---
session_id: 2026-09-22-block-032-main-pass
date: 2026-09-22
block: 032
block_name: Credit OAS mechanics
phase: Phase 1 (Credit)
status: ◐ in-progress
session_paused: true
session_closed_clean: false
resume_point: T4 re-probe (user must re-answer, question below) then session close
---

# 2026-09-22 — Block 032 main material pass — PAUSED at T4 re-probe

## What happened this session

1. **Q1 re-do (Z vs OAS identity): Working+** — party-holding verbalized first per carry-forward, labels clean, ordering correct (putable > bullet > callable). Two sharpenings taught: (a) Z *embeds* option value (not "ignores"), OAS strips it — Z−OAS = option cost in spread terms; (b) bullet Z≈OAS approximate because of model dependence (first-order) + liquidity/funding (second-order).
2. **T1 (OAS solve walk): Working+** — trial-spread direction correct (model value above market → raise spread). Two taught refinements: trial spread is a CONSTANT added at every node (flat add-back = what makes it OAS); higher-vol re-solve on callable → OAS widens via "model value fell, spread solves wider" — NOT "we'd pay less" (price held fixed; decomposition shift, not demand story).
3. **T2 (Z-vs-OAS in practice): Working+** — colleague's Z-comparison trap correctly rejected; bullet cheap / callable rich (Z150/OAS110 vs bullet 125 → callable 15bp rich). Refinement: Z *embeds* optionality. Card queued.
4. **T3 (concession ground-check): Working** — part (a) clean (20→5bp in 10 days = decent demand). Part (b): user reached for sophisticated-but-secondary pool-effect explanation BEFORE the mechanical check — old-curve 160 is likely a seasoned/different curve point; must compare interpolated 10y vs 10y. User pushback accepted as fair (probe didn't specify seasoning) — scored check-ordering slip, not substrate error. Re-test at close: version WITH maturity info given, user must spot roll/curve differential as the residual. Card queued.
5. **T4 (basis trade mechanics): teach delivered, re-probe PENDING.** User said "don't know, teach" on the live numeric walk. Teach delivered: convergence vs carry decomposition (30bp convergence + ~7.5bp carry on −45→−15 over 2mo); stress flip (index +130 vs basket +75 → basis −45→+10, flow diagnostic not credit); exit liquid-leg-first, cash side realizes worse than marks; sizing rule (convergence bet with stress asymmetry, not carry position).
6. **T4 RE-PROBE (must be re-asked, user had to leave):** Index −45 basis, short index protection / long cash basket protection, notional-matched. Stress day: index gaps to 505, basket gaps to 470. (a) basis now, flipped? (b) net P&L in bp; (c) which leg exits first + why; does the shape say credit deteriorated? [Answers: basis = 505−470 = +35, flipped; net = −35 index leg +5 basket leg = −30bp... wait: index leg you RECEIVE 420, marks 505 → −85bp; basket leg you PAY 470 vs 465 → +5bp... Mentor should recompute at resume: index widened 85, basket widened 5 → basis −45 → +35... verify arithmetic before scoring.]
7. **Cards created this session:** 4 new notes pushed + sync verified (tags `block-032 credit 2026-09-22`). Block 032 deck total: 22 notes. Credit subdeck: 55.

## Resume point

**Re-ask the T4 re-probe** (scenario in point 6 — recompute the arithmetic before scoring; the stress magnitudes were index 420→505 (+85), basket 465→470 (+5)). Then:
- T3 close re-test (curve-point version with maturity given)
- Session close protocol: resume file regeneration, competence-map updates, session-log patch, next-block pre-read assignment (Block 033 per curriculum).

## Carry-forwards

process_notes:
  - Check-ordering habit: user reached for sophisticated explanation (pool effect) before mechanical verification (curve-point match). Mentor framing: mechanical check first, sophistication second. Re-test queued.
  - Teach-then-re-probe fired (T4 "don't know, teach") — re-probe pending at resume; do not skip it (probe-teach-probe pattern).
  - Anki sync verified via negative-usn check per standing rule.
  - User convention reminder: "bonds rally" = prices up/yields down; prefer unambiguous yield-direction phrasing in probes.
to_block_033:
  - (default cycle lag — nothing yet)