Sourcecard
Sourcecard turns a creator's podcast or YouTube archive into a searchable, timestamp-linked answer index their audience can self-serve instead of asking 'which episode covers this?'
One product, one founder, thirty days. Everything below is built to be executed inside the contest window — the official rules still decide what counts.
Decide in 20 seconds
The facts that decide whether this is worth thirty days of yours.
Solo podcast hosts and YouTube educators with 50+ published episodes and an existing paid offer (course, coaching, community), English-speaking, US/UK/CA/AU, who personally answer audience questions in DMs, comments or a Discord.
Market demand 4.1/10
Narrow the buyer to a segment with an urgent, already-budgeted problem and reference the specific trigger event that creates willingness to pay.
Why this could work in 30 days: one founder can ship the MVP in 5 days; the plan targets a first sale on day 8; 7 sales at $149 clears $1,000 of collected revenue; strongest signal is speed to first revenue at 8.5/10.
Score and projections come from the generated report and are not guarantees. The $100K Hackathon is run by Ship or Die — read the official rules before you build.
Your first 24 hours
11 concrete actions. Tick them off as you go — nothing is saved or sent.
Product opportunity and positioning
Target customer
Solo podcast hosts and YouTube educators with 50+ published episodes and an existing paid offer (course, coaching, community), English-speaking, US/UK/CA/AU, who personally answer audience questions in DMs, comments or a Discord.
Why now
Transcription and embedding costs have fallen to cents per episode hour, and a working preview of the buyer's own archive can be built before any money changes hands.
Problem
Their best answers are buried inside 60-300 hours of audio and video; the same 30-80 questions arrive weekly in comments and DMs, the creator answers manually, and the archive generates no inbound traffic.
Solution
Ingest the public RSS feed and YouTube transcripts, mine recurring questions from public comments and transcript lines, retrieve the strongest segment per question, and publish one hosted index page of 60-120 two-sentence answers, each deep-linked to the exact timestamp and to the creator's offer.
1. Product opportunity and positioning
Sourcecard sells a finished artifact, not a tool subscription. The buyer is a solo podcast host or YouTube educator with 50+ published episodes and an existing paid offer. Their problem is a specific, recurring chore: answering the same 30-80 audience questions in comments, DMs and Discord, week after week, while 60-300 hours of their own best thinking sits unsearchable in an audio archive.
The wedge works because the raw material is public. The RSS feed, the YouTube captions and the public comments are all available before a single conversation happens. The founder can run the pipeline against a real prospect's archive and arrive with a live URL: a searchable index of their own answers, each card linking to the exact minute they said it. The first message is not a pitch about what software could do; it is a demo of their own show, already built. That converts a cold DM into a claim decision.
The positioning is deliberately narrow. It is not a chatbot. Live conversational AI on a creator site has a well-known failure mode: it answers fluently, cites nothing, and nobody trusts it. Sourcecard inverts that. Output is a fixed page of reviewed cards, each anchored to a timestamped deep link. Trust comes from the citation, and the citation is the product.
Compared with a generic AI FAQ generator, the defensible parts are the pre-built preview, the citation check that rejects any card lacking a valid deep link, and the founder review queue that catches misquotes before a link is ever sent. Compared with doing nothing, the alternative is answering the same DM for the fourth time. The page is promotional to the host: it drives listeners back into episodes and pushes them toward the paid offer via a CTA button the host controls.
Commercial reality check: JEV scores this concept 61.5 (band C). Speed to first revenue is the strongest criterion (8.48 normalized) because the deliverable is buildable before the ask. Market demand (4.09) and monetization potential (4.58) are the weak points. A one-time $149 charge is the fastest path to verified profit inside 30 days, but it caps the ceiling: nothing recurring is sold in this window. The plan therefore treats the one-time claim as the wedge and the $299 Studio tier as the only in-window expansion, with recurring refresh revenue explicitly deferred past day 30.
Smallest sellable MVP
One hosted, searchable index page for a real show, built by the founder from that show's public RSS and YouTube data, with 60+ citation-locked answer cards and a Stripe claim link.
