01 / 17

StockScore · showcase

StockScore

Score the business. Read the market. StockScore does both: a 100-point score for quality, value and moat, and a score-free read on the sentiment, IPOs and events around it.

100point score, fully shown2scoring engines: stocks & ETFs3score-free surfaces: sentiment · IPO · event

scroll, or press → to present

02 / 17the app

Meet StockScore

stockscore.app/

Search any company or ETF. Every card you see is live data.

03 / 17the product

One score, fully shown

stockscore.app/score/NVDA

Every score shows its working: the numbers behind it, and how complete the data is.

04 / 17methodology

Three pillars, one number

FUNDAMENTALS · 45

VALUATION · 30

MOAT · 25

45 / 100

Fundamentals

the compounder engine

Nine checks on growth, margins, balance sheet, and how well cash is reinvested.

30 / 100

Valuation

priced against its peers

Is it cheap or expensive next to other companies in the same sector?

✦ GRADED BY AI25 / 100

Economic moat

Hamilton's 7 Powers

How durable is the advantage? The one pillar an AI model helps grade.

100 points in all, split 45 / 30 / 25. Code scores the numbers; only the moat is graded with help from AI.

05 / 17coverage

A parallel engine for ETFs

stockscore.app/score/VWCE

ETFs get their own score, ranked by percentile against funds in the same category.

06 / 17depth

Go deeper than the score

stockscore.app/compare

stockscore.app/score/NVDA

the earnings read coversexecutive summaryrevenue & growthmargins & cash flowguidancerisks & red flags

Compare companies side by side, and read an AI breakdown of the latest earnings.

07 / 17signals & events

A context layer, without a score

DIRECTIONAL

Sentiment Radar

Seven signals across options, expectations and prediction markets — a direction and a confidence, never a number.

MECHANICS

IPO Radar

Range moves, deal lifecycle and S-1 facts for a company with no reported quarter to score yet.

EDITORIAL

Event Brief

A hand-curated, fully-sourced read on a marquee event — what is actually known.

NO 0–100 SCORE · NO GAUGEdirection lives on a blue ↔ violetaxis that never touches the score's emerald / amber / rose

Three surfaces add the context around the score: where the crowd leans, how a deal is pricing, and what is actually known. All deliberately score-free.

08 / 17sentiment radar

Where the crowd is leaning

stockscore.app/score/AAPL

the radar reads · 3 families · 7 signalsexpected moveput / callIV skewanalyst revisionsnews toneprediction markets

A clear direction on the blue ↔ violet axis, read from seven signals across options, expectations and prediction markets — and the same verdict rides beside each holding in your watchlist.

09 / 17ipo radar

A deal, before there's a quarter to score

stockscore.app/ipo/fcbm

the radar showsprice-range trackerdeal lifecycleS-1 structure factsrisk factorsscored peers

Track a deal as it prices: range moves, S-1 filings, and the lifecycle to listing.

10 / 17event brief

What is actually known

stockscore.app/events/atlas-cloud-ipo

the brief carriessourced factsdeal timelinebull / bearprediction-market snapshotwhat happens after

A hand-curated, fully sourced read on a marquee event, with both sides laid out for you to weigh.

Part IIhow it's built

11 / 17architecture

One request flow, four destinations

browserYOU

→

Next.js clientRENDER

→

API proxyTIMEOUTS · AUTH STRIP

→

FastAPIROUTE DISPATCH

→

cache / DBIN-PROCESS · POSTGRES

→

AI layerGEMINI · BOUNDED

↳ FastAPI dispatches by route — one path scores, three add score-free context

/scoreSCORE ENGINE · 100 PT

/sentiment-radarSENTIMENT ENGINE

/ipoDEAL PIPELINE · EDGAR

/events · /calendarCURATED CMS · FEED

The same proxy, cache and AI layer serve every route: one path scores, three add score-free context.

12 / 17data & caching

Many sources, cached in two layers

scoring sources → one record

yfinancePRIMARY FEED

Alpha Vantage · FinnhubFALLBACKS

StooqKEYLESS

justETF · API NinjasETF REFERENCE

signals & events sources · flag-gated

Options & short interestPOSITIONING

News headlinesFINNHUB · CLOUD-SAFE

PolymarketPREDICTION MARKETS · GAMMA

SEC EDGARS-1 / S-1-A FILINGS

Curated briefscontent/events/*.json

→

resolveSCORE RECORD + SURFACE PAYLOADS

→

two-tier cache · every route

hot cachein-process now · Redis-ready

↓ READ-THROUGH ↓

Cloud SQL Postgres90-day source of truth · scores

scores refresh when the rules change or earnings land; surface reads expire on their own TTLs, fail-open to “absent”

Scoring sources resolve to one record; the surfaces add their own (options, Polymarket, EDGAR, curated briefs), all cached, all fail-open.

13 / 17operating model

From intent to production

Intake — Jira & Notionwhere teams would raise work

Claude Coworkturns the idea into a clean ticket

Claude Designmockups, tokens & specs

∥ IN PARALLEL

GitHub issuelinks the ticket and the design together

Claude Codeone issue · one branch · one worktreeYOU APPROVE THE PLAN

develop → staging → mainpromoted through protected branchesYOU APPROVE THE RELEASE

needs you automated / designed-to

Cowork structures the ticket while Design supplies the mockups, in parallel, and only two steps need you: the plan, and the release.

14 / 17configuration

Configured for Claude Code

CLAUDE.md

Loaded every session — the project's standing instructions.

repo root

Branch model

issue-N → develop → staging → main.

WORKFLOW.md

Worktree isolation

One issue · one worktree — parallel sessions can't collide.

git rev-parse --git-dir

Six-block prompt

context · task · refs · constraints · success · halt.

issue contract

Autonomy + halt

Self-execute by default; escalate on risk.

CLAUDE.md

Testing gate

Failing-first · pre-commit · detect-secrets.

settings.json

A CLAUDE.md file and project settings define how every coding session behaves.

15 / 17de-risking

De-risked before any code

01Propose the approachin the Claude Project

→

02Challenge with a second modeltwo rounds

→

03Human decisionweighing the trade-offs

→

04Crystallize into a briefthen implement

Significant approaches are stress-tested against a second model — twice. The final decision is always human.

16 / 17cloud & ci/cd

Secured by default

Cloud RunBACKEND + FRONTEND

Cloud SQLPOSTGRES · VPC

RedisPLANNED · IN-PROCESS NOW

Secret ManagerRUNTIME SECRETS

GeminiFLASH BEFORE PRO

CI pipeline · WIF, no JSON keys

PR checks→

Gemini-call guard→

deploy workflow→

Cloud Run

No keys stored in the pipeline; every secret kept in a managed vault.

17 / 17the takeaway

Built to inform every decision

Every score shows its working. Where the number stops, a score-free read carries the context. Either way, you see the why and decide for yourself.

Score a stock →Get in touch →