# Kalshi AI — Mispricing Audit (v4, archetype-targeted) — 2026-08-04

Date: 2026-08-04
Source: https://kalshi-audits.pages.dev/2026-08-04-kalshi-ai-mispricing-audit-v4

---
# Kalshi AI — Mispricing Audit (v4, archetype-targeted)

Report date 2026-08-04 · Capital $2,500 · Horizon 45 days (to Sep 18) · Universe: targets.json allowlist (117 series, generated 2026-08-04 15:06 UTC) · Category focus: AI (advisory — no allowlisted AI-category series had qualifying open markets in horizon; slate worked Tier-1-first as designed)

## Verdict

**1 pick, 11 logged rejects, $81 deployed (3.2%), $2,419 cash (96.8%).**
  Stage 0 produced only 6 mechanically-clean markets across two theses — the Blanche AG-confirmation
  complex and the brand-new Pirro-out-as-DC-US-Attorney ladder. The Blanche market repriced +56 points
  in 48 hours and now sits where the process math puts it; the Pirro ladder still carries residual
  departure premium after she survived yesterday's White House meeting, and its Sep 1 leg offers a
  precedent-backed NO in the core band. An efficient week; cash is the position.

## Pick 1 of 1 — BUY NO · KXPIRROOUT-26AUG-26SEP01

### Will Jeanine Pirro announce her departure as U.S. Attorney for D.C. before Sep 1, 2026?

BUY NO @ 64¢ limit
MEDIUM conviction
archetype: multi-news-other (Tier 2)
cluster: pirro-out-dc
Prob (YES): 32% pre-market → 33% final (blended slightly toward a liquid book) → NO side 67% vs ~63¢ cost
Entry context: YES 38/39 · NO fills 62–64¢ · depth 388 contracts at/below limit · vol24h 18,061 · OI 11,605
Size: 129 contracts (depth cap 388/3), ~$81 max cost — 3.2% of capital, under the 7% MEDIUM cap
EV for the record: +4.2¢/contract on 62.8¢ avg fill ≈ +6.7% on cost

**The situation.** Trump is publicly furious at Pirro over her office's Aug 1 court filing
  concluding the Lincoln Memorial Reflecting Pool damage was a flawed renovation, not vandalism — he said she
  "choked," sources told CNN removal is "highly likely," and he "fumed all weekend" (CNN, WTOP, Forbes, Aug 3).
  But at the Aug 3 White House meeting she was _not fired and did not resign_, and she blamed career
  officials for the filing (CNN, Aug 3–4). No successor has been floated. She is Senate-confirmed
  (Aug 2025, after Ed Martin's nomination collapsed), so removing her reopens a confirmation problem.
  The market listed Aug 1 near 97¢, crashed to 38–40¢ on the meeting-survival news, and sits at 38/39 —
  still pricing a ~2-in-5 departure inside four weeks.

**Why NO at 62–64¢.** The precedent store shows exactly this shape resolving NO repeatedly:

> Precedent: KXKASHOUT-26APR-JUN01 resolved NO — Kalshi priced the rung at 67% off one week of
>   anonymous "matter of time" sourcing; lesson: "sell the spike into the news cycle rather than buying it."
>   KXKASHOUT-26APR-JUL01 (priced 77%) also resolved NO.

> Precedent: KXLUTNICKOUT-26JUL01 resolved NO — "even cross-party resignation calls plus documentary
>   scandal do not move this archetype… absent a named successor, price any single month near the base rate."

This case is stronger than those panics in one respect — the President himself, not anonymous sources, is
  attacking her — which is why the number is 33% and not the sub-25% raw base rate. The live
  `rules_secondary` was read this run: resolution is _announcement-based, effective date irrelevant_,
  and any cessation counts (removal, resignation, reassignment out of the role). That widens YES paths — a
  face-saving transfer, the exact treatment Ed Martin got, would resolve YES — and is priced into the blend.

**Resolver profile & risk pricing.** Resolver: Kalshi multi-news desk on a public-announcement
  standard; boundaries are unusually explicit in `rules_secondary` (temporary suspensions/recusals don't
  count; death → last-fair-price). Forecasting the resolver's verdict, not the vibes: nothing short of an actual
  announced departure pays YES. Deductions: −3 pts for "officially communicated" edge cases on a volatile principal
  (e.g. an ambiguous Truth Social post the desk must interpret). Total deductions 3 pts < 15 — passes.

**The live way to be wrong.** One Truth Social post fires her and NO loses 63¢/contract
  instantly; the early-close clause settles it same-day. That risk is why this is MEDIUM, depth-capped,
  and a single leg — the Aug 5/6/7 rungs of the same thesis are logged as cluster runner-ups, not stacked.

