Magic Lamp

State the wish. The lamp runs the loop.

wish: 50,000 installs · ≤ $2.40 CPI · US / CA / UK · by Sept 1
  1. reading the market … 1,412 competitor creatives studied
  2. staking hypotheses … 5 creative theses clustered
  3. casting creatives … 30 videos, six per thesis
  4. launching … 5 ad groups live, structured by thesis
  5. watching … day 3 — “resource anxiety” at $2.11 CPI
  6. reallocating … winners fed, losers cut
  7. exploring … 12 variants cast around the winner
  8. again

— a rehearsal. The numbers are illustrative, not customer data.

Magic Lamp is an autonomous performance-marketing engine. You set a goal and a budget. It researches the market, stakes creative hypotheses, produces the ads, launches real campaigns, reads the results, and reinvests — smarter every round.

see how the lamp works ↓

the problem

Every ad platform automated the buying. Nobody automated the learning.

Targeting is machine-run. Bidding is machine-run. Delivery is an auction settled in milliseconds. But the part that decides whether any of it works — what to say, to whom, and why it worked — still runs on humans, decks, and hunches. It moves at the speed of meetings.

Teams burn weeks producing creative they cannot explain, test it in campaigns structured so the results mean nothing, and start the next quarter no smarter than the last. The bottleneck in paid growth is no longer media. It is the speed at which an advertiser can learn.

the loop

What happens when you make a wish

  1. i · research

    The lamp reads the market.

    It studies the creatives your competitors are running, what the platforms are rewarding this month, the cultural hooks your audience already responds to — and what your product truthfully delivers.

    1,412 creatives studied · 9 working patterns isolated

  2. ii · theses

    Hypotheses, not hunches.

    Ideas arrive as clusters of testable campaign theses — each one an audience belief, an emotional trigger, and a hook mechanic with a reason to exist.

    resource anxiety · the protector · one more pull · the comeback · idle hands

  3. iii · creatives

    A portfolio per thesis.

    Multiple videos for every thesis, with structured variation across hook, story, pacing, character, and call to action — a portfolio designed to be read, not a pile of assets.

    30 videos cast · six per thesis

  4. iv · launch

    Real campaigns, structured by thesis.

    Ad groups are organized around the hypotheses they test, on the networks you already buy — so when results arrive, they mean something.

    5 ad groups · 2 networks · live

  5. v · readout

    It watches everything.

    Spend, click-through, conversion, cost per install, creative fatigue — read continuously, at the thesis level, where the signal actually lives.

    resource anxiety $2.11 · the protector $2.38 · one more pull $4.07

  6. vi · reallocate

    Winners fed. Losers cut.

    Budget flows toward theses that prove themselves and away from those that don’t — automatically, daily, without a meeting.

    resource anxiety the protector one more pull — cut

  7. vii · variants

    Explore around the winners.

    New creatives are cast adjacent to what worked — holding the why steady while the surface varies — to map the territory around a proven idea.

    12 variants of “resource anxiety” · 3 new hooks · 2 new settings

  8. viii · again

    Each round starts smarter.

    Everything the campaign learned — which beliefs moved which audiences, at what cost, for how long — compounds in the lamp’s memory. Round two begins where round one ended.

    round 2 · CPI $2.40 → $1.96 · memory retained

every number above is illustrative — a walkthrough of the loop, not a customer's results.

why it compounds

The unit of progress is not a video. It is a tested thesis.

one more pull resource anxiety the protector the comeback idle hands 12 variants · the why held steady again … round 1 · the explosion round 2 · the walk round 3 · again wish
the same campaign, drawn as a search: five theses staked, three cut, the winner explored. knowledge banked either way — illustrative.

Anyone can generate fifty random ads; the platforms will happily take the money. The lamp stakes five hypotheses, tests them honestly with real budget, and keeps what it learned. Knowledge your competitors pay for again every quarter, it banks.

That memory — which beliefs move which audiences, which hooks decay, which ideas travel between markets — is the asset. The videos are just how it asks the question.

where it starts

First: games.

Mobile game user acquisition is the most demanding creative-testing discipline in advertising — $25 billion a year of spend, video-first, measurable in days, and creatively exhausted on a brutal cycle. UA teams already think in theses, variants, and cost per install. They just run the loop by hand.

If the lamp wins here, it wins anywhere paid creative is measured: apps, ecommerce, subscriptions — every advertiser with a goal and a budget.

Old stories ration the magic — three wishes, choose carefully. This lamp grants one wish, ten thousand times, better each time.

The lamp is lit.

We are taking a small number of launch partners with live user-acquisition budgets.

by invitation, for now