# Educational social library
Draft concepts and opening copy. Pair every market mockup with a visible DEMO DATA label. Do not imply current observations, validated outcomes, or available features that remain proposed.

## LinkedIn: 20 post concepts
1. **The horizon behind the chart.** “A short-term rise can sit inside a longer decline. The first question is which observation window your conclusion describes.” Explain lookback versus holding period. Link the horizon article.
2. **Uncertainty as output.** “When evidence is missing, a label can become misleading. An uncertain state tells the researcher something the average score can hide.” Show a missing-input callout.
3. **The limits of a positive trend.** “Positive momentum is an observation about a selected window. Persistence is a separate claim.” Explain the qualification boundary.
4. **Three clocks.** Separate event time, data receipt, and first-known time with a simple timeline. Ask which timestamp a backtest uses.
5. **Frozen training.** Describe why an evaluation should consume retained thresholds instead of silently refitting on its own history.
6. **A percentile is relative.** Compare two synthetic training distributions with the same percentile but different absolute volatility.
7. **A spread is a snapshot.** Explain displayed bid–ask distance, visible depth, and why neither guarantees a fill.
8. **When estimators disagree.** “A disagreement between R/S and DFA is evidence to inspect, not a contest to select the more appealing result.” Link Hurst education.
9. **The story of a missing bar.** Walk through how one missing observation changes a complete-window diagnostic into an unavailable result.
10. **A proposed score needs a contract.** Introduce FMSI as a research concept. Discuss absent formulas and the need for qualification before a composite number.
11. **Price volatility versus structural context.** Explain why a HighVolatilityStress flag does not establish correlation or liquidity stress.
12. **Temporal honesty in research.** Compare retrospective use of corrected data with what a researcher could actually have known at a decision cutoff.
13. **Daily does not mean universal.** Explain why crypto calendar conventions cannot silently carry over to exchange-traded securities.
14. **Cross-venue agreement has limits.** Explain receipt-bounded quote checks without claiming synchronized source events or independent upstream ownership.
15. **Different participants, different priorities.** Introduce FMH through a day trader and long-horizon investor reacting to the same observation.
16. **Asset risk is not portfolio risk.** Discuss exposure, cash constraints, correlation, and execution assumptions as additional evidence requirements.
17. **Backtest denominators.** Explain why missed samples and failed captures belong in the result rather than disappear from the chart.
18. **Research before labels.** Show current uncertainty/stress states alongside proposed directional states. Explain why guarded code is not qualified runtime output.
19. **An understandable API.** Demonstrate null plus a reason, exact timestamps, and preserved decimal strings. Label the public contract as proposed.
20. **The questions shaping our beta.** Ask which missing context matters most to researchers. Invite interest without announcing a launch date.

## X: 20 post concepts
1. A timeframe is a choice. Before interpreting a chart, ask what its window includes—and what it leaves out.
2. Positive momentum and persistence are different claims. The first describes a window. The second needs qualified evidence.
3. “Unavailable” is a useful value when the alternative is an invented number.
4. A market date does not tell you when the information became available. Backtests need both.
5. A frozen model should stay frozen during evaluation. Otherwise the test quietly changes.
6. The 90th percentile is relative to a sample. It is not a 90% probability of a market event.
7. Displayed depth can change before an order arrives. A book snapshot is not a fill guarantee.
8. R/S and DFA can disagree. Keep the disagreement in the research record.
9. One missing bar can invalidate a complete-window calculation. Filling it changes the assumptions.
10. FMSI is a proposed Fractal Intelligence concept. No production formula or calibrated composite score exists yet.
11. Volatility stress describes price-history behavior. It does not establish observed liquidity stress.
12. Yesterday’s data, corrected today, is not necessarily yesterday’s available information.
13. Crypto calendar assumptions do not automatically apply to equities.
14. Matching quotes do not prove matching source timestamps.
15. The same price can matter differently to a day trader and a long-horizon investor.
16. An asset diagnostic is only one input to portfolio research.
17. Count the missed samples. A denominator is part of the evidence.
18. A state machine can contain a label that the current runtime is not qualified to produce.
19. A future API should preserve why a value is absent. Zero and unavailable are different meanings.
20. What market context is hardest to inspect in your current workflow? Fractal Intelligence’s early-access interest list is open. Beta timing is unannounced.

## Educational chart/post briefs: 10
1. **One path, three windows.** Synthetic normalized prices with 7-, 30-, and 180-day highlighted windows. Caption: “Changing the window changes the evidence.” No performance implication.
2. **Three distinct time concepts.** Diagram source-bar interval, calculation lookback, and intended holding period. Use actual Tactical/Swing/Structural definitions.
3. **Same percentile, different scale.** Two invented volatility distributions. Mark the 90th percentile in each. Caption explains reference dependence.
4. **Unavailable versus zero.** Two metric tiles, one null with a missing-history reason and one actual zero. Label both synthetic; explain the semantic difference.
5. **Observed, received, known.** An illustrative timing diagram of a market event, data receipt, and source revision. Do not invent historical events.
6. **Regime eligibility.** Show how missing qualified persistence blocks a directional label despite positive momentum. Use runtime v1 terminology.
7. **Displayed versus executable.** Annotated synthetic order book with retained depth boundaries and no extrapolation beyond visible levels.
8. **Frozen-fit separation.** Training window and later evaluation segment; retain a clear marker for fit availability. Explain retrospective cutoffs.
9. **Estimator disagreement.** Plot hypothetical R/S and DFA slopes with the “experimental” label. Do not call surrogate intervals Hurst confidence intervals.
10. **FMSI research concept.** Labeled candidate dimensions without a numeric composite. Footer: “Proposed framework; formula and calibration unavailable.”

## Reuse rules
Use one core teaching point per post. Vary the format between question, diagram, short explanation, and example. Link only to pages that actually exist. Quotes from external research need attribution and permission-compatible excerpts. None of these drafts has been posted or scheduled.