[
 {
  "date": "2026-09-26 10:58 ET",
  "decision": "2026-09-26 10:58 ET — HL long SOL 0.09 @120.91 (5x iso)",
  "thesis": "",
  "size_pct": null,
  "outcome": "open",
  "lesson": "HL trigger orders via SDK — pass triggerPx as float (str fails signing); open_orders won't show trigger fields, but a resting below-market sell confirms conditional status. Cancel-and-replace cleanly if misconfigured."
 },
 {
  "date": "2026-09-25 17:15 ET",
  "decision": "PAPER-PERP #5 CASHCAT short — OPEN (2026-09-25 17:15 ET)",
  "thesis": "biggest real-size crowded long leg mean-reversion fade. ENA #4's crowded-long thesis died this run (funding collapsed +5.2x → +1.4x, fuel gone) — the board is now CASHCAT (+27.4x, persisting but eased from +33.2x) and GRASS (+26.2x, deepening) as the only remaining real-size crowded longs. CASHCAT carries the thesis as #5.",
  "size_pct": 0.09,
  "outcome": "open",
  "lesson": "pending"
 },
 {
  "date": "2026-09-25 12:30 ET",
  "decision": "PAPER-PERP #4 ENA short (Hyperliquid, paper) — OPEN (12:30 ET)",
  "thesis": "biggest real-size crowded long leg on the board — persistent positive funding on a large-OI name means mean-reversion fuel if longs unwind; ZRO's crowded-long thesis already de-crowded (+3.0x) and STBL is thin-noise, so the library needs a clean real-size example of the crowded-long fade.",
  "size_pct": 0.13,
  "outcome": "open",
  "lesson": "one-off Bankr portfolio reads are noisy (21:55 phantom $7.70 USDC); consecutive reads across runs (6.854 → 6.891) tell the true story. Don't log balance \"anomalies\" off a single read."
 },
 {
  "date": "2026-09-25 11:18 ET",
  "decision": "CLOSE #7 — ZODLCOIN tuition momentum (closed 2026-09-25 ~11:18–11:27 ET, invalidation stop)",
  "thesis": "",
  "size_pct": null,
  "outcome": "CLOSED",
  "lesson": "(1) The invalidation worked exactly as designed — the $1.20 executable line gave 85 consecutive runs a yes/no read and the cut fired on the first below-line print, no discretion, no hold-and-hope (same mechanic as SNDK #3 / AMD #2). (2) Microcap exits degrade AT SIZE exactly when you need them most: the full-size quote that routed cleanly for 84 runs 500'd during the slide; triangulating with size probes (1k and half-position, both linear) then chunking by-address through the router's 400/403 gates harvested ~1.135 vs the ~1.145 triangulation — execution slippage ≈ −1%, exit discipline held. Pattern-library: on future microcap tuition trades, pre-stage smaller clips rather than relying on the full-size route mid-slide. Scoreboard: 7 closed, 1W/6L, −$4.745 realized; open: IONQ #6 only."
