Day 28: SapphireInvestments

Day 28 of 30 — the Investing arc closes

SapphireInvestments is live today at sapphireinvestments.io, built with Jim Rozich as operator-validated co-founder and an advisory board including Caroline Cooley (Summit TX Capital, Dallas; former Goldman Sachs VP, 20+ years long-short equity portfolio management) and Adam Benson (Minneapolis; $3B small-cap fund manager). The product: an AI-powered quantitative-signal platform for self-directed retail investors, with five pluggable strategy engines, paper-trading-validated performance disclosed as paper-trading, regulator-correct architecture (Sapphire never holds client funds; Alpaca holds the assets), and an accredited-investor-gated autonomous-execution tier shipping Q3 2026.

Day 28 afternoon closes the Investing arc. Sapphire is Jim Rozich’s quantitative-signal vision built as a multi-strategy AI engine, with the regulatory architecture matched to the product reality claim-for-claim.

The companion morning launch on Day 28 is ConnoisseurPlatform / Unshucked — Untappd for oysters, also operator-validated co-founder pattern (Kirby Winters as GTM Lead), also IRL-Premium-arc-closer. The double-launch architecture demonstrates the Velocity Process thesis: when AI compresses build time enough, one launch day can carry two ventures in two arcs, one founder can co-found multiple ventures across different verticals, and the audience can see both proof points in the same news cycle.

The problem

A self-directed retail investor today sits between two bad options.

On one side: the robo-advisors. Betterment, Wealthfront, M1 Finance, and the dozens of platforms that built around the 2010s automated-allocation thesis. They allocate, rebalance, tax-loss harvest, and target a fee floor. They are explicitly not alpha-generation products — the robo-advisor model is custody-plus-allocation-plus-the-median-return, and that is fine for the audience whose job-to-be-done is “I want to set this and forget it.” But it is the wrong product for the audience who pays attention.

On the other side: the institutional quant systems. The systems that actually find asymmetry — the long/short equity stat-arb, the trend-following CTAs, the regime-switching multi-strategy hedge funds — are walled off behind two-and-twenty fees, accredited-investor minimums, lockups, and institutional plumbing. The systems are good. The access is restricted by design. Most of the audience who would use them cannot.

The middle is where the audience lives. Self-directed investors who do the work — read filings, follow technicals, track sector rotation, want active signal generation — but who are not running $50M and do not want a 2-and-20 fee for the privilege of accessing the institutional toolkit. The middle has no real option today. Pattern day-trading platforms surface raw data without strategy framing. SeekingAlpha and Trading Central are publisher-side commentary without strategy attribution, confidence scoring, or risk parameters baked into every recommendation. TradingView gives you tools but no signals. The middle is empty.

Sapphire is built for that middle. Five strategy engines, signal stream with confidence scoring and risk parameters, paper-trading-validated performance with explicit disclosures, accredited-investor-gated autonomous execution on the roadmap, Alpaca on the custody side so Sapphire never touches client funds. The competitive narrative compressed: institutional-grade quantitative signals, retail-accessible price, regulator-correct architecture.

What Sapphire does — five strategy engines, one signal stream

The signal engine is a Python service (src/signal-engine/) implementing a shared BaseStrategy interface:

class BaseStrategy(ABC):
    @abstractmethod
    def generate_signals(self, market_data: MarketData) -> list[Signal]: ...
    @abstractmethod
    def backtest(self, historical_data: HistoricalData, params: dict) -> BacktestResult: ...

Five strategies plug into that interface today:

→ TrendFollower — momentum and trend signals across thousands of equities, with confidence scoring and regime context. Built on the classical trend-following literature, adapted to the multi-timeframe + AI-confidence-scoring shape the platform requires.

→ MomentumScanner — short-window momentum patterns scored against fundamental data and sentiment indicators. The cross-pollination of technical momentum and fundamental quality is the differentiator vs. pure-momentum scanners.

→ MeanReversionEngine — statistical mean-reversion candidates with risk parameters baked into every signal. Stat-arb shape, pairs-and-clusters logic, regime-aware entry / exit boundaries.

→ VolatilityRegime — regime detection across rolling volatility windows so every signal carries regime context, not just a point prediction. The “is the market in a high-vol or low-vol regime today” question is answered by the engine and routed to the other strategies as context.

→ FundamentalScreen — fundamental data overlaid on technical signals so quality screens hard-prune the universe before signals fire. Earnings quality, balance-sheet strength, accruals quality, sector context.

