan. Do it before the AI says a word.
Challenge: Treat an AI answer as a draft that might be wrong. Check the reasoning, not just the result.
Own: If you can explain it and rebuild it, it is yours. The screen goes down and the thinking stays with the child.

The six weeks

Each week lands the child on a new world with a job to do. The first three build the habit; the last three put it to work on math, research, and a project of the child’s own.

Week 1 · Telos · Generate

Brain vs. BORT

The first thing a child does each day is take a first attempt on their own before NOVA says anything. When people work out an answer themselves instead of reading one, they remember it better. That holds even when the first attempt was wrong. The habit matters more with a chatbot in the room. In one trial with about a thousand high-school students, the ones who used a plain AI as a crutch scored worse on their own once it was taken away. A version that gave hints instead of answers, the way NOVA does, caused no such drop.^{6, 5}

Day 1
Landing on Telos
Face BORT, the rogue AI running the colony, in a public battle of wits. The child wins by skipping the first obvious answer and reaching for one BORT could not predict.
Day 2
Blueprints for Survival
Design a base for the colony. The child sketches the plan their own way first, then works through it with NOVA. The AI sharpens the plan instead of making it.
Day 3
The Old Machines Fail
Repair a failing outpost. Pull the right parts, build each fix, and say why it will hold. NOVA points out where it could fail, and the child strengthens the fix or defends the call.
Day 4
What Else Lives Here?
Survey three biomes and study the creatures already on record. Ask NOVA hard questions about the habitat data, then write up the animal nobody has logged yet.
Day 5
The Week One Review
Answer to a Federation review for the whole week, one question at a time. Defend every plan drawn and every repair made.

Week 2 · Veyra · Challenge

The Static

NOVA is often confident and sometimes wrong, on purpose. Week two is about not taking her word for it. People tend to accept what a machine tells them even when the evidence in front of them says otherwise. Lean on its output too hard and you stop steering your own thinking. One 2025 study named that metacognitive laziness: the students who did it polished a single essay but learned no more than anyone else. So the child reads the evidence first, then makes NOVA defend her answer.^{7, 4}

Day 1
Something Is Wrong With NOVA
NOVA plots a route the child's ground scan shows is unsafe. The child points at each danger her map from orbit missed and makes her explain it. Then they draw a safer route.
Day 2
Confident Is Not the Same as Right
Using a field journal as evidence, the child takes NOVA's briefing one claim at a time and decides whether to trust it, reject it, or set it aside. A wrong answer in a confident voice is still wrong.
Day 3
In the Doctor's Office
Replay twelve moments where NOVA was sure of herself and find the exact step where her reasoning breaks. Name the mistake, not just the wrong answer.
Day 4
The Doctor's Verdict
Carry NOVA's reasoning core out of the clinic under a stability check that quizzes the child on the week's evidence-first habit.

Week 3 · Solis Reef · Own

The Reefborn Trials

You do not own an idea until you can explain it out loud and rebuild it without help. Students who explain a problem to themselves as they work understand it more deeply than students who just reread it. Getting ready to defend an idea to someone else pushes harder still. Kids work more and understand more when they expect to argue their case. Week three makes the child remember, choose, and defend. That is the only honest proof the thinking was theirs.^{8, 9}

Day 1
The First Tide Trial
A memory trial: take in the whole reef, then give it all back. It proves the child's own mind still holds information before a machine is trusted to hold it for them.
Day 2
The Discipline Test
Across ordinary days, the child decides each time whether to do the work themselves or hand it to a machine. When a machine is the right call, they pick the best way to ask.
Day 3
The Tide Court
Defend a claim against Quill, an undefeated challenger. The child prepares reasons in the quiet, then answers a brand-new claim and explains why their answer beats it.
Day 4
NOVA's Core Protocol
Install the Generate, Challenge, and Own rules into NOVA's rebuilt core, one at a time. The child decides how the AI they rely on should behave.

