Which Mental Models Can You Train on Too Hard Pile?

Updated

Too Hard Pile trains 13 mental models across 1,300 practice drills, and the model explanations themselves are free to read. The five models named explicitly on the site are inversion, first principles, incentives, base rates, and margin of safety — with second-order effects also called out as a selectable model in the drill flow. If you want to know whether the specific tool you care about is covered, the free model library is the fastest way to check before creating an account.

The models named on the site

The homepage and drill flow name these models directly:

Model Where it appears
Inversion Listed in the "13 models" line and as a drill option
First principles Listed in the "13 models" line
Incentives Listed in the "13 models" line and as a drill option
Base rates Listed in the "13 models" line
Margin of safety Listed in the "13 models" line
Second-order effects Named as a pickable model in step 1 of "How it works"

That's six models with direct textual support. The site states the total is 13, so the remaining seven aren't enumerated in the material available — the free model explanations are where the full list lives.

How each model is trained

The training loop is the same regardless of which model you pick, and it's designed so the model fires without a prompt eventually:

  1. Pick a model. You choose the one that's "messing with your decisions this week" — inversion, incentives, second-order effects, or another from the 13.
  2. Drill it. You get a short, real scenario. You answer before the drill shows you anything. The site's stated rationale: "The struggle is the part that teaches."
  3. Watch it go automatic. You rate yourself, the drill comes back on a schedule, and the goal is that one day you don't need the prompt anymore.

The key mechanic is commit-before-you-peek. Every drill makes you answer first, so you can't skim a paragraph and move on. That's what separates this from reading a list of biases.

What "1,300 drills across 13 models" actually means

Roughly 100 drills per model if evenly distributed — but the site doesn't break down the count per model, so treat that as an estimate, not a spec. What's confirmed: 13 models, 1,300 drills, 5 minutes a day as the suggested pace.

The free tier gives you 6 drills, 5 journal entries, and the whole Toolbox, with no card and no time limit. So you can sample several different models before deciding whether the full set is worth it.

The journal is the part that checks whether the model worked

Drills train the tools; the journal tells you if they're working. You write down the call you're about to make — the decision, what you predict happens, how sure you are — and weeks later it comes back and asks whether you were right. The site's own framing: "Almost nobody does the second part — this makes you."

This matters for model selection because a model you feel you've learned isn't the same as one that improved your calls. The journal is the only part that keeps score.

How to decide if your model is covered

  • Check the free model explanations first. They're free, and they're the authoritative list of all 13.
  • Use the 6 free drills to test the training feel. Try a drill with no signup, then decide if the commit-first format works for you.
  • If your model isn't in the six named above, don't assume it's missing — the site only names a subset in the material available. Verify against the free library before concluding.