This week: neuroscientist Read Montague walked Huberman through the dopamine prediction-error system that runs every decision you make, Ido Portal reappeared in the Essentials vault with a movement philosophy that treats the body as a skill, and a Nature Medicine study on GLP-1 drugs landed a cognitive side-effect finding no one expected.

GM.

The episode most worth your time this week was Read Montague on Huberman. Montague studies dopamine computationally — not what it feels like, but what it calculates. The distinction matters more than most health podcasts acknowledge. Your reward system isn't a feeling. It's a running account of how reality compared to expectation, updated thousands of times a day.

⏱ ~5 min read

Two ideas surfaced in the same week. One: the dopamine system doesn't track pleasure — it tracks prediction errors, the gap between what you expected and what you got. Two: the body isn't a machine you maintain. It's a skill you practice. One came from computational neuroscience. The other from a man who treats a handstand the way a musician treats a scale.

The thread this week: the systems that run you are learning systems — and most people have never looked at the curriculum.

80%
The estimated share of daily decisions driven by dopamine prediction-error signalling rather than conscious deliberation.

Read Montague walked Huberman through the computational neuroscience behind dopamine — not the pop-science version where dopamine equals pleasure, but the engineering version where dopamine encodes prediction errors. When you get something better than expected, dopamine spikes. When you get exactly what you expected, it stays flat. When reality disappoints, it dips. The system isn't rewarding you. It's teaching you.

Montague's lab has mapped this signal across decision tasks, addiction models, and social exchanges. The core finding: the brain doesn't optimise for happiness. It optimises for prediction accuracy. The dopamine spike isn't the reward — it's the error signal that updates the model. Over thousands of repetitions, your preferences, habits, and cravings are all downstream of this update loop.

The serotonin system, Montague argued, operates on a different timescale — tracking slower, mood-level baselines that modulate how aggressively the dopamine system updates. When serotonin is low, prediction errors hit harder. When it's stable, the system learns more evenly.

The implication for anyone trying to change a habit: you're not fighting willpower. You're fighting a prediction model that's been training on your behaviour for decades.

Dr. Read Montague on Huberman Lab · February 2, 2026

The dopamine system doesn't care what you want — it cares what you predicted
Read Montague on computational models of dopamine, serotonin, and why your brain treats every outcome as a data point.
Dr. Read Montague · Huberman Lab · February 2, 2026

Montague is a computational neuroscientist at Virginia Tech, not a wellness figure, and it shows. The conversation is dense but precise. Huberman lets him run, and the result is one of the clearest explanations of reward-prediction error you'll find outside a textbook.

🎙️ Listen → Huberman Lab, Dr. Read Montague, February 2, 2026

Montague gives you the mechanism. The week gives you the question: what is your prediction model trained on?

The pop-science version of dopamine is a motivation molecule — hack it with cold plunges, fasting, and dopamine detoxes. Montague's version is stranger and more useful. Dopamine is a teaching signal. Every craving, every habit, every preference you have is the downstream output of a model that's been updating itself since birth.

The practical consequence: if you want to change behaviour, changing the reward doesn't work well. What works is changing the prediction. Montague's lab has shown that reframing expectations before an experience can shift the dopamine response to that experience. Cognitive behavioural therapy, he suggests, works partly because it rewrites the predictions the brain is testing against reality.

Ido Portal's Essentials episode, which Huberman pulled from the vault the same week, lands on the same idea from a different angle. Portal doesn't talk about dopamine. He talks about practice. But his core argument — that the body is a skill you develop, not a vehicle you maintain — maps onto the prediction-error framework surprisingly well. A movement practice that constantly introduces new challenges keeps the prediction model updating. A routine that never changes lets it calcify.

The GLP-1 finding from Nature Medicine adds a pharmacological wrinkle. Semaglutide appears to reduce not just appetite but compulsive behaviours across multiple domains — food, alcohol, shopping. If the dopamine prediction-error model is correct, GLP-1 drugs may be dampening the error signal itself, not just the hunger signal. The mechanism is still speculative, but the pattern is consistent.

The question this week isn't whether your dopamine system is working. It's what curriculum you've been feeding it — and whether that curriculum is still serving you.

P.S. If you could see your dopamine prediction-error log for the past week — every spike, every dip — what would surprise you most? Hit reply and tell us. Best responses go in a future issue.

What did you think of today's BrainBite?