Must-have features
- RSS + YouTube ingester producing timestamped transcript segments
- Question miner over public YouTube comments and interrogative transcript lines
- Embedding retrieval plus two-sentence answer generation
- Hard citation gate: no card ships without a valid timestamp deep link
- Static per-show index page with keyword search and topic filters
- Stripe Checkout claim flow ($149 Claim, $299 Studio)
- Magic-link editor for answers, titles, CTA button and slug
- Embeddable iframe snippet and per-card shareable link with OG image
- Founder review queue with approve/regenerate/delete and noindex on unclaimed previews
Explicitly out of scope
- No live chat widget or conversational chatbot
- No automatic re-indexing when a new episode drops (founder-run refresh only)
- No analytics dashboard, lead CRM or email marketing
- No multi-language support beyond English
- No team seats, permissions or agency console
- No automated custom-domain provisioning, CNAME instructions only
- No free tier or trial
Demo flow (what the buyer sees)
- Founder pastes the show's RSS URL and YouTube channel handle
- Pipeline pulls episodes, captions and public comments
- Transcript segments are stored with start/end timestamps
- Questions are clustered and ranked by recurrence
- Best segment is retrieved per question and a two-sentence answer is generated
- Any card failing the timestamp check is discarded
- Founder reviews the queue and approves the page
- Page deploys to a unique preview URL, noindexed
- Host receives the URL, clicks claim, pays $149 via Stripe
- Host edits cards via magic link and embeds the snippet
Technical implementation plan
Stack
Architecture
A batch job, not a service. A founder-run CLI takes a feed URL, writes normalized episode and segment rows into Postgres, embeds segments, clusters candidate questions, generates answers with a retrieval step, and validates every timestamp against the segment range before writing a card. A Next.js build step then renders each show's approved cards into a static page served on a preview subdomain. The claim app is the only dynamic surface: magic-link auth, card editing, Stripe Checkout, and a webhook that flips claimed shows from noindex to index.
Build steps
- Scaffold Next.js app and Supabase project; create episodes, segments, questions, cards, shows tables
- Write RSS parser and YouTube caption fetcher; normalize to timestamped segments
- Add Deepgram fallback for feeds with no captions and store duration offsets
- Embed segments with OpenAI; store in pgvector with show_id and episode_id
- Mine questions from comments and interrogative transcript lines; cluster by embedding similarity
- Retrieve top segments per question and generate two-sentence answers
- Implement the citation gate: reject cards without a resolvable deep link into the segment range
- Build the static index page template with keyword search and topic filter
- Add Stripe Checkout, webhook, magic-link claim flow and card editor
- Add iframe embed snippet and OG image generation for shareable cards
- Build the founder review queue with approve/regenerate/delete and noindex default
- Deploy a first real preview and run the full flow end to end
AI components
- OpenAI embeddings for segment and question retrieval
- Small chat model for two-sentence answer generation
- Embedding clustering for question deduplication
- Speech-to-text for feeds lacking captions
Pricing and monetization
Claim
Entry offerFor Solo podcast hosts and YouTube educators with an existing paid offer
- Hosted searchable index with 60-120 cited cards
- Magic-link editor for answers, titles and CTA buttons
- Embeddable snippet
- Shareable per-card links with OG images
Studio
For Hosts with a bigger archive who want it branded and refreshed
- Everything in Claim
- Custom domain setup instructions and branding
- One founder-run refresh of the next 20 episodes
Model: One-time purchase per show: $149 to claim and keep the index live, $299 to add custom domain branding and one founder-run refresh.
First revenue offer: I already built your index from your last 120 episodes — here are the 3 questions your audience asks most, each linked to the minute you answered it: [URL]. $149 and it stays live under your CTA; otherwise I take it down in 14 days.
Expected ACV: $149
Upsell path: Studio upgrade sold to every Claim buyer within 72 hours of purchase, then a $49 per-quarter paid refresh of new episodes once the contest window closes.
First-customer acquisition
Cold DMs on X and Instagram to podcast hosts whose show already has a public RSS feed and 50+ episodes, each message carrying a live pre-built preview index of their own archive.