The architecture is operator-explicit: each strategy is a module, the interface is published, new strategies plug in, old strategies retire when they stop pulling weight. Sapphire’s audit-chain-by-construction discipline applies here — every signal is logged with strategy attribution, confidence score, and risk parameters, so backtesting and post-hoc analysis are reproducible by construction.

The architecture — why it sits on the right side of the regulatory line

The Investing vertical is the highest-scrutiny vertical in the run, and the architectural decisions are the load-bearing claim. Five structural elements:

1. Sapphire never holds client funds. Capital remains in the user’s personal brokerage account at Alpaca — a FINRA member, SIPC-insured. SIPC coverage extends up to $500,000 per account (of which $250,000 may be cash claims). Sapphire produces signals; Alpaca holds the assets. The custody question never sits inside Sapphire. This is the structural reason Sapphire can operate as a signal-generation platform without registering as a broker-dealer.

2. Autonomous execution is gated on accredited-investor verification. Per the live landing page: “Autonomous execution is available only to accredited investors as defined by SEC Rule 501 of Regulation D. Non-accredited users must place all trades manually through their own brokerage account.” The non-accredited audience gets signals; the accredited audience gets the option of letting the platform place orders on their behalf through their connected Alpaca account. The platform respects the regulatory distinction by construction.

3. Hard drawdown stops at the platform level. Drawdown limits, dynamic position sizing, correlation monitoring, portfolio-level risk budgets — every guardrail sits at the platform, not at the user’s discretion. The user can configure risk parameters within published ranges; the user cannot override the hard floor. Same pattern as the Day 19 Planwright audit-chain-by-construction discipline and the Day 20 CrewSheet platform-enforced workflow safety.

4. All signals and trade recommendations logged and archived (7-year retention). SEC Rule 204-2 sets the seven-year retention floor for investment-adviser-class records. Sapphire’s data layer enforces the retention by construction — the audit log is part of the data architecture, not a feature toggle.

5. Sapphire is not a registered investment advisor. Explicit footer disclosure: “Sapphire Investments is not a registered investment advisor. The information provided is for informational purposes only and should not be considered as investment advice. All trade recommendations are generated by automated systems and may not be appropriate for your individual financial situation.” The platform is on the publisher side of the Investment Advisers Act of 1940 distinction — same regulatory shape platforms like SeekingAlpha and Trading Central operate under. When (and if) the product evolves to a configuration where individualized advice is being given, the registration question gets answered. Today, the architecture is the regulator-correct shape for the product Sapphire actually is.

Performance posture — honestly framed

The landing page’s hero stats bar shows Sharpe ratio, win rate, signal count, and max drawdown. Every number is qualified — twice — as paper-trading performance since July 2025. The disclosures appear in three places on the live page:

“Paper trading performance since July 2025. All signals logged and archived. Past performance does not guarantee future results.”
“Performance shown is from paper trading and simulated execution. Actual results may differ due to slippage, fees, and market conditions. Past performance is not indicative of future results.”
“Trading involves substantial risk of loss and is not suitable for all investors. Past performance, whether actual or indicated by backtesting, is not indicative of future results.”

The 10-year backtested results in the strategies section are presented as backtests, with walk-forward optimization and out-of-sample validation named as the validation methodology. This is the floor of regulatory disclosure framing in the Investing vertical, not the ceiling — and the discipline matters more here than in any other category this run has shipped.

Operator-validated co-founder pattern

Jim Rozich — Founder & CEO of WheelUp Sales, quantitative-investing practitioner. Jim brought the product conviction (institutional toolkit accessible to the audience that does the work to use it), the strategy methodology, and the regulator-correct architectural framing that made Sapphire’s middle-ground positioning structurally possible. His role at Sapphire is named here only with his explicit consent; the Investing-vertical scrutiny on “founder” framing is high and the title used must match the on-paper relationship.

Advisory Board:

→ Caroline Cooley (Summit TX Capital, Dallas) — 20+ years long-short equity portfolio management. Former VP at Goldman Sachs. Deep expertise in risk management and institutional-grade signal development.

→ Adam Benson (Minneapolis) — $3B small-cap fund manager. Brings small-cap value and growth-at-a-reasonable-price perspectives to the FundamentalScreen strategy and the broader strategy roadmap.

The operator-validated-founder pattern lands harder in the Investing vertical than anywhere else in the run because the audience does the highest level of diligence on the people behind the system. PIM’s Day 27 launch leaned on Karen Rands’s three-decade angel-investing track record as the credibility wedge; Sapphire’s Day 28 launch leans on Jim Rozich’s practitioner conviction plus Caroline’s institutional risk-management track record plus Adam’s small-cap quant pedigree as the multi-source credibility wedge. Same Theme #5 (Investing — who gets funded changes), different shape.

