Real, verified people. Not bots, not your friends being nice. You earn them by rating 10 strangers' posts.
100 / 100
Takes about 4 minutes. One human, one quota.
The cold start is brutal. Feeds show new posts to almost nobody unless you are already big.
Paid boosts buy bots. You get a number, not an audience, and no idea if the work is good.
Here, follower count does not matter. Good work climbs to 1,000, then 10,000 raters.
Step 1 of 3 · Prove you are one person
One human. One quota. No bot farms.
We never learn your name or see your face. A proof-of-personhood app sends us a yes and a private code that stops you joining twice.
Opening your verification app
Checking you have not verified before
Zero-knowledge proof received. No name, no face stored.
Verified human. Quota slot #48,213,907 is yours.
Step 2 of 3 · Earn your viewsAccuracy 70
Hook: did the first second grab you?1 meh, 5 hooked
Clarity: did you get it?1 lost, 5 crystal
Would you share it?
Predict the crowdHow many of 100 will say yes to sharing?
What would make you share it?Optional, anonymous
Quota earned
0
honest views are yours.
You rated 10 strangers from 10 cities. Somewhere, 100 strangers are about to return the favour.
78rater accuracy
1.1xyour vote weight
0notes you left
How your accuracy works
Each "predict the crowd" guess is checked against the final consensus. Accurate raters count for more, so lazy or hostile ratings fade out on their own.
Step 3 of 3 · Your post
Show us what you made.
Ask your raters one thingExtra views
Your 100 views happen inside the First Hundred feed only. We never push views onto other apps. You get a results link to share instead.
Getting it in front of people
Finding your hundred
Live
0 / 100
verified views follower-blind, 14 countries
Waiting for the first rater...
Watch-through
Share of raters still watching, second by second.
Biggest drop: 0:07 to 0:11. The mixing bowl shot. 30 people left right here. Cut it to one second.
0.0hook
0.0clarity
0%would share
What would make me share this
Best anonymous answers, weighted by rater accuracy.
Score vs the ladder line
Ladder line 7.0
Your score: 7.6 / 10.Above the line. You climb, no followers needed.
Share the proof, not a fake number
Your quality badge
Rated by 100 humans
Funmi's Oven
7.6 score
64% would share
14 countries
firsthundred.app/r/funmi-agege-7q2
Verified people, follower-blind, rated inside First Hundred. Now climbing to 1,000.
What people see when they tap it
Your public results page: score, the curve, three anonymous notes, and a "Get your first hundred" button. That button is how the next creator finds us.
You, this week
Keep the hundred coming.
Level 4 · Sharp Eye
Your guesses beat 71% of raters. 32 more accurate ratings to Level 5, where your votes count 1.3x.
Weekly streak
6 weeks
Rate 10 a week to keep it. Your quota refills Sunday.
0views in the bank
10to rate for the next 100
3.00USDC top-up balance
Your posts on the ladder
Agege bread at 5am7.6 · climbing to 1,000Rung 2
How we price a loaf6.4 · 100 rated, notes savedRung 1
Puff-puff, 40 seconds8.1 · reached 10,000Rung 3
The grand mission
An open, follower-blind attention layer for the internet.
Any app plugs in and every new thing made by anyone gets a fair first hundred, from real people, before an algorithm decides its fate.
Year 1
The app. 1 million verified raters trade attention with each other. Bakers, poets, tutors and indie devs in 60 countries get honest feedback on day one.
Year 2
The open API. Any platform requests "100 verified, follower-blind raters" for a post and pays per batch in USDC over x402. Raters get paid for accuracy, not for being nice.
Year 3
The public rater graph. A portable reputation for good taste. Feeds everywhere can start new work from a fair baseline instead of follower count.
The number that matters
Today, most new posts are seen by fewer than 10 strangers. Our goal: every post made by a verified human gets 100. That is 10x more fair chances, every day, for everyone.
Blueprint
1The mission
Give every new post from every real person a fair first audience: 100 honest views from verified humans, regardless of followers. Good work then climbs on merit, not on who you already know.
