Frame
Niche behaviorStart from a repeatable interaction and a specific audience, then define a rights-safe original angle.
Case-backedBuild & LaunchMixedVaries
GeoSports Case Study: How a Shareable Web Game Was Built with Claude
GeoSports is a daily sports-geography game. Its founder reports assembling the first shareable version over a weekend with Claude and later reaching about 700K monthly active users. This source-backed case study separates Claude's reported role from the distribution, audience, and monetization factors that cannot be isolated.
What is the GeoSports Claude-assisted web game workflow?
Start with a narrow behavior and audience, write a bounded product spec with an AI assistant, ship a no-install daily loop, verify that people return and share, instrument the behavior, recruit aligned hosts, and add monetization only after repeat usage survives.
Case-backed workflow; revenue, traffic, subscribers, build time, and channel attribution are Founder-reported.
- 01. Use AI to turn domain knowledge into a written product spec before asking it to generate code.
- 02. Prefer a no-install web experience when the acquisition moment happens inside a feed or group chat.
- 03. Treat the share artifact and next-day return as product requirements, not post-launch marketing tasks.
- 04. Separate the interview-era founder stack from current services disclosed in the product's privacy policy.
Results and operating figures on this page are attributed to the cited source case. AI LinkBase has not independently audited them, and they are not a forecast for another product.
This case-backed workflow was selected for evidence quality and transferable operating value, not commission potential. The cited product and vendor links are direct official or primary-source links in this candidate. If AI LinkBase later introduces a tracked commercial link, it must be labeled, must not be presented as part of the founder's historical stack unless sourced, and must not change the recommendation or evidence status. Disclosure details
Core loop
Niche behavior → spec → web MVP → private share test → daily return → host expansion → monetization
Founder-named interview stack
Claude Max, Vercel, MapLibre, Upstash Redis, Google Sheets, and Stripe
Current disclosed additions
Neon, Resend, PostHog, X measurement, Playwire, and Google authentication
What not to copy
The revenue headline, audience size, build-time claim, content, pricing, or traffic attribution without local evidence
Prices reviewed Official Anthropic, Vercel, Upstash, Stripe, MapLibre, GeoSports, and U.S. App Store pages checked September 28, 2026; usage, tax, region, domain, ad, and store fees vary. · Full Stack reviewed September 28, 2026
Decision profile
The 8-Part Architecture Behind a Shareable Web Game
Eight connected capabilities move from a narrow behavior to staged monetization. The order matters more than any individual service.
In brief: frame the behavior, specify the loop, make sharing native, earn the next-day return, then add distribution and monetization.
Specify
AI-assisted product briefTranslate the interaction, data, states, edge cases, and launch boundary into a written spec.
Build
No-install web loopMake the first core action playable from the same surface where discovery happens.
Persist
Lightweight stateStore questions, scores, streaks, leaderboards, accounts, and payments with replaceable services.
Distribute
Share artifactTurn completion into a compact result people can post to feeds and group chats.
Return
Daily synchronized cadenceGive everyone the same bounded challenge and a clear next-session reason.
Expand
Audience-aligned hostsInvite credible participants to contribute within a reviewed template and reach adjacent audiences.
Monetize
Ads plus optional ProProtect the free habit loop and measure net revenue after fees, churn, and content cost.
Source case
What GeoSports Is—and Which Facts Are Verified
GeoSports is a daily sports-geography game created by sports storyteller Frank Michael Smith. In the direct founder interview, Smith describes studying MapTap's core interaction, writing a specification with Claude, building a web-first playable version over a weekend, privately testing whether people shared it, and then distributing it through his sports audience. The product later added daily hosts, advertising, subscriptions, and a broader GeoGames mobile app. This page reconstructs the operating sequence without treating AI-generated code, the founder's audience, or reported revenue as universally transferable proof.
In brief: the product is live; its business outcomes, build-time claim, traffic mix, and causal story remain Founder-reported.