Where to find them
- Apple Podcasts and Spotify category pages for indie SaaS marketing shows
- YouTube search for educator channels with 50+ videos and comment-heavy uploads
- Listen Notes keyword search for endurance-training and running podcasts
- Freelance-business and consulting podcast directories
- Podcast guest-swap and indie-hacker communities on X
- Comment sections of shows whose hosts reply personally to viewers
Exact outreach motions
- Open with one screenshot card: the top recurring question plus the timestamped answer.
- Send the live preview URL and one sentence on what it is.
- Name the takedown window (14 days) so there is no ambiguity about consent.
- Follow up once at 72 hours with a second, different card.
- On reply, answer in one line and link straight to Stripe Checkout, no call.
- After payment, ask for one peer host intro in the confirmation email.
Message template
Hey {name} — I indexed your last {N} episodes. The question your audience asks most is "{question}", and here's the minute you already answered it: {deep_link}. All 74 cards are live here: {preview_url}. $149 and it stays up with your course link on every card, or I'll take it down in 14 days. No subscription.30-day execution roadmap
Week 1 · Build & validate · 7 days
Accounts and prospect list
- Create Vercel, Supabase, Stripe and Resend accounts; confirm Stripe is linked to TrustMRR before any charge.
- Build a 150-show target list across three niches with RSS URLs and episode counts.
- Scaffold the Next.js repo and deploy an empty skeleton.
Deliverable: 150-row prospect sheet plus a deployed skeleton.
Feeds to timestamped segments
- Parse RSS and pull YouTube captions via the Data API.
- Store normalized segments in Postgres with pgvector.
- Add Deepgram fallback for feeds with no captions.
Deliverable: Segment tables for three test shows.
Question mining and clustering
- Collect public YouTube comments and interrogative transcript lines.
- Embed and cluster into 60-120 candidate questions.
- Rank clusters by frequency across comments and transcripts.
Deliverable: Ranked question list per test show.
Citation-locked cards and index page
- Retrieve best segment per question; generate a two-sentence answer.
- Reject any card without a valid timestamp deep link.
- Render the static index page with search and topic filters.
Deliverable: One live preview URL for a real show.
Claim flow end to end
- Wire Stripe Checkout for $149 and $299.
- Send magic-link claim email via Resend.
- Ship the card editor and CTA settings.
Deliverable: Test purchase and claim completed in Stripe test mode.
20 previews built and reviewed
- Run the pipeline overnight on 20 shows.
- Review every card in the founder queue; delete weak ones.
- Set all unclaimed previews to noindex.
Deliverable: 20 reviewed preview URLs with 3 screenshot cards each.
First 20 DMs
- Send 20 DMs each carrying one screenshot card and the live URL.
- Log every send with a follow-up date.
- Draft the 72-hour follow-up with a different card.
Deliverable: 20 preview DMs sent and logged.
Week 2 · Launch & first sale · 7 days
Close the first claim
- Build 25 new previews.
- Send 25 DMs and every follow-up that is due.
- Reply to inbound within two hours.
Deliverable: First paid claim targeted; reply log current.
Message and consent
- Rewrite the DM opener based on stated reply reasons.
- A/B the first line across the next 50 sends.
- Unpublish previews older than 14 days.
Deliverable: Two DM variants and a reply-reason tally.
Embed and share polish
- Ship the embeddable iframe snippet.
- Generate OG images for shareable card links.
- Fix the top three claim or checkout complaints.
Deliverable: Embed snippet and OG cards live.
Niche two push
- Build 25 previews from niche two.
- Send 30 DMs.
- Run follow-ups on days 1-4 sends.
Deliverable: 45 cumulative DMs and 45 previews.
Convert warm repliers
- Offer a 15-minute walkthrough to every warm replier.
- Send the Studio upsell to anyone asking about domains or refresh.
- Ask each buyer for one peer host intro.
Deliverable: Second and third paid claims targeted.
Studio to existing buyers
- Email all Claim buyers the $299 Studio upgrade with domain steps.
- Deliver the first Studio refresh manually.
Deliverable: Upsell sent and one refresh delivered.
Funnel numbers
- Compute DM-to-reply and reply-to-claim rates.