What is actually new (honest version). Paid, verified-human attention already exists for brands (reward-for-attention platforms, World ID gated attention experiments). Nobody gives ordinary creators a guaranteed, follower-blind first audience. The new parts: give-to-get (rate 10 to earn 100), one proof of personhood equals one quota, raters weighted by how well they predicted consensus, and a ladder to 1,000 and 10,000 that money cannot buy.
2Why people can't resist it
Guaranteed audience. The cold-start fear disappears. "100 people will see this" is a promise no feed makes.
Reciprocity. You rate 10, you get 100. It feels fair, and rating strangers is oddly fun with the crowd-guess game.
Status you can prove. A "Rated by 100 humans" badge and a ladder rung are a credible signal of quality, not a vanity count.
3How it works for real
Proof of personhood. World ID SDK (or passport NFC with a zero-knowledge proof) returns an app-scoped nullifier. One nullifier, one quota. No name or face stored.
Upload. Video to object storage, transcoded to a 480p HLS preview. Player emits watch events every 0.5 seconds for the retention curve.
Captions. A Whisper-class speech model transcribes for subtitles and search.
Safety and tags. A fast cheap model through OpenRouter (Gemini Flash or DeepSeek) classifies text and topic; a vision model samples 4 frames for nudity, violence and spam.
Matching. Random, follower-blind sample of verified raters, stratified by region and language, excluding anyone who follows or has rated you recently.
Rater weighting. Each "predict the crowd" guess is scored against the final consensus (peer prediction, in the spirit of Bayesian Truth Serum). Weights update nightly.
Feedback summary. A frontier LLM (Claude or GPT via OpenRouter) clusters the 100 notes into the top 3 actionable ones and finds the drop-off moment's cause from captions plus the curve.
Ladder. Weighted score above 7.0 opens the next 1,000 raters, then 10,000 at 7.8.
Rails. USDC on Base for top-ups, a smart wallet with email login so crypto stays invisible, and x402 for other apps buying rater batches per request.
4Cost per use
Personhood check (once per user)
$0.00
Transcode and serve 100 views of a 22s reel
$0.01
Transcription, 22 seconds
$0.003
Safety plus vision frames (Flash class)
$0.002
Comment clustering and notes (frontier LLM)
$0.015
Total per post, first 100
about $0.03
Free: rate 10, get 100 views, up to 3 posts a week. The currency is your attention, so free is cheap to run. Paid: 1 USDC per extra 100 views at rung one (most goes to accurate raters as micro-rewards), or Creator Pro at $6 a month for region breakdowns and A/B testing two hooks. Ladder rungs are never for sale. No generated video here, so no $0.10 to $0.50 per second costs.
5Growth loop
Creator posts the badge link on other platforms. Viewers tap through to a public results page with "Get your first hundred". To post, newcomers rate 10 first, so every new creator adds 10 ratings of supply before they take 100 views of demand. Rater streaks keep the feed full between posts.
6Risks and guardrails
Platform rules. Views happen only inside First Hundred's own feed. We never inject views or plays into other platforms. The badge is a link to our results, not a fake metric.
Bots and farms. Proof of personhood, minimum watch time, attention checks, and speed limits on ratings.
Rating rings. Follower-blind random matching, graph detection of reciprocal pairs, and accuracy weighting that buries coordinated scores.
Paying raters. Rewards are for accuracy against consensus, never for positive ratings.
Privacy. No biometrics stored, only a nullifier. Notes are anonymous; abusive notes are filtered before delivery and can be reported.
Safety. 18+ raters, content classifier before matching, and humans review anything flagged.
7MVP build order (this week)
Web feed plus the 4-question rating form. Seed with 200 posts from 20 creator friends.
World ID sign-in. One quota per nullifier.
The give-to-get ledger: 10 ratings unlock 100 views; a matching job serves posts follower-blind.
Results page: retention curve from player events, LLM note summary, shareable badge link.
Accuracy weighting and the ladder to 1,000. Add USDC top-ups after week one.