A film for every day’s news, without the daily hunt
Yoram Schaffer wanted a film from his library for each day’s news, without spending his mornings searching for one. Now he gets up to three ideas a day, with the post already written.
The need
Yoram runs a Hebrew library of documentaries and Israeli films. On Facebook he puts a film next to a story in the day’s news, so people meet the film while the subject is on their mind.
Finding that film takes time, every day. Some days are easy: on Holocaust Remembrance Day or a holiday, he already knows the film. The hard days are the ordinary ones, such as a tech company in trouble or a new health study. The right film for those is one he would not think of quickly, and that is where the value is.
He had three conditions. It had to cost less than paying a student to do it. It could not need a subscription. And it could not tie him to one tool.
What he gets every day
- One email a day with up to three ideas. Each idea links the news story and the film, says in a line or two why they belong together, and includes a Hebrew post he can edit and publish.
- Ideas for the stories his audience cares about: science, health, education, technology, culture and life in Israel. Politics stays out. Every idea comes from the 788 films in his own library.
- Every idea passes the questions he asks himself. Is the film about the same subject? Do the place and time fit? Does its point of view suit the moment? Is it about the same kind of people? Is it sensitive? Sensitive ideas are flagged.
- On a day with no good match, the email says so. He never gets a forced pairing.
- He replies with one tap under each idea: good idea, not for me, or posted. He can add a comment if he wants.
It learns what he likes
Yoram never writes rules. His votes and comments do the teaching. The agent started from what his own past posts show about his taste, and every vote adds to that.
When it paired a mayor’s arrest with an election film, he wrote that every politics story was leading to that one film. It stopped suggesting films for stories about politicians. When it linked a story about dogs to a film about people, he wrote, “Can’t mix dogs and humans… people won’t accept it.” It stopped doing that too.
A comment changes the next day’s ideas. A vote without a comment needs a second vote that agrees before it counts.
What we got wrong first
The first ideas matched his past posts well, and he liked them. Then he told us plainly that they were films he would have found himself.
An idea is only worth sending if he would not have thought of it. We changed the brief to “beyond the obvious”. Since then, the only measure we use is his daily votes.
He stays in control
The agent writes only to Yoram, and never more than five emails a day. It never posts anything. He chooses what goes on his page and edits it first. One switch stops it.
From first call to daily email in eight days
We talked about the idea on 6 October 2026. Yoram received his first ideas on 12 October, and the daily email started on 14 October.
It took about 5 to 6 hours of our founder’s time to build. Keeping it running takes about 5 minutes a day.
Scale & result
On 14 October 2026, the first day of the daily email, the top idea for a story about the collapse of Character.ai was the film “I Fell in Love with a Bot”. Yoram had picked that same film for that story himself the night before. His reply: “Am I a bot? 😀😆 … The other recommendations are also excellent. Good job 🙌🏆”
More AI work we have shipped
Every claim above is a shipped, production system.
Tell us what you're building.
The problem, roughly what you think it needs, and where the data lives today. We will tell you what we would actually build — including when the answer is less than you asked for.