Precedents consulted: KXKASHOUT-26APR-JUN01 (NO) · KXKASHOUT-26APR-JUL01 (NO) · KXLUTNICKOUT-26JUL01 (NO) · KXLEAVEPOWELLGOV-26JUL01 (NO) · KXTRUMPADMINLEAVE-26DEC31-TGAB (YES — announcement/effective-date mechanics)
Execution: Limit 64¢, 129 contracts; expected 25@62 + 104@63 ≈ 62.8¢ avg. Do not chase above 64¢ — this book repriced hourly through the Aug 3–4 cycle. Specialist-dense (vol24h > 5k): size stays at the depth cap.

## Position sizes & limits

| Market | Side | Limit | Contracts | Max cost | Max win | Conviction | % of capital |
|---|---|---|---|---|---|---|---|
| KXPIRROOUT-26AUG-26SEP01 | NO | 64¢ | 129 | $81 | $48 | MEDIUM | 3.2% |
| **Cash** | — | — | — | — | — | — | **96.8%** |

Reserve requirement (≥10% of $2,500) satisfied trivially. Depth cap binds before the 7% MEDIUM cost cap ($175).

## EV for the record

| Market | Side | Avg cost | Model win prob | EV/contract | EV % cost | Portfolio EV |
|---|---|---|---|---|---|---|
| KXPIRROOUT-26AUG-26SEP01 | NO | 62.8¢ | 67% | +4.2¢ | +6.7% | +$5.4 |

## Clusters (one leg per cluster)

| Cluster | Deployed leg | Runner-ups logged as rejects |
|---|---|---|
| pirro-out-dc | SEP01 NO @ 62–64¢ (core band) | AUG07 NO (88–89¢ thin favorite), AUG06 NO (~90¢), AUG05 NO (94–95¢ deep) |
| blanche-ag-confirm | _none — no edge_ | KXAGCONF-26JUN05-SEP01, KXBLANCHEJUDICIARY-27-26SEP01 |

## Conviction exposure

| Tier | Positions | Cost | % of capital | Cap |
|---|---|---|---|---|
| HIGH | 0 | $0 | 0% | 15%/pick |
| MEDIUM | 1 | $81 | 3.2% | 7%/pick |
| Cash | — | $2,419 | 96.8% | — |

## Rejected (logged for shadow-tracking)

| Market | Archetype | Price | Side/Prob | Reason |
|---|---|---|---|---|
| KXAGCONF-26JUN05-SEP01 Blanche confirmed AG before Sep 1 | confirmation/enact | 74/76 | YES · 72% | No edge at an efficient price. Repriced +56 pts in 48h (Cornyn/Tillis deal Sunday; 12-10 committee vote this morning). Punchbowl floor plan — cloture filed Wed, cloture vote Fri, final Sat Aug 8 before a 5-week recess — puts my process math (~0.90 cloture × 0.85 schedule-holds × 0.96 final ≈ 72%) inside the book. Precedent (KXAGCONF Jun/Jul rungs, all NO): clock runs from transmittal; this rung is the first with enough runway. NO side <40¢ band. |
| KXPIRROOUT-26AUG-26AUG07 | multi-news-other | 11/12 | NO · 91% | Cluster runner-up to SEP01 NO; entry 88–89¢ is the thin-favorite band against a publicly furious President 3 days out — the shape of the record's worst loss (the 94¢ "lock"). |
| KXPIRROOUT-26AUG-26AUG06 | multi-news-other | 10/13 | NO | Cluster runner-up; NO ~87–90¢ straddles the deep-favorite reject line. |
| KXPIRROOUT-26AUG-26AUG05 | multi-news-other | 5/6 | NO | Deep-favorite band (NO 94–95¢); closes in <13h. |
| KXCBDECISIONCANADA-26SEP-H0 | scheduled-process | 92/95 | YES | Deep-favorite band (≥90¢ reject) on a month-out macro meeting. |
| KXTRUMPTHUNEOUT-27JAN01-SEP | departure/ouster | 29/32 | NO | Liquidity floor: vol24h 214 < 500. Zero-volume cohort had negative realized edge; no exceptions. |
| KXJUDGECOUNT-26AUG-1 ≥2 judges confirmed in Aug | confirmation/enact | 29/30 | NO | Liquidity floor: vol24h 5 < 500, despite an attractive recess-based NO thesis (Senate gone Aug 7–Sep 7). Phantom edge in a dead book — the documented v1–v3 failure mode. |
| KXBLANCHEJUDICIARY-27-26SEP01 | multi-news-other | 0/100 | — | Broken book; committee voted 12-10 this morning, resolution imminent. |
| KXSENATECONFIRM-26APR29-BMAT | confirmation/enact | 91/99 | — | Mechanical: spread 8¢, vol24h 134; deep-favorite besides. |
| KXUSAIRANAGREEMENT-27-26SEP | scheduled-process | 2/3 | — | Mechanical: YES ask ≤5¢ dead tail (tails 0-for-8 lifetime). |
| KXHEGSETHOUT-26APR-SEP01 | departure/ouster | — | — | Already an active pick. |