- Product
- Daily sports-geography game with five shared questions, map taps, scoring, leaderboards, streaks, groups, and Pro features
- Best-supported web start
- May 3, 2026; exact first public-launch boundary remains unclear
- Mobile expansion
- GeoGames iOS version 1.0 released August 20, 2026
- Source publication
- September 27, 2026
- Reported operating snapshot
- About $40K monthly revenue, 700K MAU, 1,800 Pro subscribers, and $30K-$35K in programmatic ads
- Evidence boundary
- Business results, build time, traffic mix, and causal claims are Founder-reported
What the daily game loop looks like
A GeoSports prompt, map-based guess, score feedback, and next-question action—the compact loop this case study examines.
Editorial screenshot of the public GeoSports interface. GeoSports product content and interface remain the property of their respective owners; no affiliation or endorsement is implied.
Evidence
Reported Results: $40K/Month and 700K MAU (Founder-Reported)
Every figure keeps its evidence label so the historical case cannot be confused with an independently verified benchmark.
In brief: these figures are useful context, not an audited benchmark or a promise of reproducible results.
Monthly revenue
About $40K
Founder interview; exact period, currency, gross/net basis, tax, and fees are undisclosed.
Monthly active users
About 700K
Founder statement; calculation window and unique-user method are not public.
Pro subscribers
About 1,800
Founder statement; active/paying definition, price cohort, platform split, and churn are undisclosed.
Programmatic advertising
About $30K-$35K monthly
Founder statement; gross/net basis, partner share, fill rate, geography, and period are not public.
First-month reach
More than 1M unique players
Founder interview; separate founder post reports 366K plays and 150K peak DAU after the first week.
Current operation
Web product live and iOS 1.0.3 released September 15
Verified from first-party site and Apple platform records; not proof of business performance.
| Metric | Reported value | Context | Evidence |
|---|---|---|---|
| Monthly revenue | About $40K | Founder interview; exact period, currency, gross/net basis, tax, and fees are undisclosed. | Founder-reported |
| Monthly active users | About 700K | Founder statement; calculation window and unique-user method are not public. | Founder-reported |
| Pro subscribers | About 1,800 | Founder statement; active/paying definition, price cohort, platform split, and churn are undisclosed. | Founder-reported |
| Programmatic advertising | About $30K-$35K monthly | Founder statement; gross/net basis, partner share, fill rate, geography, and period are not public. | Founder-reported |
| First-month reach | More than 1M unique players | Founder interview; separate founder post reports 366K plays and 150K peak DAU after the first week. | Founder-reported |
| Current operation | Web product live and iOS 1.0.3 released September 15 | Verified from first-party site and Apple platform records; not proof of business performance. | Verified |
Cost control
Cost Snapshot
Fixed software is separated from variable usage, advertising, store fees, refunds, taxes, and labor. The largest cost is not necessarily a SaaS subscription.
Lean validation floor
EstimatedFrom about $40/month plus domain and usage
- ✓Claude Pro is $20 month-to-month and includes Claude Code under current public pricing.
- ✓Vercel Pro is $20/month with $20 of included usage credit.
- ✓MapLibre GL JS is open source, but map tiles, data, geocoding, or CDN usage may still cost money.
- ✓Upstash has a free entry tier; Google Sheets can serve as a temporary editorial queue.
- ✓Stripe has no standard monthly platform fee for basic U.S. card acceptance and charges per successful transaction.
Not included
Domain · AI overage · Map tiles/data · Database and email · Analytics · Content research · Security/legal review · Advertising and store fees
Founder-named higher-capacity build layer
EstimatedAt least $120/month before domain, usage, and transaction fees
- ✓Anthropic currently lists Claude Max from $100/month and offers 5x and 20x usage options.
- ✓The transcript names the 20x option, but the exact current checkout price and founder invoice were not captured.
- ✓Adding Vercel Pro creates a public-plan floor of at least $120/month before variable services.
