Hi everyone,
A quick update on Aletheia, my AI-driven research & trading project.
✅ Test Phase v1 — Summary
-
Period: early July → 3 Aug 2025 (~1.5 months).
-
Risk model: fixed 0.5% per trade.
-
Performance: +$2,274.02 net (≈ +4.55% on 50k). Profit factor 2.98, win rate 65% (13/20), largest win $889.92, largest loss −$256.93.
-
Drawdown: Max balance DD 0.67% (−$339.55).
-
Positioning: 20 shorts, 0 longs during v1.
-
End snapshot: Balance $52,274.02, Equity $52,803.14 at report time.
During the first month Aletheia climbed to roughly +5%, then a brief losing cluster appeared. I paused v1 and updated the agent to filter the reasoning patterns linked to those losses.
Full Trade History and Account reports are attached so you can audit every ticket.
🧭 Aletheia Web App — Public Preview
Explore how Aletheia turns global news into reasoned hypotheses and curated, broker-agnostic trade setups.
🧩 How the agent works (Home)
A visual, end-to-end flow of the pipeline with guardrails:
- Always-on monitor → enrich → hypothesis tree → evidence engine → position synthesis → execution & governance.
- Gatekeeper checks at key handoffs; clear status and outcomes.
📊 Dashboard
Live situational awareness in one place:
- KPIs & trendlines (win/loss, RR distribution, exposure, setup throughput).
- Global Impact Map (country-level choropleth by mentions, zoom/hover, top regions).
- Coverage by sectors, themes, regions; recent notable events and results.
🧠 Event Explorer
Drill from news events to full reasoning trails.
- Filters: Quick (Today / This week / This month) and Advanced (date range, themes, regions, sectors, text search, outcomes).
- Event page: context & expansion (entities/regions/sectors/themes), progress, decisions, branch & position counts.
- Branch page: hypothesis details, evidence & citation trail, gatekeeper verdicts, direction & latency, plus generated positions.
💹 Trade Setup Explorer
Browse all setups and jump straight to the reasoning that produced them.
- Filters: date, symbol(s), direction, status (open/closed/expired/discarded), result labels (e.g., tp2_hit , sl_hit ).
- Inline charts and compact cards (thesis, entries/targets/stops, risk model, current status).
- One-click open to the originating event or branch trail.
⚡ Quick Actions
Three cards to get moving fast:
- Dashboard → metrics & map
- Event Explorer → events → branches → evidence
- Trade Setups → browse & audit outcomes
👉 Web App: https://aletheia.giize.com
🔍 4 Example Setups From v1
Below are four real tickets from v1. Click the trail link to inspect the full reasoning in the app.
🟢 CRWD (Crowdstrike Holdings Inc) – 📍 Entry: 450.62 | TP1: 430.40 ✅ TP2: 410.18 ✅
Catalyst: Palo Alto Networks is in talks to acquire CyberArk in a deal valued at approximately $20 billion, potentially reshaping the cybersecurity market.
Reasoning trail: https://aletheia.giize.com/Event_Explorer?event=9594&branch=26994
🟢 IFX (Infineon Technologies AG) – 📍 Entry: 35.949 | TP1: 34.823 ✅ TP2: 33.697 ✅
Catalyst: Intel cancels planned semiconductor fabrication projects in Germany and Poland and consolidates testing and assembly operations in Vietnam and Malaysia.
Reasoning trail: https://aletheia.giize.com/Event_Explorer?event=8483&branch=24067
🟢 IBE (Iberdrola SA) – 📍 Entry: 15.32 | TP1: 15.69 ✅ TP2: 16.06 ✅
Catalyst: Government raises maximum price paid to wind farm developers for electricity.
Reasoning trail: https://aletheia.giize.com/Event_Explorer?event=8384&branch=23805
🟢 XRT (SPDR S&P Retail ETF) – 📍 Entry: 81.8 | TP1: 80.01 ✅ TP2: 78.22 ✅
Catalyst: U.S. companies raised prices for goods and services in July, citing tariffs as a key factor contributing to inflationary pressures.
Reasoning trail: https://aletheia.giize.com/Event_Explorer?event=8392&branch=23833
🔄 What changed after pausing v1
-
New Reasoning Auditor (continuous learning).
Scans the entire reasoning trail to validate that analyses are grounded in solid evidence. Continuously updated as more setups and outcomes arrive.
Why it matters: filters out weak, indirect logic and favors direct, causal impact between event → asset. -
Evidence gating tightened (carried forward).
Stricter source quality & recency checks; patterns linked to v1 losses are down-weighted or blocked.
Why it matters: reduces speculative, wide narratives slipping into setups. -
Websearch coverage & parsing improved.
Broader domain support + better parsers increase successful extractions and citation depth.
Why it matters: more (and cleaner) sources, fewer blind spots. -
Symbol Matcher upgraded.
Tighter ticker/venue mapping and conflict resolution to cut symbol mismatches.
Why it matters: improves asset targeting accuracy before orders are formed. -
Targets & RR optimizer (from v1 history).
Re-tuned target selection using all historical setups: shorter entry→SL, longer entry→TP1/TP2 distances.
Why it matters: raises average expected R while controlling downside. -
MT5 execution safety (lot-step aware).
Orders that were too large for broker constraints are now auto-split into smaller tickets so risk per setup remains exact.
Why it matters: consistent position sizing → more robust live behavior. -
Partial-closure reliability fix.
Edge cases that failed to trigger TP1/TP2 partials have been corrected; state sync improved.
Why it matters: predictable profit-taking and cleaner P/L attribution. -
Setup QA pass (carried forward).
Validator rejects vague or duplicate theses; enforces clarity before anything can become a trade.
Why it matters: higher-quality setups, fewer noisy entries.
Expected v2 effect: Fewer but higher-conviction trades, tighter risk, clearer audit trails.
🚀 Test Phase v2 — Starting Now
I’m kicking off v2 with the updated pipeline and the new web app.
Join v2: if you’d like to follow along, review setups, or stress-test reasoning, jump in: