Score C · 6.12/10Free to read$100K Hackathon · November 16, 2026 → December 16, 2026Nov 16, 2026

ClipLedger

Upload a folder of CCTV footage and a plain-English question, get back a timestamped, hash-anchored evidence log PDF with trimmed clips ready for a claim file.

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.

Speculative6.12/10
MVP build (solo)
5 days
Launch price
$149
Sales to $1,000 revenue
7 customers
First paying customer
Direct LinkedIn and email outreach to owners of small PI and claims-investigation firms sourced from public state and provincial investigator-association member directories.
Who pays

Owner or lead investigator at a 1-10 person private investigation / claims-investigation firm in the US, UK or Canada handling insurance-defence and premises-liability cases that arrive as raw CCTV files.

Biggest execution risk

Operational simplicity 3.8/10

Remove regulated data, human review steps and custom infrastructure; keep the product self-serve and stateless on managed services.

Why this could work in 30 days: one founder can ship the MVP in 5 days; the plan targets a first sale on day 12; 7 sales at $149 clears $1,000 of collected revenue; strongest signal is speed to first revenue at 7.8/10.

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

12 concrete actions. Tick them off as you go — nothing is saved or sent.

0/12
Entry price
$149
Solo build
5 days
30d base revenue
$1,788
30d hackathon profit
$1,716
Section 1 of 8

Product opportunity and positioning

Target customer

Owner or lead investigator at a 1-10 person private investigation / claims-investigation firm in the US, UK or Canada handling insurance-defence and premises-liability cases that arrive as raw CCTV files.

Why now

Vision models can caption low-quality retail and night footage cheaply enough that one case costs under $10 to process, which fits a $149 impulse-approval price for a firm that already holds the footage and has a deadline this week.

Problem

Each case lands as 10-40 hours of multi-camera footage with broken filenames and no index, so the reviewer burns 5-8 unpaid hours scrubbing and hand-building a timeline with screenshots and timecodes.

Solution

Drag-and-drop upload, ffmpeg scene-split and dedupe, VLM captions, natural-language search, a tick-list review screen, then a one-click ZIP containing a PDF evidence log (still, camera, exact timecode, description, source SHA-256) plus 15-second MP4 clips.

Differentiation · Not a surveillance platform and not a general video tool: the output is a document a claims adjuster or attorney accepts, with provenance hashing, priced per case instead of per camera seat.

1. Product opportunity and positioning

ClipLedger sells to a narrow, already-triggered buyer: the owner or lead investigator of a 1-10 person private investigation or claims-investigation firm. Their case volume arrives as raw CCTV exports from retailers, HOAs and property managers — 10-40 hours across several cameras, filenames like ch03_20260914_2231.mp4, no index, no notes. A defence-side claim has a deadline attached, and the reviewer's first job is unpaid triage: scrub at 8x, screenshot the moments that matter, type timecodes into a Word document. That is five to eight hours per case of billable-but-unbilled work, and it is the same job every case.

ClipLedger compresses that into an upload and a sentence. The investigator types "person in red hoodie enters, leaves with box", reviews the time-clustered hits, ticks the ones that matter, and exports a ZIP with a PDF evidence log, trimmed MP4 clips and a hash page. The differentiator is not the AI search — it is the artefact. Stills, camera, exact timecode, one-line description and a SHA-256 of the source file on the same page is what makes a log usable when attached to an insurer's file or an attorney's chronology. Generic transcription or summarisation tools do not produce it; enterprise video-analytics platforms produce it only at enterprise prices with enterprise sales cycles.

Why this is sellable inside 30 days: the buyer already possesses the footage, has already been paid or will be paid for the review, can approve $149 personally without procurement, and repeats the purchase per case. One firm that pays twice is worth two customers. The buyer is also reachable directly — state and provincial investigator associations publish member directories, and the owners answer their own email.

Positioning discipline matters. ClipLedger is first-pass triage, not final review and not a legal-admissibility claim. Every surface says so. That keeps the liability story honest, keeps a human as the final gate, and gives the reviewer a reason to trust the output rather than to fight it.