Not included
Exact Max 20x price · Domain · Upstash overage · Map services · Payment fees · Current additional processors · Labor
Reported production operation
Founder-reportedNot disclosed
- ✓The interview does not publish invoices, ad-serving costs, database/email/analytics spend, content labor, contractor costs, refunds, or total operating expense.
- ✓Current first-party policy disclosures show the production stack is broader than the six founder-named interview tools.
Not included
Founder compensation · Ad-network share · Store commissions · Taxes · Chargebacks · Support · Moderation · Legal and insurance
Operating system
The 10-Step Workflow: From Spec to Retention
Keep the order: choose a narrow behavior and rights-safe content boundary, write the spec, define the smallest daily loop, build a no-install version, add measurement and operational controls, run a private share test, launch where the audience already exists, strengthen the share and return loops, recruit aligned hosts, and monetize only after retention and contribution economics are visible. AI may draft code and explain screenshots, but a human owner must approve architecture, security, data rights, tests, and releases.
In brief: validate the share and return loop before adding hosts, advertising, subscriptions, or a mobile app.
01
Choose a narrow behavior and rights boundary
Identify a repeatable interaction that a specific audience already understands without copying protected content or branding.
Components
Comparable products, user ritual, niche angle, content ownership, map/data licenses, trademark and trade-dress review
Output
A one-sentence concept and documented originality boundary
↓
02
Write the product spec with AI assistance
Turn domain knowledge into explicit states, rules, data, and acceptance criteria before generating code.
Components
User journey, question schema, scoring, edge cases, loading/error states, abuse cases, privacy needs, out-of-scope list
Output
A versioned implementation brief
↓
03
Define the smallest daily loop
Make one complete session valuable without an install, account, or tutorial where possible.
Components
Five-question round, one-tap guess, immediate feedback, score, completion state, next-round timing
Output
A thin end-to-end playable slice
↓
04
Build with a small transparent stack
Use replaceable services and keep AI-generated changes reviewable.
Components
Claude or another coding assistant, Vercel, MapLibre, lightweight question store, cache/state, payments only when needed
Output
A deployed preview with tests, rate limits, secrets controls, logging, and rollback
↓
05
Instrument trust and product health
Measure the loop before public traffic arrives and protect users from avoidable failures.
Components
Start/completion/share/return events, uptime, latency, errors, consent, account deletion, content corrections, moderation, backup restore
Output
A launch dashboard and incident checklist
↓
06
Run a private share test
Verify that people finish, understand, and voluntarily share the result without founder prompting.
Components
Small mixed-skill cohort, observation, completion, share rate, next-day return, qualitative confusion, failure log
Output
A go, revise, or stop decision
↓
07
Launch into the existing audience context
Remove handoff friction between discovery and first play.
Components
Web-first link, audience-native post/reply, no forced install, fast first paint, share preview, traffic cap and rollback
Output
A measured public cohort
↓
08
Strengthen the share and return loops
Make completion useful socially and give users a clear reason to return tomorrow.
Components
Compact score artifact, comparable daily round, leaderboard, streak, group context, reminder consent, cohort retention
Output
A repeatable distribution and retention loop
↓
09
Recruit audience-aligned hosts
Expand content and reach through people already demonstrating genuine interest.
Components
Engagement signal, host brief, question template, fact and rights review, scheduling, disclosure, post-round attribution
Output
A reviewed hosted edition and measurable adjacent cohort
↓
10
Monetize and scale only after evidence
Add revenue without destroying the free habit or confusing gross revenue with profit.
Components
Ad load, optional subscription, paywall, churn, net ad revenue, payment/store fees, content cost, privacy and accessibility review
Output
A contribution-margin decision and staged operating plan
Tool-by-tool guide
What Every Tool Does in This Workflow
Each component has one job. Original-case tools are kept separate from optional affiliate and expert-execution routes so the historical record stays accurate.
