---
session_id: 2026-09-19-block-032-probepass
date: 2026-09-19
block: 032
block_name: Credit OAS mechanics
phase: Phase 1 (Credit)
status: ◐ in-progress
session_paused: true
session_closed_clean: false
resume_point: Q1 re-do (user request) then main material targets
---

# 2026-09-19 (11:23 session) — Block 032 probe pass — PAUSED at close of pass

## What happened this session

1. **P0 retention spot-check (card #10, cross-over level-vs-cascade): Working+** — tier-level read clean; divergence refinement delivered (price-vs-fundamentals moving opposite = market 'short the cliff').
2. **Probe pass — 5 of 5 probes worked:**
   - Q1 (Z vs OAS identity, 3-bond ranking): **Working** — ordering CORRECT (putable > bullet > callable) but labeling slipped (B/C mixed, callable called 'zero optionality'). Identity needs one more rep. **User requested re-do next session.**
   - Q2 (OAS-as-solve, two-dealer gap): **Working+** — model-dependence recalled clean; two failure modes TAUGHT (benchmark/input differences — 'over what?'; stale/matrixed price — kills both quotes, go to CDS). Carded (#3 in probe batch).
   - Q3 (new-issue concession): **Working** — mechanism right, sign/vocabulary flip ('rich' vs 'wide'; also MY phrasing error introduced the ambiguity — corrected, carded). Issuer-curve widening insight partially there. Carded (#4, #5).
   - Q4 (index leads, negative basis, arb constraints): **Working+** — cleanest of the session; stress hedge vs cash exit trade-off added.
   - Q5 (transition: BBB−IG 18bp × late-Sep BBB-tilted issuance window): **Working+** — taught the complication (heavy BBB supply that FAILS to widen the tier = stronger demand-driven complacency, not weaker); test named (concession sizes + grounding speed + CDX BBB bucket cross-check).
3. **8/8 new cards pushed and VERIFIED UPLOADED via usn check** (tags `block-032 credit pre-block-probe 2026-09-19`). Block 032 deck total: **18 notes**. Credit subdeck total: 51 notes.
4. Mentor error logged: 'rich concession' phrasing in Q3 introduced the trap being tested — corrected to 'full concession'; rich/cheap = value, wide/tight = level, full/light = concession size.

## Resume point

**Q1 re-do (user request, Z vs OAS identity — currently Working)** then **main material targets**:
- T1: OAS calculation walk (binomial mechanics — the solve)
- T2: Z-vs-OAS in practice (reading the option cost in spread terms)
- T3: New-issue concession pricing (the concession as trade)
- T4: Index-vs-single-name liquidity & basis trades — closing with live numeric walk (user preference: live data, name-the-check)

Probe-pass leftovers: none (all 5 worked, P0 included).

## Carry-forwards

process_notes:
  - Close protocol now includes: verify Anki sync via negative-usn check AFTER sync call, never trust clean API return.
  - Session pauses mid-probe-pass are fine — resume file carries the exact queue position (Q2 was resumed cleanly this session; same pattern next).
  - User's Q1 labeling slip: identical error class to the 031 'which leg moved more' inversions — when re-doing Q1, make user verbalize which party holds the option BEFORE ranking.
to_block_033:
  - (default cycle lag — nothing yet)