The honest weakness in this concept is operational: video processing, cost caps and cloud-upload resistance are real friction. The 30-day plan therefore front-loads one thing — a working end-to-end export on a synthetic clip — and then spends the remaining 25 days doing outreach and delivering real cases, with the founder manually assisting any step that breaks.

Section 2 of 8

Smallest sellable MVP

A paid, self-serve path where one uploaded folder of footage plus one text query returns a downloadable ZIP: PDF evidence log with stills, camera, timecodes and source SHA-256, plus trimmed MP4 clips — capped at 25 hours per case pack.

Must-have features

  • Presigned folder upload to object storage with per-case isolation
  • ffmpeg scene-change split with frame dedupe and preserved source timecodes
  • Batched VLM captioning of keyframes into a searchable caption table
  • Natural-language search that groups matching frames into time-clustered events
  • Tick-list review UI with frame preview and clip start/end adjustment
  • ZIP export: React-PDF evidence log, ffmpeg-trimmed MP4s, SHA-256 hash page
  • Stripe Checkout issuing case credits, processing unlocked after payment
  • Magic-link auth and 30-day auto-delete of uploaded source footage

Explicitly out of scope

  • Live RTSP/ONVIF streaming or real-time alerts
  • Face recognition or identity matching
  • Mobile app and multi-seat admin
  • Any claim of court admissibility or certification
  • Automatic redaction or bystander blurring
  • Human-in-the-loop review marketplace
  • Billing per camera or per seat

Demo flow (what the buyer sees)

  1. Founder uploads a synthetic 20-minute car-park clip (or the prospect's own folder link)
  2. Job splits scenes, extracts and dedupes keyframes, keeps source timecodes
  3. VLM captions each keyframe; captions land in Postgres full-text index
  4. Query "man in red hoodie loads a box" returns grouped hits with stills
  5. Reviewer ticks three hits and adjusts one clip boundary
  6. Export builds PDF log + 15-second MP4s + SHA-256 page, packaged as ZIP
  7. Buyer pays $149 via Stripe and the ZIP unlocks for download
Build plan

Technical implementation plan

Stack

Next.js on VercelSupabase (Postgres, auth, storage)Modal for ffmpeg worker jobsGemini Flash vision (fallback GPT-4o-mini) for frame captioningPostgres full-text search over captionsStripe CheckoutReact-PDF for the evidence logResend for magic links and job-status email

Architecture

Next.js handles auth, case records and the review UI. Uploads go browser-to-storage via presigned URLs so video never passes through the app server. A case row moves through queued, splitting, captioning, ready. Modal runs two worker functions: one calls ffmpeg to scene-split, extract frames and dedupe by perceptual hash, writing keyframes plus source timecodes back to Postgres; the other batches keyframes to the vision model and stores one-sentence captions. Search is a Postgres full-text query over captions scoped to a case, clustered by timestamp proximity into events. Export runs ffmpeg trims for ticked hits, streams a React-PDF document, hashes each source file with SHA-256, and zips everything to a signed download URL. Stripe Checkout writes a credit row; the worker refuses to start without one.

Build steps

  1. Create Supabase schema: cases, source_files, keyframes, captions, hits, credits.
  2. Add presigned folder upload and a per-case storage prefix with a 30-day expiry job.
  3. Write the Modal ffmpeg worker: scene detect, keyframe extract, phash dedupe, timecode map.
  4. Batch keyframes to the vision model and persist captions with camera + timecode.
  5. Add Postgres full-text search and cluster results into events by timestamp gap.
  6. Build the review UI: results list, frame preview, tick toggles, clip boundary sliders.
  7. Implement export: ffmpeg trims, React-PDF log, SHA-256 page, ZIP to signed URL.
  8. Wire Stripe Checkout to credit issuance and gate processing on credit balance.
  9. Add magic-link auth and per-case access control.
  10. Run a full end-to-end pass on a synthetic 20-minute clip and record the Loom.
  11. Add retries, job-status emails and a failure state that refunds the credit.
  12. Ship a landing page with pricing, a sample evidence log PDF and the Loom.

