Content to commerce, from AI
Fifty-six days, fifty-six reels, one person — who never wrote a script, opened an editor, or stood in front of a camera. Organic views went up tenfold, and the product those reels point at went up ninefold behind them.
Aditya Bhatt5 min read
A reel a day for 56 days on our own account, made from nothing but voice notes. 710,569 organic views against about 66,000 in the 56 days before, and the product every one of those reels points at went from about 21 people a day to 193.
This is the machine we sell, run on ourselves. Here is what it did, and here is how it works.
56 / 56
days with a reel published. Not one day skipped.
710,569
organic views, 10.8× the 56 days before.
371,100
people reached, each watching 2.0 times on average.
9×
daily users on the product at the other end of the link.
The bet
Content to commerce, from AI. Someone who knows one thing cold — a doctor, a teacher, a lawyer, a machinist — should be able to reach the people who need that thing and sell to them. Today it takes five hires they are never going to make, so the expertise sits there and reaches nobody. The expertise was never the bottleneck. The production was. We remove the production.
56 days, 56 reels
The account had been doing about two thousand views a day for months. It broke on 5 August at 137,000 views in a single day, and the part that matters more than the spike is the right-hand edge of the chart: the floor afterwards sits between twelve and twenty-two thousand a day. An order of magnitude above where it started, and it has stayed there.
| Metric | 24 June 2026 – 18 August 2026 | The 56 days before | Change |
|---|---|---|---|
| Views | 742,078 | about 149,600 | +395.9% |
| Organic views | 710,569 | about 66,000 | +977.3% |
| Reach | 371,100 | about 104,700 | +254.4% |
| Content interactions | 101,200 | about 5,300 | +1,800% |
710,569 organic views, up 977.3% on the period before. 371,100 people reached, up 254.4%. 101,200 interactions, 19× what the account was taking before — and on this account that is the number that matters most, because the call to action is a comment. Total views came to 742,078 across 56 reels, an average of 13,251 a video.
A voice note in, a finished reel out
That is the entire input. The creator talks for a few minutes into their phone, in whatever language they think in, while doing something else. Everything after that is ours.
- Workflows that decide and write. Research on what the field is actually doing this week, ideas drawn wide and selected, then a script written against the account's own measured history — which reels worked, which floored, and the mechanic each one used. Each stage hands a structured result to the next rather than an email to a person.
- An MCP server that edits. Assembly, timing, the visuals, the captions and the sound pass, exposed as tools rather than buried in an application. It is harness-agnostic: anything that speaks MCP can drive it, including ours.
- A presenter who is the creator. The face and the voice on screen are their digital twin, built once and used on every video after that. This is why the input is a voice note and not footage.
- A funnel that answers itself. Every reel carries the same call to action — comment anything, and the link arrives as a direct message within seconds, at any hour, without anybody being awake for it.
- An inference fleet sized for the whole thing. Enough compute to run all of the above every single day, for every creator on it, without a queue. Daily publishing is a throughput problem before it is anything else.
The reason a content operation is a team is that every handover is a meeting. When each stage’s output is an object the next stage reads directly, the meetings stop existing, and what is left of the week is the part that was actually making something. Which is why one person published 56 videos in 56 days, and why the list of people that person did not need runs like this:
- A content strategist to work out what is worth saying
- A scriptwriter
- A video editor
- A motion designer
- A camera, a set, and the hour it takes to record
- Somebody answering the comments
One thing is not automated and will not be: somebody watches every video before it goes anywhere. Nothing upstream of that can tell the difference between a video that is fine and a video that is good.
Then commerce
An account having a good month is a hobby with an audience. The thesis turns on the second half of the sentence, so here is the other curve: daily active users on AI College Buddy, the product every one of those reels points at.
- People using the app, daily
- Views, daily
The product sat on about 21 people a day for weeks. It left that floor on 7 August, two days behind the account’s best day, and hit 537 on 10 August. The peak is not the point — peaks decay. The point is 13 August to 19 August, which averaged 193 a day and has not come back down. Content moved the ceiling for a week. It moved the floor for good.
Daily active users is the first metric we tracked and the shallowest one. The next is revenue, and the bet underneath it is the one nobody wants to underwrite: that a person needs to own two things — the idea and the product — and everything standing between those and a paying customer can be machine. Not the strategist, not the writer, not the editor, not the camera, not the person answering the messages at midnight.
56 days, 56 reels, one person, one voice note at a time. That is the whole game, and we are only at the top of it.
Tell us what you sell. We’ll write when we open.
We are not taking on new accounts right now — the system is being improved on the ones already running, and our own. Leave your name and what you sell, and you hear from us before anyone else does.
No new accounts until the system is ready. When it opens, the first week is on us.