Session Log — Tuesday, April 14 2026¶
What We Worked On¶
- NEO Trading Engine — analyzed two 10 RLUSD live sessions that both halted early on exposure caps (RLUSD and XRP sides)
- Identified and fixed mid-price reconstruction bug in both
dashboard.pyandsummarize_paper_run.py(global median → per-tickmarket_snapshots.mid_price) - Identified and fixed terminal summary session scoping bug in
summarize_paper_run.py(all queries now filter toMAX(session_id)) - Verified both fixes with 7 RLUSD baseline session — 0% noise ticks, terminal = dashboard agreement confirmed
- Identified FLAG-013: stale requote TTL (300s) blind to skew changes, causing 9–12 bps placement mismatch under active skew
- Implemented deviation-triggered requote fix (
skew_deviation_threshold_bps) with 90s TTL backstop - Threshold calibration issue: 5 bps caused cancel churn → raised to 10 bps
- Identified and fixed stuck orders blocking placement (sessions 7 and 9 left
pending_submissionandcancel_pendingorders in DB) - Added
_startup_force_cancel_stuck_orders()to prevent recurrence - Intel exchange with Titan Ghost (competitor XRPL market maker) — logged in
[C] Intel Log.md
What Was Built or Changed¶
dashboard.py— mid-noise fix (merge_asof on market_snapshots)summarize_paper_run.py— mid-noise fix + session_id scoping on all queriesneo_engine/config.py—OrderReplacementConfigaddsskew_deviation_threshold_bps: float = 5.0neo_engine/strategy_engine.py— cacheslast_bid/ask_offset_bpsafter eachcalculate_quote()neo_engine/main_loop.py—_should_replace()deviation check +_startup_force_cancel_stuck_orders()config/config_live_stage1.yaml— TTL 300s → 90s,skew_deviation_threshold_bps: 10.0, sell offset locked at 16 bpstests/test_main_loop.py— 4 new tests for deviation trigger[C] Open Flags.md— FLAGS 011, 012, 013 added; 010 resolved; 002 updated[C] Intel Log.md— created with Titan Ghost INTEL-001 entry[C] Agent Alignment Briefing 2026-04-14.md— created for agent handoffneo-2026/INTEL/— 33 screenshots reviewed and logged
Still Open¶
- Session 11 (10 RLUSD re-test with deviation fix at 10 bps) — just started, needs to complete and be reviewed
- FLAG-012: drift-to-halt at 10 RLUSD — still needs clean re-test data after FLAG-013 fix
- FLAG-008: WAC rebuild bug — still blocks capital injection path
- FLAG-002: anchor 73% capping — monitor; review if exceeds ~85%
- Intel follow-up: ask Titan Ghost about the -42 XRP controller failure root cause
Start Here Tomorrow¶
Review session 11 terminal summary — it should show deviation-triggered requotes firing fast, fill count higher than the 2-fill baseline, and no cancel churn. If it passes, FLAG-013 is resolved and 10 RLUSD is validated; next decision is whether drift-to-halt (FLAG-012) reappears with clean signal.