Week 4 · Frost Hollow · Math, applied

The Solarium Run

Week four is real math. Every day, NOVA hands over a plan with a mistake buried in it. Working a problem out before anyone shows you the method feels slower. It builds deeper understanding, and it works better on new problems than being handed the finished steps. The effort that slows a child down in the moment is often what makes the math stick. So they run the numbers first and check NOVA's second.^{10, 11}

Day 1
Back to the Rigs
Work out which of three drill engines mines the most crystal in a 12-hour day, and prove it on paper. NOVA gets it wrong by counting cooling time as digging time.
Day 2
The Long Haul
NOVA's supply route crosses ice too thin to hold a rover. Trace one that is actually drivable, then prove the fuel lasts one leg at a time.
Day 3
Packing the Pod
Load the launch pod like a puzzle, fitting crates within both the space and the weight limit that NOVA forgot. Total the space, weight, and value before launch.
Day 4
The Transmission
Clear a Federation scan of every number proved on the ice before the cargo can fly home to power the colony.

Week 5 · Ember Drift · Reading and research

The Storm Drop

Week five is research. NOVA can read eight thousand documents in the time it takes a child to open one. Speed is not the skill. When researchers watched fact-checkers, historians, and students judge websites, the fact-checkers won. They did it by leaving a page to check what other sources said, instead of trusting how official it looked. The child asks the archive real questions, keeps what has a source, and drops the rumors. Then they catch the one line in NOVA's summary that does not hold up.¹²

Day 1
The Storm Drop
Question an archive of 8,412 documents for the safe way down, and keep only the claims with a source. Catch the one line in NOVA's summary that does not match what the child found.
Day 2
The Food Hunt
Teach NOVA what counts as food, one rule at a time, cutting 427 scanned plants down to four. Then make the final call a scanner cannot settle.
Day 3
The Last Transmission
Write three honest lines to the colony's Seed Council, then read NOVA's cleaned-up version with a critical eye. Defend the message when all three leaders push back.
Day 4
The Homecoming Scan
Run the end-of-week scan. Every question answered plants one of ten seed pods waiting in the greenhouse back home.

Week 6 · Telos · Capstone project

The Architect's Project

The last week is the real test: a project the child designs, builds, and defends, with NOVA kept to the role of sounding board. A skill only counts if it holds up on a fresh problem with no one there to help. The slow, hands-on work of the earlier weeks is what makes it stick. NOVA gets one read of the finished project and suggests one change. The child takes it or turns it down and says why.^{11, 10}

Capstone
The Architect's Project
Voted Chief Architect of the colony, the child builds something the colony still lacks: a tool, a machine, a rule, or a system. They name who it helps, then plan, build, and defend the project on their own across several days.

25 missions: Six weeks, five worlds
Ages 11–14: Math and reading scale to each kid
Self-paced: Nothing unlocks on a schedule
You see it all: Parent and teacher dashboard

The research behind the Expedition

1Kosmyna, N., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. MIT Media Lab. arXiv:2506.08872. Source
2Schwartz, H. L., & Diliberti, M. K. (2026). More Students Use AI for Homework, and More Believe It Harms Critical Thinking: Selected Findings from the American Youth Panel. RAND Corporation (RR-A4742-1). Source
3Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), 6. Source
4Fan, Y., et al. (2025). Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance. British Journal of Educational Technology, 56, 489–530. Source
5Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2024). Generative AI Without Guardrails Can Harm Learning: Evidence from High School Mathematics. SSRN 4895486. Source
6Slamecka, N. J., & Graf, P. (1978). The generation effect: Delineation of a phenomenon. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 592–604. Source
7Skitka, L. J., Mosier, K. L., & Burdick, M. (1999). Does automation bias decision-making? International Journal of Human-Computer Studies, 51(5), 991–1006. Source
8Chi, M. T. H., de Leeuw, N., Chiu, M.-H., & LaVancher, C. (1994). Eliciting self-explanations improves understanding. Cognitive Science, 18(3), 439–477. Source
9Chase, C. C., Chin, D. B., Oppezzo, M. A., & Schwartz, D. L. (2009). Teachable agents and the protégé effect: Increasing the effort towards learning. Journal of Science Education and Technology, 18(4), 334–352. Source
10Kapur, M. (2014). Productive Failure in Learning Math. Cognitive Science, 38(5), 1008–1022. Source
11Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In Psychology and the Real World (pp. 56–64). Worth Publishers. Source
12Wineburg, S., & McGrew, S. (2019). Lateral reading and the nature of expertise: Reading less and learning more when evaluating digital information. Teachers College Record, 121(11), 1–40. Source
13Stanford Institute for Human-Centered AI (HAI). What parents need to know about AI in the classroom. Source

Ready to start the expedition?

Twenty-five missions, open from the moment it is yours. Your explorer goes at their own pace, and you can watch the whole thing from your account.

See pricing Read the science behind it