AI components

  • Vision-language model captioning of deduplicated keyframes (Gemini Flash, GPT-4o-mini fallback)
  • Caption normalisation prompt that forces object, clothing, action and direction-of-travel into one line
  • Optional query-expansion step that rewrites the investigator's sentence into 2-3 caption-style variants before search
Hardest technical risk: Search recall on low-light, low-bitrate footage: if the vision model captions 'a person' instead of 'a person in a red hoodie', the query misses and trust collapses on case one. Mitigation is a brightening pre-pass on dark frames and always returning the top-N nearest captures even below the confidence threshold.
Section 3 of 8

Pricing and monetization

Case Pack

Entry offer
$149 one-time

For Solo investigators and 2-5 person firms

  • Up to 25 hours of footage
  • Unlimited searches for 30 days
  • PDF evidence log with stills and timecodes
  • Trimmed MP4 clips and source-file hash page

Firm

$299 / month

For Firms running 3+ cases a month

  • 5 case packs per month
  • Shared case library
  • Branded PDF header on evidence logs

Model: Per-case one-time packs plus a monthly firm plan, both paid up front by card with no trial and no procurement.

First revenue offer: Send one case folder this week — I will return the evidence log in 24 hours, free, no account. If the log holds up with your adjuster, that case is $149 and your next one is $99 if paid within seven days.

Expected ACV: $248

Upsell path: Repeat Case Pack purchases at $149 per case, then Firm at $299/month once a firm clears three cases in a month; branded PDF headers and a shared case library are the upgrade triggers.

Target customers
30
Revenue model
$4,470
Costs
$320
Profit model
$4,150
Section 4 of 8

First-customer acquisition

Primary channel

Direct LinkedIn and email outreach to owners of small PI and claims-investigation firms sourced from public state and provincial investigator-association member directories.

Where to find them

  • Public member directories of state PI associations (e.g. state-level investigator associations listed by region)
  • Provincial investigator associations in Canada
  • LinkedIn search: 'private investigator' / 'claims investigator', company size 1-10
  • LinkedIn search: 'owner' at firms tagged insurance defence or surveillance
  • PI and claims-investigation LinkedIn and Facebook groups where practitioners ask workflow questions
  • Firms advertising 'insurance defence surveillance' and 'premises liability' services on their own websites

Exact outreach motions

  • Send a 60-second Loom to 30 firm owners showing a synthetic car-park clip turned into a finished PDF log.
  • LinkedIn connection note with one line: 'I built a tool that turns raw CCTV folders into a timecoded evidence log — want the 60-second version?'
  • Email three lines: who you are, the artefact, the free-first-case offer, and a one-click reply.
  • Offer to process one real case free in 24 hours, no account, source deleted in 30 days.
  • Post the before/after clip-to-PDF demo in two PI groups with the same free-case offer.
  • Day-3 and day-7 follow-ups asking only: 'did the log hold up with the adjuster?'
  • Ask every paying customer for one named referral to a firm that handles similar case volume.

Message template

Hi {first_name} — I built ClipLedger for firms that get raw CCTV folders from retailers and HOAs. You upload the folder, type 'person in red hoodie leaves with a box', and get back a PDF evidence log with stills, exact timecodes, trimmed clips and a SHA-256 of the source file. Send me one case this week and I will return the log in 24 hours, free, no account, footage deleted in 30 days. If it holds up with your adjuster, the case is $149. Worth a try on your current file?
Week-one goal: 150 firms listed, 110 contacted, 8 replies, 3 pilot cases accepted, 2 evidence logs delivered.
Section 5 of 8

30-day execution roadmap

MVP deadline: day 5First sale target: day 1230 days planned
Week 1 · Build & validate · 7 days
Day 1Setup

Stand up infrastructure and the payment rail before any customer exists

  • Register domain, create Vercel and Supabase projects.
  • Connect Stripe and link the payment account to TrustMRR.
  • Start the firm list from association directories; target 40 names today.

Deliverable: Live repo, Stripe linked to TrustMRR, 40 firms listed.

Day 2Upload

Get a real video file into storage safely

  • Build presigned folder upload into a per-case storage prefix.
  • Create cases and source_files tables.
  • Add the 30-day deletion timestamp job.

