How to write a useful interest stream
A useful stream request says what you want to follow and what would make an update relevant to you. A couple of ordinary sentences is enough to start.
Start with the subject
Name the thing you want to keep up with. It can be a field, a place, an industry or a particular question. Use natural language: a clear subject and the angle you care about can be more useful than instructions to an imaginary expert.
‘I want to follow new battery research’ is a reasonable starting point. ‘I want to follow new battery research, especially recycling methods and results that might scale’ makes the angle clearer.
Add what matters, and what doesn’t
A boundary helps separate an interesting development from a merely related headline. You might care about early scientific results, practical applications, a particular country, or changes affecting a profession.
Keep me up to date on battery research and recycling. Include useful lab results and explain what still needs to work at scale. Skip stock-price stories.
That request contains a subject, two useful angles and an exclusion. It does not need a long list of rules. Too many exclusions can hide developments you would have wanted to see.
For a different interest: ‘Follow independent film releases in the UK, especially documentaries. I want release dates and a little context, without plot spoilers.’ These are example requests, not promises of complete coverage.
Revisit exclusions as well as subjects. A request to skip market stories might filter out a funding decision that affects the research itself; choosing a boundary always involves a trade-off.
Know when to make another stream
Split subjects when you would want to browse, edit or pause them independently. Battery research and independent films are easy candidates. Closely related questions may belong together if the same stories answer them.
Separate streams can make it easier to change an interest without rewriting everything. In an earlier Carrot alpha, a roundup combined developments across streams and a side panel linked to each topic.
Refine it after the first brief
Read the first set of stories before making the request more elaborate. Which story was useful? Which was only loosely related? Edit the stream around that difference.
If the topic is right but an explanation is too technical, you may want a different writing style or more or less detail. These are separate choices from changing what a researcher looks for; available controls differ between services.
A practical revision is ‘focus more on recycling processes than new battery chemistries’. A vague revision such as ‘make this better’ gives the researcher much less to work with.
The earlier Carrot alpha screenshots show an example of a voice or text stream request. For any AI-written brief, use these source checks before relying on a claim.
Changing a request is a way to experiment, not proof that every relevant development will appear. Periodically read a broader source and see which questions your stream did not lead you to ask.