Oracle Bones

Stub entry — May 3, 2026. To be developed.

In one sentence

An Oracle Bone is a dated, falsifiable, written-down prediction filed before the event resolves, scored after it does, and re-read months later — making prediction accountable in the only way it can be: across time.

Where the name comes from

Shang dynasty divination, c. 1600–1046 BCE. Ox scapulae and turtle plastrons were heated until they cracked; the cracks were read by diviners; the question, the diviner’s reading, and often the outcome were inscribed on the bone — and kept.

The keeping is the discipline. Anyone can predict. Few file the prediction with the date and the reasoning, and almost nobody returns to grade it.

The Shang oracle bones survive 3,000 years later because the inscription itself was the practice. The Oracle Court — the small advisory council of five scholars we run inside our learning system to commission and grade predictions — borrows the name and the discipline directly. Its five characters each hold a single verb: 老司天 the Astronomer (observe), 占者 the Diviner (read), 史 the Scribe (record), 史官 the Historian (remember), and 滑稽 the Jester (jest). The Diviner writes the bone; the Jester argues the opposite before it is filed; the Historian scores it months later. The Scribe keeps the register.

Why the practice exists

Three failures repeat across every prediction tradition without filed predictions:

  1. Hindsight rewrites memory. Without a written record, the predictor remembers their accurate calls and forgets their wrong ones, drifting toward an unfounded confidence that erodes future calibration.
  2. The framework cannot improve. A framework that produces predictions but never grades them has no feedback loop. It will reproduce its own errors indefinitely.
  3. Grey Swans hide. Without a record, you cannot tell after the fact whether an event was a true black swan or a Grey Swan — the prediction was either filed or it wasn’t.

The Oracle Bone is the smallest unit of accountability that solves all three.

What an Oracle Bone actually contains

Each bone is a single markdown file in learning-memos/oracle-bones/YYYY-MM-DD-<institution>-<short-name>.md, with a fixed structure:

  1. The institution. What entity is the bone reading? (Apple, OpenAI, AACSB, your own household, etc.)
  2. The window. What date range is the prediction for? (Specific. “Within 90 days” is the minimum; an explicit deadline is better.)
  3. The convergence reading. Which of the six vectors are lit up, and on what evidence? Which are dark to us, and why?
  4. The prediction. A specific, falsifiable claim. “Cook will announce a successor by July 1, 2026” is a bone. “Apple will face challenges this year” is not.
  5. The diviner. Which member of the Court (or which agent) read the bone? Whose call is this?
  6. The reasoning. Two paragraphs. Enough to reconstruct the call later.
  7. The scoring slot. Empty until the window closes. Then filled in: fulfilled / partially fulfilled / refuted / untestable, with a brief note on what we got wrong about the reasoning, even when the prediction was right.

The file is committed to the repo at the moment of filing. The git timestamp is the bone’s authentication; it cannot be backdated without leaving traces.

A working example

As of May 3, 2026, no bones have been filed yet. The first will be commissioned this week, almost certainly on a convergence-vulnerable institution that emerged from the May 3 Apple learning memo (candidates: Anthropic, John Deere, AACSB).

The empty bones folder — learning-memos/oracle-bones/ — is itself a kind of accountability. It documents the gap between writing the framework and using it.

The ritual

Three timing layers, deliberately staggered:

The annual reading is where framework versioning happens. Convergence Detection v0.1 → v0.2 will be driven by what the first year’s bones reveal, not by armchair revision.

Why it matters in a teaching context

Strategy and forecasting classes routinely teach students to make predictions. Almost none teach them to file and grade predictions. The Oracle Bone practice — even in its lightest form, three predictions per semester filed in a shared notebook with a scoring meeting in finals week — closes the loop that the rest of the curriculum leaves open.

It also serves a deeper pedagogical purpose: it teaches students that being wrong, when filed honestly, is more valuable than being right by luck. The student whose three predictions were all wrong but who learned why is the one developing a usable framework. The student whose three predictions were vaguely correct in a hand-wavy way has learned nothing.

Trade-offs and warnings

See also

Status: stub, May 3, 2026. The first bones will be filed this week; this entry will be expanded with a real worked example once we have one.

Return to Dictionary All Entries (A–Z)