Treat Yourself Like an Agent
One of the best productivity ideas Kevin has found: treat yourself like an agent. You have a system prompt — your subconscious and identity, running under every response — and you don't get to edit it directly. The unlock is realizing these variables aren't fixed; they can be fine-tuned for yourself the way you'd tune an agent. private page
The system prompt you can't open
Every agent runs on a system prompt it doesn't control. For a human, that prompt is the subconscious + identity running underneath every "response" you produce. You can't open the file and edit it directly. private page
What does edit it is the "woo stuff" — sleep, food, exercise, meditation. These are not self-care for its own sake; they are the only write-access you have to the file you can't open. This reframes Kevin's health philosophies — Sleep, Nutrition, Strength Training, Recovery, Spirituality, Biohacking, Low-Cortisol Love, the The Five Pillars — as system-prompt maintenance: performance inputs, not indulgences. Neglect them and the prompt degrades.
A degraded prompt can't be fixed by awareness
Knowing you're in a bad mood or a bad headspace doesn't fix the mood. "The system prompt is degraded and your conscious mind inherits it." Willpower and self-talk run on top of the prompt — they can't overwrite it. The implication: you manage the prompt upstream (state, recovery, inputs), not by trying to consciously override a degraded run. private page
Identity engineering
You can deliberately change the prompt's contents. Kevin spends real time managing his: reading biographies of successful founders and loading their mental models into his head so he can access them whenever a decision comes up — actively changing his identity to influence his behavior. Identity is an input you can edit, not a fixed given. (World Models, Want It Bad Enough) private page
Context engineering for humans
The same Context Engineering discipline used on agents applies to yourself. Humans have very short context windows that degrade with too much information. So before a task or a day of work, Kevin actively decides what context to load into his own window — and what to leave out. Curating what you let in is as important as what you add. private page
The catch: you're not actually an agent
The metaphor has a hard limit. Agents have consistent performance regardless of time; humans are extremely inconsistent. The same task or prompt given to yourself produces different outputs at 9am vs 9pm. You can only apply "high effort" to a task when you're in the right headspace. So scheduling, state, and timing aren't logistics — they gate which work is even possible. The metaphor tells you what to tune; the inconsistency tells you when you can. private page
The unlock
Agentic engineering and optimizing humans is actually not that different. The biggest unlock is realizing that all of these variables are not fixed and can be fine-tuned for yourself as well. private page
This is the meta-frame behind Personal Principles (System Instructions for Myself): that list is Kevin's literal "system instructions for myself," and this thesis is why maintaining them works. It mirrors the wiki's own architecture — the agent's identity and rules live in Agent Soul; Kevin treats his own subconscious as the same kind of editable-but-not-directly file. The same skill set that makes agents perform — Agent Ethos, context engineering, identity/harness design — is the skill set for making yourself perform.
Timeline
- 2026-06-22 | Captured Kevin's "treat yourself like an agent" thesis as a follow-up to Personal Principles (System Instructions for Myself). Components: the uneditable system prompt (subconscious + identity), the woo stuff as its only write-access, awareness ≠ override, identity engineering via founder bios, context engineering for humans, and the inconsistency catch (humans aren't time-invariant). Reframes the health philosophies as system-prompt maintenance. private page