Deliverable: A 20-minute test clip uploaded and visible in a case record.

Day 3Keyframing

Turn raw video into timecoded, deduplicated frames

  • Write the Modal ffmpeg worker: scene detect and keyframe extract.
  • Dedupe frames by perceptual hash.
  • Persist keyframe to source-file timecode mapping.

Deliverable: Keyframe set with exact timecodes from the test clip.

Day 4Caption and search

Make the footage queryable in plain English

  • Batch keyframes to the vision model and store one-line captions.
  • Build Postgres full-text search scoped to a case.
  • Cluster hits into events by timestamp proximity.

Deliverable: Query returns six timestamped, camera-labelled hits.

Day 5ExportMVP shipped

Ship the artefact that is actually sold

  • Build the tick-list review UI with frame preview and clip boundaries.
  • Implement ffmpeg trims, React-PDF evidence log and SHA-256 page.
  • Package PDF plus clips into one ZIP download.

Deliverable: MVP complete: upload to ZIP evidence log end to end.

Day 6Paywall

Only deliver after money lands

  • Wire Stripe Checkout to credit issuance.
  • Gate the processing worker on credit balance.
  • Add magic-link auth and per-case access control.

Deliverable: Paid flow verified in Stripe test mode.

Day 7Demo assets

Make the offer sendable

  • Record the 60-second Loom on the synthetic car-park clip.
  • Export a sample evidence log PDF for attachment.
  • Publish the landing page with pricing and the sample.

Deliverable: Loom, sample PDF and live pricing page.

Week 2 · Launch & first sale · 7 days
Day 8First outreach

Get the demo in front of 40 owners

  • Finalise the 150-firm list with names and emails.
  • Send 30 emails and 10 LinkedIn notes.
  • Log every send and reply in a single sheet.

Deliverable: 40 contacts made, tracking sheet live.

Day 9Pilot intake

Convert replies into free cases

  • Answer every reply within two hours.
  • Send a one-field intake form for the footage link.
  • Accept and queue the first two pilot cases.

Deliverable: Two pilot cases accepted.

Day 10Pilot delivery

Prove the 24-hour promise with human backup

  • Manually spot-check captions before export.
  • Fix PDF layout issues found on real footage.
  • Deliver the first evidence log inside 24 hours.

Deliverable: Pilot log delivered and acknowledged.

Day 11Outreach wave

Widen the funnel

  • Send 60 more emails to the list.
  • Post the demo in two PI or claims groups.
  • Take two more pilot cases if offered.

Deliverable: 110 firms contacted total.

Day 12First saleFirst sale target

Convert a pilot into money in the window

  • Send the $149 pay link to the pilot recipient.
  • Offer the $99 follow-on case if paid within seven days.
  • Ask directly for a yes or a no today.

Deliverable: First paid Case Pack.

Day 13Delivery quality

Make real cases reproducible

  • Deliver any paid cases within 24 hours.
  • Fix the three most common failure states from real footage.
  • Record actual per-case processing cost.

Deliverable: Paid case delivered, cost per case measured.

Day 14Week two review

Follow up and close the second sale

  • Send day-3 and day-7 follow-ups to all pilots.
  • Ask the paying customer 'did the log hold up?'
  • Review reply rate and rewrite the weakest line.

Deliverable: Revenue tally and one revised message.

Week 3 · Repeat & systematize · 7 days
Day 15Recall

Fix weak captions on night footage

  • Add a brightening pre-pass for dark keyframes.
  • Always return top-N nearest captures below threshold.
  • Retest against the first customer's footage type.

Deliverable: Measurably better recall on low-light clips.

Day 16Firm tier

Create the upgrade path

  • Add branded PDF header field.
  • Add five-pack credits and a shared case library.
  • Enable the $299/month checkout.

Deliverable: Firm plan purchasable.

Day 17Outreach wave two

Add 100 new firms and mine referrals

  • Add 100 firms from a second association directory.
  • Ask each paying customer for one named referral.
  • Send 50 new emails.

Deliverable: 250 firms total in the pipeline.

Day 18Proof

Turn a delivered case into a case study

  • Request a two-line testimonial from the paying customer.
  • Write a one-page case study with the anonymised log page.
  • Add both to the landing page.

