Company Fundamentals
Revenue, margins, guidance, filings — the base layer of every research view.
Market Coverage
Follow the evidence across every dimension of the market.
Each tracked asset is researched across eight dimensions — then cross-checked by a second, independent engine before anything is published.
Valuation
Multiples, growth-adjusted context, and scenario-based valuation ranges.
Technical Structure
Trend, momentum, and key levels — as research method, never as a promise of accuracy.
Capital Flows
Volume behavior and positioning context around tracked names.
Options Context
Options activity as background evidence for the research picture.
Macro & Industry
Rates, cycles, and sector dynamics that frame each watchlist group.
Market Sentiment
News and social tone, scored and time-stamped as a research input.
Cross-Source Verification
A second engine independently re-derives the numbers before anything ships.
The System
Dual-Engine. Multi-Model Council.
Two AI engines — Hermes Agent and Codex — split the work across fundamentals, technicals, and macro, and cross-check each other before a council of models decides. Every part earns its place; nothing runs that can't be explained.
01 · Collect
12+ data sources
Prices, news, filings, social sentiment, market-structure signals, insider flow, and supply-chain mapping — pulled by 100+ scheduled jobs.
02 · Reason
Two engines, split roles
Hermes Agent leads fundamentals, macro & sentiment; Codex runs technicals — and independently re-derives every key number before it ships.
03 · Verify
Maker / Checker
No single model is allowed to be right alone. A second engine independently re-derives the key numbers; divergence is surfaced, not hidden.
04 · Deliver
Two languages, two channels
English to Telegram, Chinese to Feishu — every report carries a compliance line. WATCHLIST_ONLY by architecture.
One model can be confidently wrong.
A second engine catches it. A multi-model council decides.
Engine 1 — Hermes Agent
The orchestrator and lead analyst. Runs the all-weather pipeline on one Mac mini (Apple Silicon), leads fundamentals, macro & sentiment, and drafts every report.
Engine 2 — Codex
The independent checker. Re-derives Engine 1's key numbers — valuation, growth, technicals — and prints any divergence instead of smoothing it over. The single most defensible feature in the system.
The Multi-Model Council
DeepSeek V4 Pro
GLM-5.2
GPT-5.6
+ Qwen3 14B (local)
Fallback chain: GLM-5.2 → GPT-5.6 → DeepSeek V4 Pro → Qwen3 14B (local, free)
What TradingMapClaw is
- An AI research workflow for US equities — fundamentals, valuation, technicals, sentiment, and macro across your own watchlist
- A product we sell as skill packs, tutorials, and consulting so builders can run the system themselves
- A transparent, cross-verified pipeline that shows its work — every number re-derived by a second engine
What it never touches
- No buying, selling, or routing of orders — there is no broker API and no trade execution
- No stock recommendations and no personalized investment advice
- No managed money and no promise of returns — WATCHLIST_ONLY, research & education only
In short: we sell the AI stock-research system, not stock picks. Everything here is for research and education — not investment advice.
Latest Intelligence
This week's radar, straight from the pipeline.
Data-driven, not decorative — this block reflects the actual latest published period. See the full All-Weather Run →
PERIOD 2026-W29 · DATA AS OF 2026-07-10
Weekly Market Radar —
Desensitized watchlist observations across 82 tracked assets, with bullish / neutral / bearish scenario labels — research context, never trade instructions.
Latest update: 2026-07-13 · Next scheduled publication: Weekly Market Radar — Monday, July 20, 2026
Free Archive
What this desk publishes — on a fixed, honest cadence.
This is a working desk, not a brochure. Updated weekly, never “live”.
WEEKLY · MONDAYS
Market Radar
Desensitized watchlist observations across 82 tracked assets with bullish / neutral / bearish scenario labels — research context, never trade instructions.
View the archive →
WEEKLY
System Build Log
What broke, what got fixed, what it cost. The unglamorous engineering diary of running 100+ workflows on a $55/month budget.
Read the build log →
EVERY TWO WEEKS
Engine Disagreement Notes
Where Hermes Agent and Codex reached different conclusions on the same question — and how the council resolved it.
See the notes →
TWICE MONTHLY
Deep Analysis
Long-form research walkthroughs: supply-chain mapping, S-Factor scoring, sentiment structure — the method shown end to end.
See the method →
OCCASIONAL
Research Method Releases
Selected methods from inside the system — published when they are stable. Part of the method stays public; part stays commercial.
Browse methods →
WEEKLY · METHODOLOGY LOG
Verifiable Run Records
Which workflows ran, what was produced, what failed — a methodology log you can audit. No fabricated returns, no cherry-picked wins.
Audit the system →
The Story
A man who refused to accept the dimensions of the room he was given.
Eight surgeries. One working hand. Years inside global-bank operations. Then a dual-engine research system built from the ground up in two months, on a single Mac mini.
Read the full story Brachial plexus injury, eight surgeries, the back offices of five global banks — and how it all became one system.Prefer it raw? The unabridged STORY.md lives in the same GitHub repo as the system itself →
Get the weekly radar brief
One email a week from inside the system: what broke, what got fixed, what it cost, and the engineering patterns you can reuse. The first email includes a sample deep-dive report.
Free · One email a week · Unsubscribe anytime · Never investment advice