Pricing — published, regulator-aware, and explicitly NOT the “$25K minimum” framing

The actual launch pricing on sapphireinvestments.io:

Tier	Price	What it gets you	Execution
Standard	$99/month	Signal stream across all five strategy engines, dashboard access, risk-management dashboard, backtest explorer (Coming Soon)	Manual through your own Alpaca account
Premium	$299/month	Standard + accredited-investor verification track + advanced risk-management features + prioritized signal delivery	Manual through your own Alpaca account; verification opens the path to autonomous execution when Q3 2026 ships
Autonomous	Coming Q3 2026	Premium + fully automated trade execution through connected Alpaca account	Accredited-investor-gated per SEC Rule 501 of Reg D — non-accredited users are not eligible at any price

The “$25K minimum on Premium” framing referenced in some earlier planning surfaces (the SEQUENCING.md Featured-ventures table item 11, the Day 18 → Day 28 CONTENT-CALENDAR row) was an aspirational price point from an older business-case document. The actual launch pricing is $99 / $299 / Q3 2026 autonomous. The earlier framing is being corrected across every launch surface as part of the Day 28 cross-source sweep.

The Velocity Process notes

What Claude Code handled. The Next.js 14 frontend (React + TypeScript + Tailwind + shadcn/ui + Recharts) with the hero stats bar, the strategies section, the performance chart, the pricing tier comparison, the advisory-board / institutional-experience block, the FAQ accordion, the footer disclosures, and the HubSpot email-capture form on the CTAs; the Python signal-engine service (FastAPI + pandas + numpy + scikit-learn) implementing the BaseStrategy interface across the five strategy modules (TrendFollower, MomentumScanner, MeanReversionEngine, VolatilityRegime, FundamentalScreen); the AWS five-stack CDK using the VelocityStack base class with all seven mandatory tags (Project / Client / Environment / BillingCode / ManagedBy / DeployedBy / CostCenter) — Data layer (VPC, RDS, TimescaleDB pattern, ElastiCache Redis, S3, Secrets Manager), API layer (API Gateway, Cognito with MFA, Lambda), Signal Engine layer (ECS Fargate, ECR, EventBridge, CloudWatch alarms), Frontend (S3, CloudFront, Route53, ACM), Budget & Cost Controls; the Alpaca paper-trading integration; the Polygon.io market-data ingest with the REST + WebSocket dual-protocol approach; the Claude API integration with the prompt-caching discipline (90% input cost reduction on system prompts), the Batch API routing for end-of-day analysis (50% discount), the REALTIME / BATCH-ELIGIBLE tagging on call sites, the explicit max_tokens limits, and the Sonnet 4.5 / Haiku 4.5 routing by task class.

What required human judgement. The decision to ship signals-only at launch (Phase 1) and explicitly future-state the autonomous-execution tier (“Coming Q3 2026”) rather than implying autonomous is live; the call to brand the launch around the regulator-correct architecture (Alpaca custody, accredited-investor gating, “not a registered investment advisor” disclosure) rather than around the performance numbers, because performance claims in the Investing vertical land differently than in any other category and the disclosure floor is the foundation; the discipline on the pricing — $99 / $299 / Coming Q3 2026 reflects the actual product state at launch and explicitly corrects the aspirational “$25K minimum on Premium” framing from older planning documents; the operator-and-advisor naming sequence (Jim first, Caroline + Adam as advisors, named only with explicit consent given the Investing-vertical scrutiny on founder / advisor naming); the doubleheader architecture with the morning Connoisseur launch to demonstrate the Velocity Process portfolio thesis (one launch day can carry two ventures in two arcs, when build time stops being the bottleneck).

