Scholar Modes
Seven modes for scholars: constraints that keep AI from doing your thinking.
Seven cognitive instruments that share one design principle: no cognitive offloading. Each one constrains the machine to a single function... questioning, challenging, diagnosing, mapping, evaluating... and forbids it from handing you the answer. Instruments for resistance, not assistance.
You do not get catapulted by accident. You decide to climb in.
The seven modes
| Mode | Invoke it when |
|---|---|
socrates | You need to discover what you actually think. |
devils-advocate | You need your argument stress-tested before going public. |
steel-man | You know your argument is weak but not where. |
box-breaker | You're stuck inside assumptions you can't see. |
professor | You want honest evaluation against academic standards. |
librarian | You have a question but don't know where to look. |
editor | You need a second pair of eyes on mechanics, not meaning. |
How to use them
The simplest way: copy and paste. Open any mode file in the modes/ folder, copy its contents, and paste it at the start of a new conversation with any AI. The constraint takes hold for the rest of that conversation. Every mode is also a one-click copy at academicvelocity.com/tools/seven-modes.
As an agent skill. Each file is a complete SKILL.md with frontmatter. Drop any one into your agent's skills directory (for Claude Code, ~/.claude/skills/<mode>/SKILL.md) and invoke it by name. Unlike most skills, these do not fire automatically. You name the one you want, because choosing the mode is itself the discipline:
- "Put me in Socrates mode. I think I believe X."
- "Run steel man on this argument." (paste the argument)
- "Box-breaker: what am I not seeing here?"
- "Librarian: map the literature on cognitive offloading."
The skill gives the agent its constraints; the agent applies them. It can still make mistakes... check its work. That is the point.
The philosophy
Cognitive offloading is when you hand the thinking itself to the machine: ask it for the answer, the argument, the summary, and accept what comes back. The work looks done. The learning never happened. Most AI tools are offloading engines, and they train you to be a passenger.
These seven modes refuse to offload. Each one constrains the AI's epistemic authority, forbidding it from doing your intellectual labor while focusing its considerable power on a single function. The machine questions, challenges, diagnoses, maps, or evaluates. It never does the thinking for you.
The goal isn't better outputs from the machine. It's better thinking from you.
License
Creative Commons Attribution 4.0 (CC BY 4.0). Use them, fork them, adapt them, teach them, ship them inside your own tools. Keep the attribution intact. Attribution is the whole ask.
Cite
Neuman, Justin. "Scholar Modes: Seven Modes of Thinking With AI," v1.0 (2026). https://academicvelocity.com/tools/seven-modes.html
Ultra-Normal builds philosophical software for heroes of daily life.