Claude / Claude Code
Specify and buildClaude can turn a detailed product idea into a spec, draft implementation, explain screenshots, and help debug bounded issues.
- ✓Draft and refine the product specification
- ✓Implement reviewable slices
- ✓Explain errors and propose tests without owning approval
Founder named regular Claude and a Claude Max 20x plan; transcript renders the name inconsistently.
Review Claude CodeVercel
DeployVercel hosts the web app, previews changes, and supports fast rollback and traffic-aware operations.
- ✓Deploy preview and production builds
- ✓Keep secrets out of client code
- ✓Monitor latency, errors, and usage
Founder-named and currently disclosed in the privacy policy.
Review Vercel pricingMapLibre GL JS
Render the gameMapLibre renders the interactive globe or map where players place guesses.
- ✓Render map/globe interaction
- ✓Support tap location and camera behavior
- ✓Keep attribution and data licensing correct
Founder-named; tile and geographic-data providers were not disclosed.
Review MapLibre docsUpstash Redis
Cache live stateUpstash Redis can hold fast-changing leaderboard, session, or statistics data.
- ✓Cache bounded game state
- ✓Protect hot paths with TTLs and rate limits
- ✓Budget for viral command volume
Founder-named and currently disclosed in the privacy policy.
Review Upstash RedisGoogle Sheets
Manage question contentA spreadsheet can act as a temporary reviewed queue for daily questions and answers.
- ✓Store question drafts and sources
- ✓Track reviewer and publish status
- ✓Prevent public or broad write access
Founder-named for questions; exact permissions and plan were not disclosed.
Review Google SheetsStripe
Collect web subscriptionsStripe can collect web payments and manage recurring subscription billing.
- ✓Create products and prices
- ✓Handle checkout and webhooks
- ✓Reconcile refunds, chargebacks, tax, and access state
Founder-named and current terms say GeoGames Pro is billed through Stripe on the web; mobile app purchases are a separate platform path.
Review Stripe pricingPostHog
Measure product behaviorPostHog can measure completion, sharing, return, and feature usage when configured with minimized data.
- ✓Define event names and denominators
- ✓Build retention and funnel cohorts
- ✓Exclude unnecessary personal or gameplay content
Named in the current privacy policy, not in the founder's interview-era six-tool summary.
Review PostHogPortfolio
Original Tools vs Current Portfolio
Original Tools preserve what the case reported. Current Portfolio reflects present workflow fit. A Required slot can remain empty when evidence is not strong enough to name a Current Primary.
Concept and rights briefFrameRequiredSelected
Define audience, interaction, originality, and content permissions
Original case
Sports remix of MapTapFounder domain expertise
Current Primary
Versioned product and rights brief
A comparable-product insight is useful only when the team documents what is original, licensed, or prohibited.
AI build assistantSpecify and buildRecommendedSelected
Turn a detailed brief into reviewable implementation
Original case
ClaudeClaude Max 20x
Current Primary
It matches the reported product family, but start with the lowest-capacity plan that fits measured workload.
Important condition: Human owner retains architecture, testing, security, and release approval.
Web deploymentBuildRequiredSelected
Preview, deploy, observe, and roll back the no-install product
Original case
Vercel ProCustom domain bought through Vercel
Current Primary
It matches the founder-named stack and current privacy disclosure.
Map interactionPlayRequiredSelected
Render the globe and collect guesses
Original case
MapLibre
Current Primary
It directly covers the disclosed interaction layer.
Important condition: Verify map style, tiles, location data, attribution, and usage rights separately.
Content operationsPublishRequiredUnder review
Review and schedule questions, answers, locations, and sources
Original case
Google Sheets
Current Primary
A spreadsheet is enough for a tiny reviewed queue; migrate when permissions, audit, localization, or volume outgrow it.
Fast stateOperateRecommendedSelected
Leaderboards, session state, and hot statistics
Original case
Upstash Redis
Current Primary
It matches the reported role and can start small, but viral traffic requires explicit budgets and fallbacks.