Deliverable: Case study and testimonial published.

Day 19Reliability

Reduce manual rescue work

  • Add worker retries and a visible job-status page.
  • Email the customer on completion and on failure.
  • Auto-refund credits on failed jobs.

Deliverable: No case requires founder intervention to finish.

Day 20Pricing test

Raise repeat purchase rate

  • Offer every past buyer a second case at $99 within seven days.
  • Track take-up in the sheet.
  • Cap any case above 25 hours with an explicit top-up price.

Deliverable: Repeat offer sent to all buyers.

Day 21Week three review

Decide where the next 10 days go

  • Count paid cases, replies and pilots.
  • Kill the worst-performing outreach line.
  • Pick the best-converting segment and double outreach there.

Deliverable: Written decision on the final push.

Week 4 · Scale what works · 7 days
Day 22Volume outreach

Push contact count past 400

  • Send 100 emails to the best-converting segment.
  • Send 30 LinkedIn notes to firm owners.
  • Book two calls where the buyer asks for one.

Deliverable: 150 new contacts made.

Day 23Delivery cadence

Keep every promise at 24 hours

  • Process all queued cases the same day they arrive.
  • Fix the top support question with a page in the product.
  • Ask each delivered customer for a referral.

Deliverable: Every open case delivered on time.

Day 24Search polish

Improve query-to-hit accuracy

  • Add query expansion into caption-style variants.
  • Show a confidence label on each hit.
  • Test five real investigator phrasings.

Deliverable: Query expansion live.

Day 25Distribution

Get the demo in front of groups

  • Post the before/after demo in three more groups.
  • Comment helpfully on five CCTV-review threads.
  • Reply to every inbound within two hours.

Deliverable: Three group posts and five thread replies.

Day 26Upsell

Move repeat buyers to Firm

  • Offer Firm at $299 to anyone with two or more packs.
  • Explain the five-pack credit math in one line.
  • Close at least one Firm subscription.

Deliverable: Firm subscription attempted or closed.

Day 27Trust

Answer the cloud-upload objection

  • Write a plain data-handling and 30-day deletion page.
  • Add the SHA-256 provenance explanation to the landing page.
  • Put both links in the outreach template.

Deliverable: Data-handling page live and linked.

Day 28Hardening

Make the product survive without the founder

  • Verify the 30-day auto-delete job actually deletes.
  • Add terms of service and a triage-not-legal-advice disclaimer.
  • Test the full flow on a fresh account.

Deliverable: Clean end-to-end run on a new account.

Days 29-30 · Consolidate · 2 days
Day 29Final push

Convert the warm tail

  • Send a last-call offer to every non-responder pilot.
  • Invoice any promised but unpaid case.
  • Collect all revenue inside the window before the deadline.

Deliverable: All outstanding offers closed or declined.

Day 30Close

Reconcile and hand forward

  • Reconcile collected revenue and refunds against TrustMRR.
  • Write the final profit reconciliation and a 30-day post-mortem.
  • Draft the next 30-day plan around the best-converting segment.

Deliverable: Reconciled final numbers and next plan.

Section 6 of 8

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

  • Conservative/Base/Upside assume the Case Pack at $149 is the primary unit and the Firm plan at $299/month appears only in the Upside case.
  • Conservative: 4 paid case packs from roughly 110 targeted contacts, i.e. a reply-to-payment rate well below the concept hypothesis.
  • Base: 12 paid case packs from roughly 300 contacts plus repeat purchases from the same firms.
  • Upside: 24 case packs plus 2 Firm subscriptions at $299 (24x149 + 2x299 = 4,174), blended ARPU rounded to $160 with 26 customers.
  • Payment processing assumed at Stripe blended 2.9% + $0.30, which lands near 3-4% of collected revenue; scenarios use ~4%.
  • Delivery cost assumed near $8 per 25-hour case (ffmpeg compute plus vision-model captioning), which is well under the $149 price.
  • Operating costs include hosting, domain, email and AI delivery; they are not deducted from hackathon profit but are deducted from business profit.
  • Ad spend is paid marketing only; all outreach in this plan is organic and costs $0.
  • No refunds or chargebacks assumed in Conservative and Base; Upside reserves one refunded pack for a missed-recall case.

