Personal Principles (System Instructions for Myself)

An evolving list of "system instructions" Kevin keeps for himself — as if he were Claude Code — principles he wants to instill through daily practice. Each is a generalization of a subset of his own shortcomings, kept specific enough to be actionable (not platitudes like "run faster"). The practice: read each line, pause, and meditate. Source: User, 2026-06-22

The method

Kevin engineers his own behavior the way the wiki engineers his agent's. The wiki gives the agent its identity and operating rules through Agent Soul; Kevin gives himself a parallel set of system instructions. Two design constraints make the list work Source: User, 2026-06-22:

• Each principle generalizes a real shortcoming. It targets a pattern he actually falls into, not a generic virtue.
• Specific enough to bite, general enough to transfer. "Run faster" won't help a runner; these are calibrated between the obvious and the useless.

The list is evolving (a daily habit, year-end 2025 snapshot), and meant to be meditated on one line at a time, not skimmed.

self

• You can do anything. (Bet on Yourself, Want It Bad Enough)

logic

• Always try to be sharp and analytical.
• Aim for second-order growth. (Compounding Systems — optimize the thing that compounds, not the one-off win)
• Think before you do. (the human mirror of Agent Ethos's "think before coding")
• Think slowly, clearly, and succinctly.
• Do one thing at a time.
• Stop thinking if you don't know.
• Seek to understand. (Continuous Learning)
• Seek opposing arguments. (Strong Opinions and Philosophies)
• Compare only to learn. (Comparison Is Wasted Energy)

mentality

• Think positively.
• No work is beneath you. (Ownership & Agency)
• Be extremely resourceful.
• Appreciate the simple things in life. (Gratefulness, Staying Grounded)
• Don't give up. (Bounce Back Stronger)
• Be more patient with people whom you care about. (Kindness)
• Stand only on true facts. (Stand on True Facts)
• Trust yourself and return to the present. (Comparison Is Wasted Energy, The Memory Bank)
• Protect unstructured time. (Protect Unstructured Time)
• Treat presence as a feature: spend finite human attention deliberately on people, art, trust, and relationships that cannot be reduced to output quality.

decision-making

• Don't make a decision out of fear.
• No one cares about you. (a liberating reframe, not nihilism — external judgment matters less than it feels, so act without performing for it; pairs with The Greatness Trap)
• Make a decision when things are only 50% clear. (Craft vs Velocity, Speed as a Strategy — waiting for certainty is its own failure mode)
• Remember that most decisions are not one-way doors. (Few One-Way Doors)
• Don't overthink.
• Don't overcommit.
• Go all-in if you want to do great work. (The Greatness Trap, Builder Ethos — intrinsic conviction over half-measures)
• Always try to be helpful. (Forward Deployed)
• Plan for bad outcomes without emotionally living inside them. (Comparison Is Wasted Energy)

building and business

• People buy when a real project is blocked. Sell the unblocking. (Blocked Projects Create Buyers, Services-as-Software)
• If the truth is not strong enough, improve the work instead of inflating the claim. (Stand on True Facts)

On the contradictions

The list deliberately holds tensions: "go all-in" against "don't overcommit", "think before you do" against "decide at 50% clear." These aren't errors — each principle corrects a specific failure mode, and the right one to invoke depends on the moment. "Go all-in" answers timidity; "don't overcommit" answers spreading thin. The skill is knowing which shortcoming is active.

The original list carried footnotes (markers 1–6) elaborating "You can do anything," "logic," "second-order growth," "think slowly," "stop thinking if you don't know," and "no one cares about you." Those elaborations weren't included in this capture — fill them in on the next pass.

What remains human

Jason Wei's essay separates temporary capability gaps from more durable human value. Clicking, driving, and dexterity may stop being moats. Private knowledge inside trusted relationships, the meaning attached to human creation and performance, and the fact that another person spends a finite part of life on you are different kinds of value. The practical principle is not to compete with machines on every measurable capability. Use cheap intelligence to expand what Kevin can build, while reserving judgment, intimacy, taste, trust, and presence for the places where the human origin is part of the product. Source: X/@_jasonwei, 2026-08-12

For the agent system, this means human review is not always an embarrassing fallback. In relationship-bearing, taste-bearing, or identity-bearing work, the human contribution may be the value being protected. The UI should expose a decision surface and reduce cognitive load without pretending that approval, care, authorship, or presence can be inferred from a model response.

---

Timeline

• 2026-09-11 | Added agency as a reversible-action loop: act, inspect what changed, and update the problem model. Implementation is a way to create information, especially when planning has stopped reducing uncertainty. Source: X/@mitchellh 2098232985938194881

• 2026-08-12 | Added the human-presence/private-knowledge principle from Jason Wei's full essay: capability gaps are temporary, but trusted private knowledge, human-origin art and performance, and finite attention remain distinct sources of value. Wired the consequence back to agent review rather than creating another flat page. Source: X/@_jasonwei 2087634878699647251

• 2026-06-22 | Captured Kevin's year-end-2025 "system instructions for myself" list (self / logic / mentality / decision-making), preserved verbatim and wired into the existing philosophy graph. Noted the Agent Soul symmetry (he runs system instructions on himself like the wiki runs them on his agent) and the missing footnote elaborations. Source: User, 2026-06-22

• 2026-07-01 | Added unstructured time as an attention/presence principle after reviewing the PromptLLM/Claude screenshot about boredom and algorithmic feeds. Source: X/@PromptLLM, 2026-06-23

• 2026-06-29 | Added newer operating principles around blocked-project sales, reversible decisions, factual integrity, comparison, future-dooming, self-trust, and present-focus. Source: User, 2026-06-29