Durable dataOperateRequiredUnder review
Accounts, durable game history, purchases, and groups
Original case
Not disclosed in interview
Current Primary
Neon is named in the current privacy policy; the historical database and schema are unknown.
Product analyticsMeasureRequiredUnder review
Completion, share, and return cohorts
Original case
Not named in interview summary
Current Primary
It is currently disclosed, but a new team should choose only after defining the minimum event and privacy model.
Web paymentsMonetizeOptionalSelected
Subscription checkout and recurring billing
Original case
Stripe
Current Primary
Use after free-loop retention is measured; web and app-store entitlements need reconciliation.
AdvertisingMonetizeOptionalUnder review
Serve and measure ads without overwhelming the daily loop
Original case
Programmatic ads; partner not named in interview
Current Primary
Playwire is currently disclosed, but a new product should not add ad complexity before repeat usage, consent, and net economics are proven.
Replacement choices
Cheaper Alternatives and Their Trade-offs
These are comparison directions, not automatic recommendations. Replacing one component can change integrations, operating burden, attribution quality, or the economics of the whole workflow.
In brief: choose by operating burden, data rights, reliability, and economics—not by a tool's tutorial popularity.
AI build assistant
Lower-cost / direct path
Claude Pro or pay-as-you-go use for bounded tasks
Specialized path
Claude Max or another repository-focused coding agent for sustained sessions
Trade-off
Higher capacity reduces interruptions but does not replace tests, security review, or product judgment.
Deployment
Lower-cost / direct path
Vercel Hobby for a personal non-commercial prototype
Specialized path
Vercel Pro, Cloudflare, or another production platform with explicit observability and limits
Trade-off
Choose from traffic shape, team controls, runtime needs, and predictable overage—not tutorial convenience alone.
Question operations
Lower-cost / direct path
A permissioned spreadsheet
Specialized path
A small CMS or database-backed editorial workflow
Trade-off
Spreadsheets are fast but weak for granular roles, audit history, localization, and complex validation.
Leaderboard state
Lower-cost / direct path
A relational database with indexed score tables
Specialized path
Managed Redis for hot ranks, counters, TTLs, and rate limits
Trade-off
Adding a cache improves latency but creates consistency, eviction, recovery, and cost-control work.
Monetization
Lower-cost / direct path
No monetization during validation
Specialized path
Ads, web subscription, app-store purchases, sponsorships, or a blended model
Trade-off
Each revenue layer changes privacy, UX, reporting, fees, moderation, and entitlement complexity.
| Slot | Lower-cost / direct path | Specialized path | Replacement trade-off |
|---|---|---|---|
| AI build assistant | Claude Pro or pay-as-you-go use for bounded tasks | Claude Max or another repository-focused coding agent for sustained sessions | Higher capacity reduces interruptions but does not replace tests, security review, or product judgment. |
| Deployment | Vercel Hobby for a personal non-commercial prototype | Vercel Pro, Cloudflare, or another production platform with explicit observability and limits | Choose from traffic shape, team controls, runtime needs, and predictable overage—not tutorial convenience alone. |
| Question operations | A permissioned spreadsheet | A small CMS or database-backed editorial workflow | Spreadsheets are fast but weak for granular roles, audit history, localization, and complex validation. |
| Leaderboard state | A relational database with indexed score tables | Managed Redis for hot ranks, counters, TTLs, and rate limits | Adding a cache improves latency but creates consistency, eviction, recovery, and cost-control work. |
| Monetization | No monetization during validation | Ads, web subscription, app-store purchases, sponsorships, or a blended model | Each revenue layer changes privacy, UX, reporting, fees, moderation, and entitlement complexity. |
Transferability
What to Copy—and What Not to Assume
What is transferable
- ✓Translate domain expertise into a written spec before asking an AI assistant to write production code.
- ✓Choose the delivery surface that minimizes friction at the actual acquisition moment.