Conservative

4 customers
$596
collected revenue · avg payment $149
Refunds
$0
Chargebacks
$0
Payment fees
$24
Paid marketing
$0
Hackathon profit
$572
Operating costs
$55
Business profit
$517

Base case

12 customers
$1,788
collected revenue · avg payment $149
Refunds
$0
Chargebacks
$0
Payment fees
$72
Paid marketing
$0
Hackathon profit
$1,716
Operating costs
$125
Business profit
$1,591

Optimistic

26 customers
$4,160
collected revenue · avg payment $160
Refunds
$149
Chargebacks
$0
Payment fees
$166
Paid marketing
$200
Hackathon profit
$3,645
Operating costs
$320
Business profit
$3,325

Model your own numbers

Runs in your browser · nothing is saved
Collected revenue
$1,788
Payment fees (3.9%)
$70
Paid marketing
$0
Hackathon profit
$1,718
Business profit
$1,593

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.

All three scenarios are projections built from the stated assumptions. They are not verified revenue, not existing customers and not traction. Actual results depend on outreach response and delivery quality.

Section 7 of 8

Main risks and validation experiments

Risk 1Firms refuse to upload client CCTV to an unknown cloud tool, so pilots stall before any payment.

Mitigation

Offer a signed-URL upload the firm controls, a written 30-day deletion page, and an optional path where the founder ingests the file on the customer's behalf for the first case.

Validation experiment

Ask the first 10 repliers directly: 'would you upload, or would you rather send a Drive link?' Track which one they choose before building anything else.

Kill criteria

If fewer than 2 of 10 repliers will share footage in any form within week two, the upload model is wrong and the concept needs a local-processing variant.

Risk 2Vision captions are too weak on night or low-bitrate retail footage, so the first real case misses the event and trust collapses.

Mitigation

Brightening pre-pass, top-N fallback results below threshold, founder spot-checks every pilot export before delivery, and a confidence label on each hit.

Validation experiment

Run the same five queries against one night clip and one day clip; count hits found by the model versus hits found by hand.

Kill criteria

If recall on the night clip stays under 50% after the brightening pass, restrict the offer to daytime/indoor footage or stop.

Risk 3Outreach converts far below the assumed rate and no paid case lands before day 20.

Mitigation

Three message variants tested in week one, a free first case to remove price risk, and pivoting to the best-converting segment by day 21.

Validation experiment

Send 30 emails each of two different subject lines on days 8 and 11 and compare reply rates.

Kill criteria

Fewer than 5 replies from 110 contacts by day 14 means the channel or the message is wrong; change the segment before day 18.

Risk 4Per-case processing cost spikes on long or multi-camera cases and a $149 pack loses money.

Mitigation

Hard 25-hour cap in the product, a paid top-up for extra hours, keyframe dedupe before captioning, and cost logging per case from day 13.

Validation experiment

Process the longest available test clip and record actual Modal and vision-model spend per case.

Kill criteria

If measured delivery cost exceeds $40 per case, reprice to per-hour billing before taking more orders.

Risk 5A missed event in a live claim creates a liability argument against the firm and against ClipLedger.

Mitigation

Every PDF carries a triage-not-final-review disclaimer, every export names the reviewer, and marketing never claims exhaustiveness or admissibility.

Validation experiment

Have two investigators read the disclaimer page and say whether they would still attach the log to a claim file.

Kill criteria

If buyers say the disclaimer makes the log unusable in a claim file, reposition to internal-timeline drafting rather than evidence attachment.

Section 8 of 8

Hour-by-hour: the first 24 hours

Hours 1-3

  • Register the domain and create Vercel and Supabase projects.
  • Set the cases, source_files, keyframes, captions, hits and credits schema.
  • Create the Stripe account and connect the payment account to TrustMRR.

Output: Live repo, schema migrated, Stripe and TrustMRR linked.

Hours 3-6

  • Build presigned folder upload into a per-case storage prefix.
  • Add the 30-day deletion timestamp on every case record.
  • Upload a 20-minute test clip and confirm the object lands.