- Drop the worst-performing niche.
- Set the week-three send quota.
Deliverable: One-page funnel table with next week's quota.
Week 3 · Repeat & systematize · 7 days
Raise outreach volume
- Build 30 previews.
- Send 35 DMs.
- Run second follow-ups on non-repliers.
Deliverable: 75 cumulative DMs.
Single-niche saturation
- Target 40 hosts in one niche with question-first lines.
- Log objections verbatim.
Deliverable: Niche-specific objection list.
Objections and takedowns
- Add an FAQ block answering the top three objections.
- Add a visible takedown link on every preview.
Deliverable: Updated claim page with takedown flow.
Preview production and refresh runbook
- Build 30 new previews.
- Send 30 DMs.
- Ship the manual refresh script for Studio.
Deliverable: Refresh runbook and 105 cumulative DMs.
Turn buyers into sources
- Add the referral ask to the claim confirmation email.
- Offer $50 credit per referred claim.
Deliverable: Referral line live in the claim email.
Empty every open thread
- Send a same-day close offer to all open threads.
- Follow up by voice or call with warm repliers.
Deliverable: Every open thread closed or marked dead.
Niche comparison and fresh list
- Compare claim rate by niche.
- Rebuild the target list with 100 fresh shows.
Deliverable: Fresh 100-show list.
Week 4 · Scale what works · 7 days
New segment, new previews
- Run the pipeline on 30 new shows.
- Send 35 DMs.
Deliverable: 140 cumulative DMs.
Studio for large archives
- Offer Studio to buyers with 150+ episodes.
- Deliver one custom domain setup end to end.
Deliverable: One Studio delivered with domain live.
Spot-check citations
- Spot-check 50 random cards against the audio.
- Fix any mis-timestamped card and log the failure mode.
Deliverable: Accuracy report at 98% or better.
Final broad send
- Build 30 new previews.
- Send 30 DMs.
- Follow up every thread older than 72 hours.
Deliverable: 170 cumulative DMs.
Reconcile revenue
- Reconcile Stripe payouts against claim count.
- Confirm every charge landed inside the contest window.
- Send receipts to all buyers.
Deliverable: Reconciled revenue sheet.
Harvest intros
- Chase every promised intro.
- Post one build-in-progress thread with three concrete results.
Deliverable: Five warm referred hosts in pipeline.
Warm second-degree list
- Send 40 DMs to the referred and warm list.
- Offer 48-hour delivery on any claim.
Deliverable: All final threads opened.
Days 29-30 · Consolidate · 2 days
Zero undelivered orders
- Deliver every outstanding claim and Studio job.
- Fix any broken embed or dead deep link.
Deliverable: No undelivered orders.
Freeze and report
- Freeze code after 12:00 UTC; ship nothing new.
- Export the verified profit sheet.
- Noindex every unclaimed preview.
Deliverable: Final verified profit export.
Revenue and profit scenarios
How the numbers are defined
Hackathon profit — Collected revenue − refunds − chargebacks − payment processing fees − paid marketing spend.
Business profit — Hackathon profit − operating costs (hosting, domains, tooling, AI/API delivery costs).
The contest ranks entrants on hackathon profit only. Hosting, tooling and AI/API delivery costs are not deducted by the organisers, but each sale still has to cover them — so we show both. Definitions verified on 2026-11-15 against the official rules.
Stated assumptions
- All figures are PROJECTIONS from stated assumptions, not verified revenue or existing customers.
- Blended ARPU of $149 in Conservative and Base; Upside assumes 26 Claim at $149 plus 6 Studio at $299 (32 orders, $5,668 collected, $177 effective ARPU).
- Payment fees modeled at a blended 4.5%: Stripe 2.9% + $0.30 plus roughly 1.5% international and currency surcharge.
- Ad spend is zero in Conservative and Base because the channel is organic DM outreach; Upside includes $150 of paid placement in a podcast-newsletter sponsorship.
- Operating costs cover Deepgram transcription (about $1.50 per 100-episode archive at pay-per-minute rates), OpenAI embeddings and generation (about $2-3 per archive), plus fixed monthly Vercel, Supabase, Resend and domain fees of $28-60.
