new file mode 100644
index 0000000..23be4c7
@@ -0,0 +1,47 @@
+---
+visibility: public-edit
+---
+
+# information gathering
+
+how much research should you do before making a decision? the answer is almost never "all of it."
+
+## the diminishing returns curve
+
+information gathering follows a steep diminishing returns curve. the first 30 minutes of research on a topic often gives you 70% of the useful information. the next 3 hours might give you another 20%. the remaining 10% could take days.
+
+the question isn't "do i have enough information?" — you never will. the question is "will more information change my decision?"
+
+## the 70% rule
+
+i aim to make decisions at roughly 70% information. that's enough to be directionally right without burning time on diminishing returns. for [[reversible-vs-irreversible|type 2 decisions]], even 50% might be fine — just try it and course-correct.
+
+for type 1 decisions, push closer to 85-90%. but never 100%. 100% is a fantasy, and waiting for it is its own form of risk.
+
+## failure modes
+
+- **research as procrastination** — the most common one. "i need to learn more" becomes a way to avoid the scary part: committing. if you've been researching for a while and your decision hasn't changed, you're procrastinating.
+- **anchoring on the first thing you find** — the opposite failure. making up your mind immediately and then only seeking confirming information. see [[research-workflow]] for how to avoid this.
+- **asking too many people** — every new opinion adds noise. after 3-4 informed perspectives, more opinions usually just create confusion.
+- **not distinguishing signal from noise** — some information sources are much more valuable than others. one conversation with someone who's done the thing is worth 20 blog posts about the thing.
+
+## the best information sources
+
+ranked by signal density:
+
+1. **doing the thing** — nothing beats direct experience. often faster than research. if you can prototype or test cheaply, do that instead of researching. see [[vibe-coding]] for how this applies to software.
+2. **talking to someone who's done it** — high bandwidth, can ask follow-up questions, they know what actually mattered vs what seemed important
+3. **case studies / postmortems** — real outcomes, not theory
+4. **structured frameworks** — models that help you think, not answers that tell you what to do
+5. **general advice** — often too generic to be actionable. "it depends" is usually the honest answer.
+
+## when to stop gathering
+
+signs you have enough information:
+
+- your decision hasn't changed with the last few inputs
+- you can articulate the key tradeoffs clearly
+- you know what you don't know, and you've decided it's acceptable not to know it
+- the cost of waiting is starting to exceed the cost of being slightly wrong
+
+this connects to [[operation-optimization]] — information gathering is itself a process that can be optimized. don't just research the decision; optimize how you research.
\ No newline at end of file