Skip to content
HWA GALLERY & CO.INTELLIGENT SYSTEMSSTART A CONVERSATION ↗
HWA ENGINE / BUILT WITH CLAUDE

Research. Decide. Review. Learn.

An operational autonomous AI stock research and paper-trading system, implemented using Claude AI and Claude Code — currently being extended toward secure, controlled broker-connected trading infrastructure.

PHASE 0 OPERATIONAL · SIMULATED TRADINGPHASE 1 IN DEVELOPMENT
03PAPER PORTFOLIOS

Swing, long-term, and meme books — each from a simulated $100k start.

06CLAUDE CODE SKILLS

The complete behavior of the agent, defined in versioned skill files.

25RECONCILED SIMULATED FILLS

As of the 2026-10-08 evidence snapshot, reconciled across ledger, portfolios, and journals.

FIGURES DERIVED FROM THE PHASE 0 REPOSITORY RECORD, AS OF 2026-10-08 · ALL TRADING SIMULATED · NO REAL CAPITAL

PHASE 0 / SANITIZED CAPTURE · ESC TO CLOSE
AutoStockTracker Phase 0 research overview: an agent that investigates stocks, simulates its decisions, reviews its own calls and keeps the lessons, with live record counts and the four-stage research cycle.
Actual AutoStockTracker Phase 0 research environment · simulated portfolios. Counts are generated from the repository’s own journals, ledger, and lessons. VIEW FULL SIZE ↗
02 / HOW THE AGENT WORKSPHASE 0 — IMPLEMENTED & TESTED

One loop, recorded end to end.

Every stage below is implemented and has operated on a scheduled weekday cadence. Each run leaves dated journals, simulated fills, and reports in the repository record — the same records this page’s figures are drawn from.

  1. 01Investigate

    Scheduled Claude workflows screen the market and research individual equities with valuation-first analysis.

  2. 02Decide

    Each call is written down before the outcome: thesis, dated evidence, falsifiable expectation, invalidation level.

  3. 03Simulate

    Decisions become simulated paper fills in one of three $100k books — never real-money orders in Phase 0.

  4. 04Review

    Retrospectives grade the original reasoning separately from the market outcome, against what was written at entry.

  5. 05Learn

    Durable lessons are recorded in persistent project memory and read by later research runs.

PROCESS DIAGRAM DERIVED FROM THE PHASE 0 METHODOLOGY · THE FIFTH STEP FEEDS THE FIRST

03 / THE CLAUDE FOUNDATIONTWO DOCUMENTED ROLES

Built with Claude, from the beginning.

ROLE 1 · BUILDING THE SOFTWARE

Claude Code as the engineering partner

The entire Phase 0 system — research methodology, journal and ledger structure, skill definitions, and reporting views — was implemented with Claude AI and Claude Code in interactive sessions under the founder’s direction.

ROLE 2 · OPERATING THE RESEARCH

Claude Code skills as the running agent

Separately, six Claude Code skills define everything the agent does, executed as scheduled workflows — research at the open, midday trade runs, close-of-day reports, and retrospective reviews. Each run writes its own dated records.

  • /analyze

    Valuation-first deep dive on one ticker, ending in a verdict per research book.

  • /market-scan

    Screener-sourced undervalued candidates, market pulse, and sentiment sweep.

  • /trade-run

    Fills pending orders, sweeps invalidations and targets, executes simulated trades, and writes the day’s journal.

  • /meme-run

    Sentiment-driven meme-book judgments — no valuation gate, tight stops.

  • /daily-report

    Close-of-day report: book status, allocation charts, valuation table, tomorrow’s plan.

  • /review-calls

    Retrospective: judges past calls against their original theses and extracts durable lessons.

PHASE 0 / SANITIZED CAPTURE · ESC TO CLOSE
Phase 0 operations page: the six Claude Code skill definitions rendered from the repository, the scheduled weekday routine cadence, and the run-coverage record including its known gaps.
How the agent runs — the six Claude Code skill definitions, the scheduled weekday cadence, and the run-coverage record with its gaps stated plainly. VIEW FULL SIZE ↗
04 / THE RECORD, IN VIEWSANITIZED PHASE 0 CAPTURES

Evidence you can zoom into.

Captures of the Phase 0 reporting environment, taken 2026-10-08. No data, wording, prices, or dates were altered; gaps in the record are shown as gaps. Select any image for a keyboard-accessible full-size view.

05 / PROOF OF LEARNINGTHE RECORDED XOM CHAIN

A system that reviews its own calls.

The record contains one complete, dated reflective-learning chain: a simulated XOM call, reviewed against its own written thesis, generalized into persistent lesson L1, and cited by name in later monitoring notes. Written memory and repeatable retrieval — not model retraining, and not yet a changed new-entry decision.

  1. 012026-08-10Simulated thesis & decision

    BUY XOM — $6,000 paper fill at $159.79 on a live supply catalyst, entered with a falsifiable expectation (≥$170 within 4 weeks) and an invalidation fixed at entry.

  2. 022026-09-08Retrospective review

    The catalyst played out — crude rose ~12% — yet XOM moved just +0.45%. Verdict: reasoning right, outcome lagging; a timing and magnitude miss, not a broken thesis.

  3. 032026-09-08Lesson L1 recorded

    Written to LESSONS.md: integrated majors lag pure upstream producers on a crude-spike catalyst — with concrete sizing and timeframe guidance for future calls.

  4. 042026-09-09 →L1 cited in later work

    Six subsequent journal entries cite L1 by name during position monitoring. The record shows the lesson informing later research — not yet changing a brand-new entry decision.

PHASE 0 / SANITIZED CAPTURE · ESC TO CLOSE
Phase 0 learning page showing the five-step XOM case study: the call, the outcome, the retrospective, lesson L1, and six later journal entries citing L1.
The learning loop as recorded: call → outcome → retrospective → lesson → later citations, every step linked to its original dated record. VIEW FULL SIZE ↗
FULL XOM CASE STUDY ON THE DEVELOPMENT PAGE →
06 / FROM PHASE 0 TO PHASE 1THE ARCHITECTURE BRIDGE

The agent works. Now the guardrails.

PHASE 0 · OPERATIONAL

Autonomous paper-research loop

The working system shown on this page: scheduled research, simulated decisions, journals, reviews, lessons.

PHASE 1 · IN DEVELOPMENT

Independent risk & approval controls

Deterministic limits and human authorization between any AI proposal and any order — separate from the model.

PHASE 1 · IN DEVELOPMENT

Robinhood MCP brokerage adapter

Permissioned broker connectivity for one strategy first. Under development — not a live capability.

PHASE 1 · TARGET

Audit & monitoring

Order-level logs, reconciliation, alerts, and verified safe-shutdown behavior before any limited live testing.

SCOPE, STATED PLAINLY

HWA Engine is founder-controlled research and development. All Phase 0 trades are simulated. There is no public brokerage service, no financial advice, and no asset-management offering — and no claim of completed live broker execution.

07 / CONTACT

Let’s talk systems.

For research collaboration, technical partnerships, or company inquiries, contact Hwa Gallery & Co. directly.

elaina@hwagallery.com ↗