- Refunds assume one $149 refund in Base and two in Upside; chargebacks assume one $149 dispute in Upside only.
- No customer, list size or reply rate below is verified; 400-600 preview DMs at a 5-7% claim rate is the hypothesis behind the Base case.
- Hackathon profit excludes hosting, tooling and inference; business profit subtracts operating_costs_usd.
- Every sale is priced above delivery cost: $149 against roughly $5 of transcription and inference.
Conservative
4 customers- Refunds
- $0
- Chargebacks
- $0
- Payment fees
- $27
- Paid marketing
- $0
- Hackathon profit
- $569
- Operating costs
- $40
- Business profit
- $529
Base case
15 customers- Refunds
- $149
- Chargebacks
- $0
- Payment fees
- $100
- Paid marketing
- $0
- Hackathon profit
- $1,986
- Operating costs
- $100
- Business profit
- $1,886
Optimistic
32 customers- Refunds
- $298
- Chargebacks
- $149
- Payment fees
- $255
- Paid marketing
- $150
- Hackathon profit
- $4,812
- Operating costs
- $220
- Business profit
- $4,592
Model your own numbers
Runs in your browser · nothing is saved- Collected revenue
- $2,235
- Payment fees (3.9%)
- $87
- Paid marketing
- $0
- Hackathon profit
- $2,148
- Business profit
- $2,048
Hypothetical arithmetic, not a forecast. The fee rate is a flat 3.9% assumption, refunds and chargebacks are left at zero here, and none of this is tracked data.
Scenario models from stated assumptions, not verified revenue. All figures are projections; no customers, traction or partnerships exist at launch.
Main risks and validation experiments
Risk 1X and Instagram DM rate limits and spam flagging throttle the only acquisition channel.
Diversify into email where a public contact address exists, cap sends at 25-30 per day per account, and keep message text varied and personalized.
Send 40 DMs across two accounts in 48 hours and record how many send attempts succeed versus get blocked.
If fewer than 60% of send attempts succeed for two consecutive days, pause and switch to email outreach before day 12.
Risk 2Unclaimed previews burn real inference and transcription money before any revenue arrives.
Cap to 20-25 previews per day, reuse embeddings, and purge unclaimed shows after 14 days.
Track actual pipeline cost per preview across the first 20 shows and compare it against the $149 price.
If cost per preview exceeds $8, reduce per-show episode depth from 120 to 60 episodes before scaling.
Risk 3Hosts object to their answers being republished and prefer to write their own FAQ.
Every preview carries a visible takedown link, is noindexed until claimed, and every card credits and links the original episode.
Log reply reasons verbatim on the first 100 DMs and count copyright or consent objections.
If more than 25% of replies cite republishing concerns, reposition to a private, host-only searchable index.
Risk 4Answer cards misquote or mis-timestamp, destroying the only trust asset the product has.
Hard citation gate in code plus a founder review queue that blocks deploy until every card is approved.
Spot-check 50 random cards against audio on day 24 and measure mismatch rate.
If accuracy falls below 95% on the spot-check, ship only cards under 20 seconds of audio and regenerate the rest.
Risk 5One-time pricing caps revenue and buyers ask for a free trial or a call instead of paying.
Lead with the finished preview so there is nothing to trial, and never offer a discount below $149 in the window.
Count how many replies request a trial or call in the first 100 DMs.
If more than 40% of positive replies demand a call, add a $49 self-serve single-episode index as a lower rung.
Hour-by-hour: the first 24 hours
Hours 1-2
- Create Vercel, Supabase, Stripe and Resend accounts.
- Connect Stripe to TrustMRR before any charge is possible.
- Confirm no customers or payments exist pre-launch.
Output: Working accounts with a clean payment ledger.
Hours 3-5
- Build the 150-show prospect sheet with RSS URLs, YouTube handles and episode counts across three niches.
- Mark the 40 shows with the highest public comment volume.
Output: Prospect sheet sorted by preview viability.
Hours 6-9
- Scaffold the Next.js app and Supabase schema for episodes, segments, questions and cards.
- Write the RSS parser and run it against five feeds.