- ✓Test whether users voluntarily share before calling a feature a viral loop.
- ✓Make the next session predictable with a bounded daily cadence and comparable result.
- ✓Recruit contributors from demonstrated product engagement, then use a reviewed template and fact checks.
- ✓Instrument completion, sharing, return, net revenue, and operational cost before scaling traffic.
- ✓Keep the early stack small, but publish a current data-flow inventory as additional services appear.
- ✓Treat screenshots and AI explanations as debugging inputs, not proof that code is correct or secure.
What may not transfer
- !The founder had more than four million social followers and years of sports-content experience before GeoSports.
- !Sports trivia, daily games, and group-chat score sharing have category-specific social behavior.
- !A web game can remove install friction in ways that regulated, hardware-dependent, or deeply personalized products cannot.
- !The reported viral reply, media coverage, and host network may be difficult for a new founder to reproduce.
- !Advertising economics depend on geography, consent, session volume, viewability, fill, and network terms.
- !An eight- or fourteen-hour prototype says nothing about the time required for reliable accounts, billing, privacy, mobile apps, support, moderation, and content operations.
Guardrails
Risks and Stop Conditions
AI-generated code, sports and trivia rights, third-party map data, sudden traffic spikes, user accounts, leaderboards, advertising trackers, subscriptions, and a potentially young audience create security, privacy, reliability, consumer-protection, content, and platform-policy risks. Viral reach is not evidence of durable retention or profitable unit economics.
Critical
AI-generated code exposes secrets, authentication flaws, payment errors, or abusive endpoints.
Mitigation: Use least privilege, server-side secrets, tests, dependency scanning, rate limits, code review, preview environments, monitoring, and rollback before public traffic.
Critical
Questions, sports marks, map data, stories, or a comparable product's trade dress are copied without rights.
Mitigation: Create original content, keep source and license records, preserve map attribution, review trademarks and media rights, and maintain a correction/takedown path.
Critical
Advertising, analytics, login, and payment processors collect data without valid notice, consent, minimization, or deletion behavior.
Mitigation: Maintain a processor/data-flow register, region-aware consent, retention limits, access controls, tested deletion, vendor agreements, and special review for children or young audiences.
High
A traffic spike causes runaway bills, broken leaderboards, lost state, or an extended outage.
Mitigation: Load-test the hot path, set budgets and alerts, cap expensive work, use TTLs and queues, define degraded modes, back up durable data, and rehearse recovery.
High
Share and retention metrics are attributed to product quality when founder audience or one social platform drove exposure.
Mitigation: Separate cohorts, record referrers and dark-social uncertainty, compare assisted and unassisted users, and require repeat behavior after the initial spike.
High
Ads or subscriptions damage trust, accessibility, speed, or the daily habit.
Mitigation: Stage monetization behind retention evidence, cap ad load, label paid features, measure churn and performance guardrails, and keep cancellation and entitlement behavior testable.
High
Host-contributed questions introduce factual errors, undisclosed promotion, bias, or moderation incidents.
Mitigation: Use a written brief, independent sources, rights checks, review and correction queues, disclosure rules, and a kill switch for scheduled content.
Stop when: Stop or narrow the experiment when private users do not finish and share, next-day return stays below the team's declared threshold, content cannot be sourced reliably, rights or privacy review fails, acquisition depends on one founder account, infrastructure cannot survive a bounded load test, or net monetization does not cover content and operating cost without degrading trust.
Frequently Asked Questions
Did GeoSports really take only eight hours to build?+
The source title uses eight hours, but the transcript says the founder worked about eight hours on Sunday and continued Monday; another detailed account says about fourteen hours. The safest description is that a first shareable version was assembled over a weekend. The later production system, monetization, mobile app, policies, and operations took additional work.
Did Claude cause GeoSports to make about $40K per month?+
No causal conclusion is supported. Claude reportedly helped with the specification, code, troubleshooting, and implementation. The founder also had niche expertise, years of game experiments, a large content audience, social distribution, strong sharing behavior, recurring content, and later monetization systems.
