49.28% win rate · +0.021R expectancy · PF 1.13 · +$6,494.60 simulated net.
Technical acquisition brief
Two transferable Python software assets
Working systems.
Clear scope.
Honest limitations.
Review the real interfaces, repository structure, verification receipts, transfer scope, and known limitations before deciding whether either asset fits your acquisition thesis.
Pre-revenue status disclosedTests and transfer scope documentedPrivate source withheldWritten diligence welcome
01 / AEGIS
A governed quantitative-research laboratory with a preserved experimental estate
Asking priceUSD 2,000negotiable
Evidence correction · 13 September 2026
The E4 BTC trend headline is withdrawn as performance evidence.
The preserved new replay contains a material execution-order defect: an intrabar stop can be followed by re-entry at that same bar's earlier open. The reported equity curve also measures realized trade events, not open-position marked-to-market risk.
The earlier 58.8% win-rate claim was incorrect. Both new-replay ledgers contain 22 profitable and 80 losing closed trades: 21.57%. These are diagnostics of a flawed simulation, not corrected strategy performance. The previous CAGR, profit-factor, mean-return, drawdown and positive-window headlines must not be relied on as trading-edge evidence. Original records remain preserved; no new valid result or live return is claimed.
22 / 102profitable / all closed trades in each new-replay ledger
16noncausal same-bar re-entries in each ledger
Not MTMreported equity omits open-position mark-to-market risk
NO-GOno validated trading edge or deployment authority
Recorded before the evidence correction. The walkthrough is retained to show the historical interface and research workflow. Its E4 performance and conservative-execution claims are superseded by the correction above; the video is not strategy validation.
Other historical research summaries · separate from E4Round 32,910 trades Later controlled training1,078 trades One-shot sealed 2025+ cohort140 trades
The remaining historical figures and test counts below have not been re-audited in this correction. They are not substitute alpha, current-code certification, or live returns. Source/version-bound verification is required before reliance.
+0.1043R expectancy and +$12,551 simulated net in the retained training span.
+0.0216R and +$298 overall; about 80% of the measured training edge decayed and results were thin and asset-dependent.
Far more than a strategy
Research Director plus Opportunity, Postmortem, and Self-Study labs.
Deterministic replay, risk and authority gates, paper/execution infrastructure, and human-approved rollback.
Preregistered hypotheses, sealed validation, hash-linked evidence, failed experiments, and postmortems.
A 113-asset, 12,035-trade breadth campaign, 222 included Binance datasets, and hundreds of MB of retained measurements.
Attractive backtest, rejected43.6% CAGRhistorical simulation
The cross-sectional momentum configuration also carried an 89.5% maximum drawdown, a negative median month, and catastrophic regime dependence. AEGIS rejected it. The system is designed to challenge impressive-looking results until risk, robustness, and real-world economics survive.
Final governed conclusionRESEARCH_NO_GO
Historical/simulated research only. No independently validated live trading edge, no live performance record, and no representation of future profitability. The final production/capital-certification conclusion remained NO-GO.
1,619tests passed
682subtests passed
113assets in breadth campaign
12,035preserved breadth trades
02 / IGNITION
Four sellable AI employees on top of a governed business operating system
Asking priceUSD 6,000negotiable
Ignition does not stop at a chatbot reply. It connects bounded AI employees to the machinery intended to find prospects, handle buying intent, provision a purchased employee after human payment verification, and keep the deployment visible after go-live.
Sellable AI employees
Try all four working demonstrations.
Each captures structured lead state, follows an industry workflow, and refuses to invent prices, bookings, guarantees, or diagnoses.
01AI employees
02Discover
03Outreach
04Classify replies
05Deal
06Human payment gate
07Deploy
08Maintain
Customer acquisition engine
Governed outreach that can slow down, suppress, reconcile, and escalate.
Campaign and vertical discovery with persistent search state, public-site validation, qualification, and deduplication.
Country/time-zone market windows, configurable daily limits, bounce/opt-out/decline suppression, and adaptive backpressure.
Exactly-once Gmail draft/send reconciliation, failure recovery, interested-reply prioritisation, and human follow-up queues.
Browser-assisted contact-form discovery and inspection, with unsafe or ambiguous actions left to a person.
Buying intent becomes a governed state.
READY_TO_BUY, PRICING, and CUSTOMIZATION replies can enter bounded package selection, onboarding capture, change-request handling, decline suppression, and progression toward PAYMENT_PENDING. Repeated ambiguity escalates to a person.
Payment authority stays human.
Only PAYMENT_VERIFIED can start onboarding checks, a tenant manifest, API-key and channel provisioning, and a website embed artifact. Missing configuration produces explicit BLOCKED evidence instead of a false success.
Go-live is not the end.
LIVE deployments can enter health sweeps, complaint and technical-issue triage, billing-question flags, blocked-deployment visibility, unresolved-event queues, and weekly operational reports.
Governed architecture
Semantic interpretation→Deterministic policy→Bounded writing→Validated action authority
The LLM can interpret and communicate. It does not own prices, business facts, permissions, bookings, diagnoses, payments, tenant boundaries, or consequential authority.
The counts below are retained historical seller-reported snapshot figures, not a fresh test of the current transfer package or production certification. Exact version, date and receipts must be checked during diligence.
1,754historically reported tests passed, 0 failed
21historically documented legacy skips
~98khistorically reported Python lines
4demonstrations linked above
Honest limitations
Pre-revenue with no paying-customer traction represented. Streamlit is demo-grade and SQLite is pilot-grade. Web chat works in the demos; WhatsApp and voice groundwork still requires provider evidence and is not certified end-to-end. Buyer accounts, provider configuration, and production hardening remain buyer responsibilities.
Continue only if there is a fit.
01Ask
Send focused technical, commercial, or transfer questions for a written response.
02Verify
Request a deeper sanitized review of the asset that matches your thesis.
03Offer
Make a written offer and identify the intended buyer entity and protected payment route.
Cloneable source, private repositories, credentials, operating data, and ownership transfer are not included in this preview. Transfer follows signed terms, agreed acceptance criteria, and protected cleared payment.
Reply in the platform conversation where you received this brief.