Output: Parsed episode rows for five real shows.
Hours 10-14
- Add transcript fetching from YouTube captions and store timestamped segments.
- Embed segments and verify retrieval returns plausible passages.
Output: Segment table with a working similarity query.
Hours 15-18
- Write the question miner over comments and interrogative transcript lines.
- Run clustering on one show and manually score the top 20 questions.
Output: A reviewed question list for one real show, ready for answer generation.
Full 30-day launch report
Read the full generated report
Sourcecard — 30-Day Launch Report
1. Product opportunity and positioning
Sourcecard sells a finished page, not access to software. The buyer is a solo podcast host or YouTube educator with 50+ published episodes and an existing paid offer — a course, a coaching program, a paid community. They answer the same 30-80 audience questions in comments and DMs every week while 60-300 hours of their best material sits unsearchable in an audio archive.
The reason this is sellable inside 30 days is that the raw material is public. The RSS feed, the YouTube captions and the public comments exist before the founder makes contact. That means the first message is not a promise; it is a live URL of the prospect's own show, indexed, with three of their most-asked questions already answered and deep-linked to the exact minute. The sale becomes a claim decision: this exists, it is yours, take it down or keep it.
Positioning is deliberately anti-chatbot. Live conversational AI on a creator site answers fluently and cites nothing, and audiences do not trust it. Sourcecard is the inverse: a fixed, reviewed page of two-sentence answers where every card carries a clickable timestamp back to the source. The citation is both the trust mechanism and the differentiator against any generic AI FAQ generator.
One honest constraint: JEV scores the concept 61.5 (band C). Speed to first revenue is strong (8.48 normalized) because the deliverable can be built before the ask. Market demand (4.09) and monetization potential (4.58) are weak. A $149 one-time charge is the fastest route to verified profit but caps the ceiling, so the plan treats Claim as the wedge and Studio as the only in-window expansion. Recurring revenue is deferred past day 30 by design.
2. Smallest sellable MVP
The sellable unit is one hosted index page for one real show: 60-120 reviewed answer cards, keyword search, topic filters, a host-controlled CTA button, and a Stripe link that claims it. Everything else is optional.
What must exist on day 4: the ingester, the question miner, the citation gate, the static page renderer, the review queue, and Checkout. What must not exist: a live chat widget, an analytics dashboard, automatic re-indexing, team seats, or automated domain provisioning. The founder runs the pipeline by hand; that manual step is a feature for 30 days, because it is the quality control that makes the citations trustworthy.
3. Technical implementation plan
A Next.js app on Vercel with Supabase Postgres and pgvector. A founder-run batch job takes a feed URL and writes episode and segment rows, embeds segments with OpenAI, mines candidate questions from public YouTube comments and interrogative transcript lines, clusters them, retrieves the best segment per question, and generates a two-sentence answer.
The critical piece is the citation gate. Each generated card carries a candidate deep link into the source segment range; if the link does not resolve to a valid segment with a start and end timestamp, the card is discarded, never published. Deepgram or Whisper covers feeds without captions, at roughly $1.50 per 100-episode archive.
Rendering is static: approved cards become a page on a preview subdomain, noindexed by default. The dynamic surface is small — magic-link auth via Resend, a card editor, Stripe Checkout, and a webhook that flips a show from noindex to indexed on payment. The hardest technical risk is feeds with no captions or unusable audio, which can multiply per-show cost and time; those shows get skipped rather than rescued.
4. Pricing and monetization
Two one-time tiers. Claim is $149: hosted index, editor, embed snippet, shareable cards. Studio is $299: everything in Claim plus custom domain instructions, branding, and one founder-run refresh of the next 20 episodes. No free tier, no trial, no subscription inside the window.
The pricing logic is simple: $149 sits under a solo host's discretionary spend and does not require procurement. Delivery cost per sale — transcription plus inference — is roughly $5, so every sale clears cost with margin. The upsell path is Studio within 72 hours of a Claim purchase, then a paid quarterly refresh after the contest closes.
