This page explains how I try to think clearly and how this site reflects that effort. It’s part manifesto, part user manual.
Why This Page Exists
Most blogs present conclusions. Few show their epistemic work: how confident the author actually is, what would change their mind, or how their thinking has evolved.
I think that’s a failure mode. If you’re going to publish ideas, you should also publish your uncertainty. This site tries to do that through structure, not just words.
The Metadata System
Every page on this site carries three metadata fields, inspired by Gwern Branwen’s approach:
Status
How complete is this piece?
| Status | Meaning |
|---|---|
| 💡 Idea | A seed. Barely formed. |
| 📝 Notes | Rough notes, thinking out loud. |
| ✏️ Draft | Has structure but needs work. |
| 🔄 In Progress | Actively being developed. |
| ✅ Finished | Complete to my current satisfaction. May still be updated. |
| 🚀 Launched | For projects/ventures that are live. |
| 🔨 Building | Actively under construction. |
| ⚡ Active | Ongoing, no “finished” state (like this page). |
Certainty
How confident am I in what I’ve written?
| Certainty | Meaning |
|---|---|
| ❌ Impossible | I’m confident this is wrong (keeping it for the record). |
| 🤔 Unlikely | Probably wrong, but worth considering. |
| ❓ Possible | Could go either way. Genuinely uncertain. |
| 👍 Likely | I’d bet on this, but I’m not sure. |
| ✓ Certain | As confident as I get. (Which still isn’t 100%.) |
Importance
How much does this matter? A 1-10 scale:
- 1-3: Minor thoughts, low stakes
- 4-6: Interesting but not crucial
- 7-8: Important ideas I’ve thought carefully about
- 9-10: Core beliefs or foundational content
The metadata isn’t decoration. It’s a contract with the reader: here’s how seriously you should take this, and here’s how seriously I take it myself.
How I Form Beliefs
I don’t have a formal epistemology degree. What I have is a set of practices I try to follow:
1. Start with uncertainty
My default state is “I don’t know.” Confidence is something I build toward, not something I start with. If I can’t articulate why I believe something, I probably shouldn’t believe it yet.
2. Seek disconfirmation
It’s easy to find evidence for what you already believe. The useful work is looking for evidence against your position. I try to steelman opposing views before critiquing them (see Rule #8).
3. Express beliefs as probabilities
“I think X” is less useful than “I’m 70% confident X is true.” Probabilities force precision. They also make updating easier: if new evidence drops me from 70% to 40%, that’s a meaningful update I can track. (See Rule #11.)
4. Update on evidence
When evidence conflicts with my beliefs, the beliefs should move. Not always immediately (sometimes evidence is weak or misleading), but the direction of update should follow the evidence, not my preferences. I track major updates in my Changelog.
5. Read primary sources
Summaries lose nuance. I try to read the actual papers, books, and documents rather than relying on second-hand accounts. This is slower but produces better-calibrated beliefs.
6. Track what I believe and why
My Beliefs page is a public record of my current positions. My Changelog tracks how they’ve evolved. My Predictions page tests whether my confidence is well-calibrated.
Influences
The epistemic approach on this site is shaped by:
- Gwern Branwen: The gold standard for epistemic metadata, living documents, and intellectual rigor. His approach to status and confidence tags directly inspired this site’s metadata system.
- Philip Tetlock: Superforecasting convinced me that calibration is a learnable skill, and that tracking predictions is the best way to improve judgment.
- Daniel Kahneman: Thinking, Fast and Slow is the manual for understanding your own cognitive failures.
- LessWrong: The rationalist community’s emphasis on Bayesian reasoning, calibration, and epistemic humility has shaped how I think about thinking.
- Buster Benson: His belief codex inspired my Beliefs page.
- The Stoics: Marcus Aurelius and Seneca weren’t Bayesians, but their emphasis on distinguishing what you can and can’t know is proto-epistemic practice.
What This Means in Practice
Concretely, this site tries to:
- Show uncertainty: Every page has a certainty rating. I don’t pretend to know things I don’t.
- Evolve over time: Pages are updated, not just published. The Changelog tracks major revisions.
- Make predictions testable: My Predictions page contains specific, dated claims with probabilities that can be checked.
- Be transparent about reasoning: When I change my mind, I say so and explain why.
- Separate confidence from importance: A post can be important (high importance) but uncertain (low certainty). These are orthogonal.
Limitations
I should be honest about where I fall short:
- I’m not perfectly calibrated. Nobody is. I’m working on it.
- I don’t always update as fast as I should. Motivated reasoning is the default human mode. Knowing about it helps, but doesn’t cure it.
- This system is new. I’ve been thinking this way for years, but formalizing it publicly is recent. The prediction track record will take time to build.
- I’m an engineer, not an epistemologist. My approach is practical, not academic. If you want formal epistemology, read the philosophers. If you want to see someone trying to think clearly in public, this is what that looks like.
Last updated: February 2026
If you have suggestions for improving the epistemic practices on this site, I’d love to hear them. That’s kind of the point.
Related
See also: What I Believe | My Predictions | How I’ve Changed | Questions I’m Exploring | My Rules