What tools did the founder report using?+
The interview names Claude Max, Vercel Pro and a custom domain, MapLibre, Upstash Redis, Google Sheets, and Stripe. The current privacy policy separately names Vercel, Neon, Upstash, Stripe, Resend, PostHog, X, Playwire, and Google authentication. These snapshots should not be collapsed into one historical stack.
How much does a similar validation stack cost now?+
A current public-plan floor can start around $40/month with Claude Pro and Vercel Pro, before domain, map data, database, email, analytics, payment, legal, and labor costs. The founder named a higher-capacity Claude Max 20x option; Anthropic currently lists Max from $100/month, but the exact 20x checkout price and founder invoice were not verified.
Should every creator build a web app before a mobile app?+
No. Web-first is useful when discovery happens in a browser, feed, or group chat and the core action can work without device-specific capabilities. Native delivery may be better when notifications, offline use, sensors, app-store discovery, or platform-native purchases are essential. GeoSports itself later added the GeoGames mobile app.
Can a large audience be ignored because the founder said most traffic was word of mouth?+
No. The founder reports that direct posts contributed only a small portion of daily traffic, but a large audience, credibility, media relationships, and access to hosts may still have accelerated initial exposure. Treat the channel split as founder attribution and validate distribution with your own cohort data.
What is the first success gate for this workflow?+
A small private cohort should complete the core loop, understand the result, voluntarily share it, and return at a predeclared rate. If those behaviors do not appear without heavy prompting, revise the product before adding hosts, ads, subscriptions, or mobile apps.
Are the reported revenue and user figures verified?+
No. They are preserved as Founder-reported because no audited financial statement, reconciled processor export, analytics definition, or independent user ledger was published. Current website and App Store records verify operation, not the business outcomes.
Transparency
Sources & Methodology
Prepared September 28, 2026 from the direct Starter Story founder interview, current GeoSports product and policy pages, Apple App Store and lookup records, a dated founder launch post, and current official vendor documentation. The source publication date, prototype date, private/public launch window, interview-relative result period, mobile release, and current operating evidence are recorded separately. All revenue, traffic, subscriber, audience, attribution, and build-time claims remain Founder-reported unless explicitly marked Verified. The title's eight-hour claim is not repeated as a settled fact because the transcript says work continued the next day and another account says fourteen hours. Current vendor pricing is a reference scenario, not a reconstruction of founder invoices. The original stack and AI LinkBase's current recommendations remain separate. No private analytics, source repository, billing export, ad-network statement, processor reconciliation, security audit, or legal opinion was available.
Starter Story — direct founder interview ↗GeoSports — official website ↗GeoSports — how scoring and gameplay work ↗GeoSports — privacy policy ↗GeoSports — terms and conditions ↗Apple App Store — GeoGames ↗Apple lookup API — GeoGames platform record ↗Founder LinkedIn — first-week GeoSports report ↗Anthropic — Claude pricing ↗Vercel — pricing ↗MapLibre GL JS — official documentation ↗Upstash Redis — pricing ↗Google Sheets — official product page ↗Stripe — pricing ↗PostHog — pricing ↗
Source case: I built this website in 8 hours with Claude. Now it makes $40K/month, Starter Story, September 27, 2026. Short excerpts are used only where necessary; this page is an original workflow analysis.
Transferable lesson
The bottom line
GeoSports validates a sequence, not a tool: own a narrow behavior, write the specification, ship the smallest no-install loop, test voluntary sharing, then add hosts, advertising, and subscriptions only after retention is visible.
That sequence can transfer. The outcome cannot. Frank Michael Smith's audience of more than four million followers and years of sports-content experience were material distribution advantages that a new team should not assume it can reproduce.
Where to start: write the first operating spec before opening a code editor. It should define the core loop, scoring, failure states, content and rights boundary, and what is explicitly out of scope.