5. First-customer acquisition
The only channel is cold DMs on X and Instagram to hosts with a public RSS feed and 50+ episodes. Each message carries one screenshot card and one live preview URL, and states a 14-day takedown window. Follow up once at 72 hours with a different card. On reply, answer in one line and link straight to Checkout — no calls, no decks.
Week one target is 20 reviewed previews built and 20 DMs sent, with at least five replies logged. Every buyer is asked for one peer host intro in the confirmation email, which keeps the preview loop fed without paid acquisition.
6. 30-day execution roadmap
Days 1-4 build the pipeline and end with one live preview URL for a real show. Day 5 wires Stripe and the claim flow. Day 6 produces 20 reviewed previews. Day 7 sends the first 20 DMs.
Days 8-14 are the first sales push: 25 new previews and 25 DMs on day 8, message iteration on day 9, embed and OG polish on day 10, niche-two volume on day 11, a second close attempt on day 12, the first Studio upsell on day 13, and a funnel retro on day 14 that drops the worst niche.
Days 15-21 scale sends to roughly 105 cumulative DMs, saturate a single niche, ship the takedown flow and an objection FAQ, and empty every open thread on day 20.
Days 22-28 push a third niche, deliver one full Studio job, run a 50-card accuracy spot-check, reconcile Stripe against claim count, and harvest referral intros. Days 29-30 are delivery and freeze: every order shipped, code frozen at noon UTC, unclaimed previews noindexed, verified profit exported.
7. Revenue and profit scenarios
All three scenarios are projections built from stated assumptions. None represent existing customers or verified revenue.
Conservative: 4 claims at $149, $596 collected, $27 in processing fees, $0 ad spend, $40 operating cost. Base: 15 claims at $149, $2,235 collected, one $149 refund, $100 in fees, $100 operating cost. Upside: 32 orders at a $177 blended ARPU ($5,664 collected, 26 Claim plus 6 Studio), $298 refunds, one $149 chargeback, $255 in fees, $150 in paid placement, $220 operating cost.
The contest ranks collected revenue minus refunds, chargebacks, processing fees and paid marketing. Under that definition the Base case clears about $1,986 and the Upside about $4,812. Hosting, transcription and inference are not deducted from hackathon profit but do reduce business profit; every sale still clears delivery cost several times over.
8. Main risks and validation experiments
DM rate limits are the first risk; the test is 40 sends across two accounts in 48 hours, and the kill switch is a send success rate under 60% for two days. Unclaimed-preview burn is the second; track cost per preview across the first 20 shows and cut episodes per show from 120 to 60 if cost exceeds $8. Consent objections are the third; log reply reasons on the first 100 DMs and reposition to a private indexed archive if more than a quarter of replies raise republishing concerns.
Citation accuracy is the trust risk; a 50-card spot-check on day 24 with a 95% floor decides whether cards are shortened and regenerated. Finally, trial-seeking behaviour: if more than 40% of positive replies demand a call, add a $49 single-episode rung rather than discounting Claim.
9. First 24 hours
Hours 1-2: accounts created, Stripe connected to TrustMRR, ledger confirmed clean. Hours 3-5: the 150-show prospect sheet built and ranked by public comment volume. Hours 6-9: repo scaffolded, schema written, RSS parser proven on five feeds. Hours 10-14: captions fetched, segments embedded, similarity query returning plausible passages. Hours 15-18: question miner run on one real show with the top 20 questions manually scored — the first artifact that can become a preview URL tomorrow.
This brief was generated from public data about established startups, then narrowed and scored for a solo 30-day launch. Revenue and profit figures are scenario models, not verified results, guarantees or evidence of existing customers. Validate demand yourself before you spend a month on it — and check the contest rules for what counts toward your score.
Related ideas
All ideas →Flowproof Audit
Paste in an n8n or Make workflow export and get a scored reliability report plus drop-in patch nodes that stop silent failures, duplicate runs and untyped AI outputs before a client notices.
Refund Radar
A weekly email that tells small Shopify stores exactly which refund requests to deny, which to offer store credit on, and which to approve, based on their own written policy.
MockSmith
MockSmith turns a tutor's syllabus and past papers into a print-ready, blueprint-weighted mock exam with answer key and OMR sheet in minutes.