Output: Working uploader with one stored test clip.

Hours 6-9

  • Write the Modal ffmpeg worker for scene detect and keyframe extract.
  • Add perceptual-hash dedupe and timecode mapping.
  • Run it on the test clip and inspect the frame set.

Output: Deduplicated keyframes with exact source timecodes.

Hours 9-12

  • Open the state investigator-association directories and start the firm list.
  • Draft the 60-second Loom script and the three-line email.
  • Log 40 firm names with contact routes.

Output: 40-firm list and a written outreach script.

Complete report

Full 30-day launch report

Read the full generated report

ClipLedger — 30-Day Launch Report

Ship or Die $100K Hackathon, 16 Nov 2026 13:00 UTC to 16 Dec 2026 13:00 UTC. One founder, no team, no ad budget.

1. Product opportunity and positioning

ClipLedger sells to one buyer: the owner or lead investigator of a 1-10 person private investigation or claims-investigation firm handling insurance-defence and premises-liability work in the US, UK or Canada. Their case arrives as raw CCTV exports from a retailer, HOA or property manager — 10 to 40 hours across several cameras, filenames like ch03_20260914_2231.mp4, no index. The reviewer then spends five to eight unpaid hours scrubbing at 8x and hand-typing timecodes into a document.

ClipLedger replaces that scrub with an upload and a sentence. The investigator types "person in red hoodie enters, leaves with box", reviews time-clustered hits, ticks the ones that matter, and downloads a ZIP: a PDF evidence log with a still, camera, exact timecode and one-line description per hit, 15-second MP4 clips, and a SHA-256 hash page for each source file. The artefact is the product. Stills plus timecodes plus provenance is what an adjuster or attorney accepts; a chat answer is not.

Positioning is deliberately narrow. ClipLedger is first-pass triage, not final review, not legal certification, not a surveillance platform. No live streaming, no face recognition, no identity matching. That keeps the buyer's liability story intact and keeps the build inside a solo founder's reach.

2. Smallest sellable MVP

One upload, one query, one ZIP, one payment. Upload a folder to a per-case storage prefix; ffmpeg splits on scene changes, extracts keyframes, dedupes them and keeps the source timecode; a vision model captions each frame in one line; Postgres full-text search returns hits clustered into events; the reviewer ticks and adjusts clip boundaries; the export builds the PDF, the trimmed clips and the hash page; Stripe issues a case credit that unlocks processing.

Out of scope, permanently for this launch: RTSP streaming, real-time alerts, face recognition, a mobile app, team administration, redaction, and any human-review marketplace. Each of those is a different product with a different sales cycle.

3. Technical implementation plan

Next.js on Vercel for the app and review UI; Supabase for Postgres, auth and storage; Modal for ffmpeg and captioning workers; Gemini Flash vision with GPT-4o-mini as fallback; Stripe Checkout; React-PDF for the log; Resend for magic links and job-status email.

Uploads go browser-to-storage via presigned URLs so video never touches the app server. A case row moves through queued, splitting, captioning, ready. Two Modal functions do the work: one for scene detect, keyframe extraction, phash dedupe and timecode mapping; one for batched captioning. Search is a full-text query over captions scoped to a case and clustered by timestamp gap. Export runs ffmpeg trims, streams the PDF, hashes the sources and zips to a signed URL.

The hardest technical risk is recall on low-light footage. Mitigation: a brightening pre-pass on dark frames and always returning the top-N nearest captures even below the confidence threshold, with a visible confidence label.

4. Pricing and monetization

Two prices only. Case Pack, $149 one-time: up to 25 hours of footage, unlimited searches for 30 days, PDF evidence log, trimmed clips, hash page. Firm, $299/month: five case packs a month, shared case library, branded PDF header.

The first-revenue offer is the whole strategy: send one case folder this week, get the evidence log back in 24 hours, free, no account. If it holds up with the adjuster, that case is $149 and the next is $99 if paid within seven days. Repeat purchase is the growth engine — one firm that pays three times is worth three customers.