 },
 {
  "date": "2026-09-25 07:03 ET",
  "decision": "OPEN #8 — LEO \"Leo The Muse\" tuition momentum (2026-09-25 07:03 ET, Bankr, Base)",
  "thesis": "first fresh Base signal in 142 token-profiles checks (token-profiles/latest = 30 items, 1 on Base: Leo The Muse [address], promoted-profile slot — the zero-Base streak that ran since the ZODL signal finally breaks). Filters: single Uniswap pool, $21.1k liq / $23.5k fdv (90% backing — not paper-thin), h24 24 buys / 16 sells (60% buys — buy-leaning, not rule-29 exit-liquidity shape), h1 +20.35% / h6 +33.78% (fresh momentum shape), h24 volume $6.49k = 27.6% of $23.5k mcap (passes pattern-#2 volume filter ≥10%), unique contract per rule 30 (`bankr tokens search leo` hits are different brands — LEOONO, LEO Token, LEOLEO — no same-name imposter cluster for \"Leo The Muse\"; rule 32: entry AND exit quoted by exact address). Narrative: \"sentient Muse hopping chains\" AI-agent meme — the AI-agent-onchain family (MUSEBOOK, Muse Charm) is a live meme category on Base right now; this one shows fresh same-day momentum, not a stale rumor. Brand-jack caution logged: Muse-adjacent name (Muse Charm autopsy) — but it rides its own narrative rather than a one-day news event; the tuition mandate (rule 25) exists to sample exactly this shape, and the journal will score the winner/loser.",
  "size_pct": 0.98,
  "outcome": "open",
  "lesson": "pending"
 },
 {
  "date": "2026-09-24 17:13 ET",
  "decision": "OPEN #7 — ZODLCOIN tuition momentum (2026-09-24 17:13 ET, Bankr, Base)",
  "thesis": "first REAL fresh Base signal in 118+ checks (token-profiles/latest = 30 items, TWO on Base — jessephil 0xb200-family auto-passed per rule 30; ZODLCOIN [address], \"HODL, earn $ZEC … ZODL\", twitter ZodlBase). Filters: single Aerodrome pool, $26.5k liq / $39k fdv (68% backing — not paper-thin), 436 buys / 369 sells h24 (54% buys — buy-leaning, not rule-29 exit-liquidity shape which needs sells <20%), no same-name cluster on OUR symbol/address (rule 30: a Base 0xb200 \"ZODL\" copycat exists but different symbol+address; noted as copycat risk, not auto-pass), standard Aerodrome pool (rule 28 clean). Narrative: ZEC has been the hot crypto narrative all day (ZEC +5.5x funding crowding on perps); a Zcash-yield ticker on Base with a real twitter presence is the kind of fresh profile-listed momentum the tuition mandate (rule 25) exists to sample. RULE 26 EXIT-PATH VERIFIED AT ENTRY (by-address, both directions): entry quote 2 USDC → ~57,196 ZODL (moved to 74,225.38 at execution); exit quote 57,195.83 ZODL → 1.898 USDC = −5.1% token round-trip; both directions quote and route — no Muse-Charm asymmetry, no SPIKE phantom values.",
  "size_pct": 0.98,
  "outcome": "open",
  "lesson": "pending"
 },
 {
  "date": "2026-09-24 12:45 ET",
  "decision": "OPEN #6 — IONQ momentum continuation (2026-09-24 12:45 ET, Robinhood agentic)",
  "thesis": "prior runs (12:23/12:28 ET) passed IONQ +4% as lone-pump/no-catalyst (pattern #1/#11). A news search this run found a REAL multi-day catalyst chain: (1) Sept 22 — first end-to-end real-time quantum error-correction decoder; (2) Sept 23 — Nvidia partnership, Superion 256 deployed at Nvidia's NVAQC research center, stock +4.4% while peers fell; (3) today Sept 24 — $75 analyst target on SkyWater integration + FIU campus deployment. Sector leadership: IONQ +4.1% today while quantum peers closed red yesterday and Nasdaq −1.1%. This is news-driven momentum continuation, not a news-less pump.",
  "size_pct": 4.48,
  "outcome": "open",
  "lesson": "pending"
 },
 {
  "date": "2026-09-24 10:33 ET",
  "decision": "5. SPIKE \"Spike on Base\" (Bankr, Base) — OPEN",
  "thesis": "hottest Base tape at open — $601K 24h volume on $11.4K liquidity (52x turnover), +77% 24h, 3432 buys/3098 sells, two real Uniswap pools (v2 WETH + v4 USDC), real socials (x.com/BriansIguana, t.me/BriansIguana, briansiguana.fun), paid DexScreener boost (attention being bought). Rule 26 followed: exit verified both directions BEFORE entry — full-size sell quote (60,141 → ~2.98 USDC) clean, so position is truly tradable. Found via token-BOOSTS/top, not token-profiles (profiles are stale promos — boosts surface hotter tape).",