What the readiness pass caught. Four items the Day 28 readiness audit (gtm/DAY-28-READINESS-AUDIT.md) surfaced and the launch addressed before going live. (1) The landing-page “What our investors say” testimonial section attributed quotes that appeared to be from advisory board members (Caroline Cooley, Adam Benson) under language that implied they were customer subscribers, including one quote with a “since subscribing” line and a “40% alpha generation” customer-performance claim. The pre-launch fix re-labels the section “What our advisors say” with the disclosure line “Sapphire Investments advisory board members. Quotes reflect personal use and observations; advisors are not paid for endorsements” — same fabrication-class catch-and-fix discipline as the Day 21 PatentFlow “200+ Law Firms / SOC 2 Type II” precedent, scaled up for the Investing-vertical regulatory exposure (SEC Marketing Rule 206(4)-1 treats customer testimonials and promoter testimonials under different disclosure regimes). (2) The “$25K minimum on Premium” framing in older sequencing notes did not match the actual launch pricing ($99 / $299 / Coming Q3 2026 autonomous) — the launch surfaces are updated to reflect the actual pricing, and the LinkedIn post explicitly corrects the framing. (3) The sapphireinvestments.com domain is parked on GoDaddy (a parking page redirecting to a for-sale lander) — the launch domain is sapphireinvestments.io, not .com, and the launch copy / cross-source sweep correctly use .io. (4) The Phase 0 / Phase 1 / Phase 2 phase labels were clarified to match the live state — Sapphire ships signals-only today (Phase 1), autonomous execution is explicitly future-state (Phase 2, Coming Q3 2026), paper-trading performance since July 2025 is the labeled performance window.

What I’d do differently. Acquire sapphireinvestments.com alongside .io and stand both up identically from day one. The current state — .io live, .com parked on GoDaddy — is fine for the cohort audience (who will type the link from the launch post) but creates an SEO-and-direct-navigation gap that will pull discovery traffic to the parking page until the .com is acquired. Same shape as the Day 14 / Day 20 / Day 21 stale-domain-sweep precedents in earlier launches. Pre-launch readiness checklists for future Investing-vertical builds should include “acquire all reasonable TLD variants of the company name” as a Day 0 task.

What’s next this week

Day 28 morning (today, 9:00 AM ET): ConnoisseurPlatform / Unshucked launched — Untappd for oysters, Kirby Winters as GTM Lead. See /builds/day-28-connoisseurplatform.
Day 28 afternoon (today, 1:00 PM ET): SapphireInvestments launches — five strategy engines, $99 / $299 pricing, regulator-correct architecture, Jim Rozich + advisory board.
Day 29 (Fri May 29): the 30-day retrospective. TFTSL Week 4 wrap drops. The Investing-arc pair (Day 27 PIM + Day 28 Sapphire), the IRL-Premium-arc closer (Day 28 Connoisseur), and the doubleheader launch-day architecture itself all get pulled into the retrospective narrative.
Day 30 (Sat May 30): Silverback Ventures public launch.
Phase 1 design-partner cohort: Standard tier signups, Premium tier accredited-investor-verification onboarding, signal-engine telemetry tuning, feedback-driven strategy weighting adjustments. Goal: validate the engagement metrics and the conversion funnel before Phase 2 (autonomous execution) ships.
Phase 2 readiness (Q3 2026): finalize accredited-investor verification flow, complete the Alpaca broker-API integration on the production path (not just paper trading), ship the autonomous-execution tier with the hard drawdown stops, position-sizing logic, correlation monitoring, and audit-log discipline in production. Pricing on the Phase 2 tier settles after Phase 1 cohort feedback.
Roadmap, named honestly as roadmap: additional strategy modules (the CLAUDE.md lists ShortSideResearch as the sixth strategy, not yet shipped); RIA registration if and when the product evolves into individualized investment advice (today it is publisher-side signal generation, not advisor-side individualized advice); additional broker integrations beyond Alpaca; Bedrock-routed inference path (the regulated-vertical AI-architecture pattern from Day 10 CounselExpress and Day 21 PatentFlow); 1,000+ subscriber threshold for Phase 2 graduation per the CLAUDE.md development sequence.

Want to talk

If you are a self-directed retail investor — open sapphireinvestments.io, browse the strategy methodology, read the disclosures, and sign up for the Standard tier ($99/month) if the signal stream is what you have been looking for.

If you are an accredited investor — the Premium tier ($299/month) opens the verification track and the path to the Q3 2026 autonomous-execution tier. Reach out via the contact surface on sapphireinvestments.io for the verification workflow.

If you are a quantitative practitioner or fund manager — the architecture is a public reference point. The five-strategy pluggable design, the audit-chain-by-construction retention, the accredited-investor-gated autonomous tier, the Alpaca custody architecture, the Claude-routed AI layer with prompt-caching and Batch-API cost discipline — every decision is documented in the CTO assessment. DM Todd on LinkedIn to talk about the architecture or partnership conversations.

If you are watching the Investing arc thesis — Day 27 PIM (Karen Rands) and Day 28 SapphireInvestments (Jim Rozich) are the two proof points of Theme #5: when AI compresses time-to-validation, the funding decision and the signal-generation problem both change shape. Different operators, different methodologies, same vertical, same operator-validated-founder pattern. The Day 29 retrospective pulls both together.

silverbackcto.com/builds for the daily.