A further ~28 in-horizon allowlist markets were cut mechanically at Stage 0 (YES ask ≤5¢ dead tails, vol24h < 500, or spread > 5¢) without Stage-2 research: KXALITOOUT, KXKASHOUT, KXIPO* (4), KXIMPEACH, KXSENATECONFIRM (5 more), KXLEAVE* (5), KXFARMBILL, KXSANCTIONRUSSIA, KXTARIFFBILL, KXRECNCH, KXMARTINDNCOUT, KXLEAVEMILLS, KXGALLEGOOUT, KXDEMSCHUMER, KXFARAGEOUTREFORM, KXBURNHAMOUT, KXLUTNICKOUT, KXNYTCHARGE, KXMCDONALDCONF, KXLAKECONF, KXCOMPANYACTIONMERGER, KXCBDECISIONCANADA (2). None showed a book worth a rules read.

## Risk profile

- **Max portfolio loss:** $81 (3.2% of capital) — Pirro is fired/announces before Sep 1.
- **Single point of failure:** one principal's mood; the early-close clause settles same-day on an announcement. No second leg exists on this thesis by design.
- **Correlation with existing picks:** none direct. KXHEGSETHOUT NO (existing) shares the "Trump doesn't actually fire people" macro-thesis; combined worst case if the administration enters a mass-firing news cycle is both legs losing, which argues for no further departure/ouster NOs this run — and none were added.
- **Efficiency note:** both researched theses repriced violently within 48h before this audit ran. v4's answer is small size, core band, one leg, 96.8% cash.

## Precedent writeback

None this run — no settled target-series market required reconstruction (the KXAGCONF and KXKASHOUT
  histories consulted are already in the store; KXBLANCHEJUDICIARY and the Pirro Aug rungs are not yet settled).
  `precedents.json` ships empty.

## Method & sources

**Pipeline:** Stage 0 allowlist scan (117 series → 95 in-horizon markets → 6 mechanical survivors) →
  Stage 1 precedent retrieval (26 rows across KXAGCONF, departure/ouster, confirmation/enact) → Stage 2 probability-before-price
  (pre-market numbers committed after rules + news, before book study) → Stage 2.5 resolver profiling → Stage 3 reject screens →
  Stage 3.5 bands → Stage 4 one-leg-per-cluster sizing. Prices are live Kalshi `*_fp / *_dollars` reads, 2026-08-04 15:00–16:30 UTC.

**Primary sources:**
[CNN — Trump doesn't fire Pirro at White House meeting (Aug 3)](https://www.cnn.com/2026/08/03/politics/jeanine-pirro-removal-reflecting-pool) ·
  [WTOP/CNN — Furious Trump considering removing Pirro (Aug 3)](https://wtop.com/dc/2026/08/furious-trump-is-considering-removing-pirro-over-reflecting-pool-filing-sources-say/) ·
  [Forbes — "she choked" (Aug 3)](https://www.forbes.com/sites/conormurray/2026/08/03/trump-reportedly-considers-firing-jeanine-pirro-over-reflecting-pool-says-she-choked/) ·
  [NPR — Judiciary Committee to vote on Blanche (Aug 4)](https://www.npr.org/2026/08/04/nx-s1-5918355/senate-judiciary-committee-to-vote-on-todd-blanches-nomination-for-attorney-general) ·
  [AP/US News — panel advances Blanche 12-10 after GOP-holdout deal (Aug 4)](https://www.usnews.com/news/politics/articles/2026-08-04/senate-panel-expected-to-advance-blanches-attorney-general-nomination-after-deal-with-gop-holdouts) ·
  [Punchbowl — Thune's pre-recess floor plan: cloture Wed→Fri, final Sat](https://punchbowl.news/article/senate/next-for-thune/) ·
  [The Hill — Blanche momentum, GOP obstacles, McConnell absence math](https://thehill.com/homenews/senate/6006947-gop-swing-votes-attorney-general/) ·
  [NPR — Blanche rescinds $1.8B fund to win Cornyn/Tillis (Aug 3)](https://www.npr.org/2026/08/03/g-s1-136864/todd-blanche-rescinds-trump-anti-weaponization-fund) ·
  [The Hill — Ed Martin exit / Pirro takeover background (May 2025)](https://thehill.com/regulation/court-battles/5298314-ed-martin-resigns-us-attorney-jeanine-pirro/)