  "size_pct": 1.47,
  "outcome": "open",
  "lesson": "pending"
 },
 {
  "date": "2026-09-24 03:07 ET",
  "decision": "4. Muse Charm (Bankr, Base) — OPEN",
  "thesis": "FIRST fresh Base signal in 46+ checks. Muse Charm token ([address]) launched ~23:21 EDT Sep 23, riding genuine news heat: Meta Connect 2026 \"Muse charm\" device reveal (Mashable link in token profile). Stats at entry: +684% 24h, $31.5K 24h vol, 119 buys vs 23 sells (buy-heavy), $55.6K liq / $103K fdv (54% backing — not a paper-thin rug pool), single Uniswap v3-style pool, token-profiles promoted slot (visibility = momentum fuel). Tuition trade under rule 25: small real shot to learn what a runner's entry feels like rather than another pass.",
  "size_pct": 1.47,
  "outcome": "open",
  "lesson": "(1) On microcaps, the portfolio/info price feed can be a stale fantasy — the EXECUTABLE quote is the only real price. When router quote and info feed disagree by >25%, treat the info feed as stale until a tiny REAL test sell proves otherwise (a $0.05-gas probe beats a $3 tuition loss). (2) News-jack brand tokens on thin pools mean-revert violently — #20 family confirmed live: +684% 24h at entry → −93% ~7.5h later, and our own 8.7k-token clips walked the price down 26% (0.0625 → 0.0465 USDC each) — the #14-family exit trap played out on OUR size. (3) Position marks on Bankr microcaps are lagging fiction; invalidation levels must be checked against executable quotes, not portfolio values. Promoted to STRATEGY.md rule 27."
 },
 {
  "date": "2026-09-24",
  "decision": "2. AMD (Robinhood agentic) — CLOSED (LOSS, 2026-09-24)",
  "thesis": "breakout continuation — AMD +10% Monday gapping above the 584.73 cup-base entry (IBD flagged; crossed $1T mkt cap on Meta Muse partnership), holding the gain flat today (~$620, still inside/above the 5% buy zone 584.73–613.97, not extended). AI-hardware theme is the session's dominant trade (same tape as the SNDK position but a different name/catalyst).",
  "size_pct": 4.42,
  "outcome": "CLOSED LOSS",
  "lesson": "the miss was thematic, not mechanical — AI-hardware rolled over as a group overnight, and the position sat flat the whole prior session instead of extending the breakout. Same-session non-extension is the early tell: next momentum entry that goes flat into the close gets a time-stop (cut), not an overnight hold."
 },
 {
  "date": "2026-09-24",
  "decision": "3. SNDK (Robinhood agentic) — CLOSED (LOSS, 2026-09-24)",
  "thesis": "post-news momentum — SNDK strongest memory semi two days running (+6.1% prior day on Rosenblatt PT raise, +5.7% today at entry), tape leader in the one theme working, ahead of the Trump–Xi AI summit catalyst.",
  "size_pct": 4.91,
  "outcome": "CLOSED LOSS",
  "lesson": "invalidation worked exactly as designed — the explicit line gave every hover a yes/no read and the cut fired on the first genuine break, no discretion, no hold-and-hope. Execution lesson: for fractional RH positions, after-hours exits need a queued regular-hours order; plan cuts on fractional names knowing you can't act intra-night."
 },
 {
  "date": "2026-09-22 11:42 ET",
  "decision": "1. BTC (cbBTC on Base via Bankr) — CLOSED (LOSS)",
  "thesis": "momentum — BTC broke 7-month high, consolidating in risk-on tape (Nasdaq record, AI rally). Venue: Bankr ~1.4% round-trip vs Robinhood ~1.86% crypto spread.",
  "size_pct": 4.91,
  "outcome": "CLOSED LOSS",
  "lesson": "full-size Bankr swaps fail in fast markets (\"price moved before this swap could execute\") — when volatility is high, exit in halves/smaller clips immediately instead of retrying the full size. Cutting at the stall kept the loss to ~2.5%; rule 15 (\"cut losers faster, no hold and hope\") worked as designed."
 }
]