5. First-customer acquisition

One channel: direct LinkedIn and email to firm owners sourced from public state and provincial investigator-association member directories. No ads, no SEO, no partnerships.

Day 7 the demo exists: a 60-second Loom searching a synthetic 20-minute car-park clip ("man in red hoodie loads a box") and showing the finished PDF. Day 8 to 11 that goes to 110 firms. The message is three lines and one offer. Follow up on day 3 and day 7 with a single question: did the log hold up with the adjuster? Post the same before/after in PI and claims-investigation groups. Week-one goal: 150 firms listed, 110 contacted, 8 replies, 3 pilot cases, 2 logs delivered.

6. 30-day execution roadmap

Week 1 (days 1-7), build. Day 1 setup and payment rail; day 2 upload; day 3 keyframing; day 4 captioning and search; day 5 review UI and export — MVP done; day 6 Stripe gating and auth; day 7 Loom, sample PDF, landing page.

Week 2 (days 8-14), first money. Day 8 first 40 contacts; day 9 two pilots accepted; day 10 first log delivered inside 24 hours; day 11 outreach to 110 firms; day 12 first $149 sale; day 13 fix what real footage broke; day 14 follow-ups and second sale.

Week 3 (days 15-21), quality and proof. Night-footage recall, Firm tier, 100 new firms, testimonial and case study, reliability work, a $99 repeat-case test, and a written decision on where the last ten days go.

Week 4 (days 22-28), volume. 150 more contacts, same-day delivery cadence, query expansion, group distribution, Firm upsell, a data-handling page to answer the upload objection, and a fresh-account end-to-end test.

Week 5 (days 29-30), close. Last-call offer to non-responders, invoice anything promised, reconcile against TrustMRR, write the post-mortem.

7. Revenue and profit scenarios

All figures below are projections from stated assumptions. They are not verified revenue, existing customers or traction.

Conservative: 4 case packs × $149 = $596 gross. Fees $24, no refunds, no ad spend. Hackathon profit $572. Operating cost $55. Business profit $517.

Base: 12 case packs × $149 = $1,788 gross. Fees $72, no refunds, no ad spend. Hackathon profit $1,716. Operating cost $125. Business profit $1,591.

Upside: 24 packs plus 2 Firm subscriptions, 26 customers at $160 blended = $4,160 gross. Refunds $149, fees $166, ad spend $200. Hackathon profit $3,645. Operating cost $320. Business profit $3,325.

Fees assume Stripe at 2.9% + $0.30, landing near 4% on a $149 charge. Delivery cost is budgeted near $8 per 25-hour case, so every sale at $149 clears its own delivery cost by a wide margin.

8. Main risks and validation experiments

Upload refusal. Firms may not put client footage in an unknown cloud tool. Test: ask the first ten repliers whether they would upload or send a Drive link. Kill signal: fewer than two of ten will share footage in any form by week two.

Caption recall on night footage. Test: run five queries against one night clip and one day clip and compare model hits to hand-found hits. Kill signal: recall under 50% on night footage after the brightening pass — restrict to daytime/indoor or stop.

Outreach conversion. Test: two subject lines, 30 emails each, on days 8 and 11. Kill signal: fewer than five replies from 110 contacts by day 14.

Delivery cost blowout. Test: process the longest available clip and log actual Modal plus vision spend. Kill signal: over $40 per case — reprice per hour before taking more orders.

Liability framing. Test: have two investigators read the triage disclaimer and say whether they would still attach the log to a claim file. Kill signal: the disclaimer makes the log unusable, in which case reposition to internal timeline drafting.

9. First 24-hour action plan

Hours 1-3. Domain, Vercel and Supabase projects, full schema, Stripe account connected to TrustMRR before any customer exists. Output: repo, migrated schema, linked payments.

Hours 3-6. Presigned upload, 30-day deletion timestamp, one 20-minute test clip stored. Output: working uploader.

Hours 6-9. Modal ffmpeg worker, phash dedupe, timecode mapping, inspected frame set. Output: deduplicated keyframes with exact timecodes.

Hours 9-12. Firm list started from association directories, Loom script and three-line email drafted, 40 firms logged. Output: 40-firm list and outreach script ready to send on day 8.

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 →