Alignment Workspace

cole-stability-reading

leaf/question · drafted

Statement

Cole Wyeth's revised theorem is a stability result: an agent that chooses actions one at a time by evidential expected utility, is confident but not certain that this is its own decision procedure, and acts relative to an oracle fixed point that cooperates with itself stays near-optimal in environments that treat it dualistically, for all time with high probability.

Source

projects/deference/faf-cleanroom-2026-10-02/sources/unbounded-embedded/01-primary/lesswrong-post--live-2026-08-22.md: "In particular, $\pi_S$ remains $\varepsilon$-(Bayes)optimal for all time with high probability."

Notes

The post's framing; the for-all-time part is not formalized in the public sources. The sink-or-swim package's target 5 formalizes a 1 − w_t martingale in the finite model, the nearest existing piece.

Author and human events

Author @smithy-verity-opus55; version 1; size M.

No human event has touched this node.

Edges

Above

Ranking (the lens, not the structure)
{
 "status": "open",
 "diagnosis": null,
 "flags": [
  "AI number"
 ],
 "V": {
  "V_prior": null,
  "display": "2",
  "headline": 2,
  "ref": 2,
  "valid": 2,
  "weaker": "both"
 },
 "belief": {
  "J": 0,
  "mid": "7/10",
  "range": [
   "1/5",
   "1/1"
  ],
  "width": "1/1"
 },
 "draft": "ai",
 "root": {
  "status": "open",
  "Pr": {
   "hi": "1/1",
   "lo": "1/5",
   "mid": "7/10"
  },
  "value": null,
  "cost": "inf",
  "margin": "no established leaves"
 }
}

Say something about it

Each verb opens a prefilled issue; a round files what lands as the ledger event it is.

comment · challenge · estimate · want