Nufeed
A personalised news feed that lets people design their own algorithm instead of inheriting one — taken from spec to a public test-flight in weeks.
Overview
Nufeed lets people track the topics they actually care about and have AI rewrite what it finds to their own instructions, instead of accepting whatever an engagement-driven feed surfaces.
The problem
The idea was clear before any of it existed as software: monitor the web for a topic, let a user set rewrite instructions in plain language, publish through an editor good enough to use daily. None of it existed yet, and the founder needed a working product in front of early testers, not a deck.
What we did
We built the three pieces as one pipeline rather than three separate features: topic monitoring, an AI rewrite step driven by the user’s own instructions, and a Notion-style editor to polish and publish. The first version stayed deliberately narrow — enough topics, enough rewrite control, enough of an editor to be genuinely usable — so it could reach a closed test-flight rather than wait for a longer feature list to finish.
Outcome
- Working product shipped from spec to a closed test-flight in two months
- Feed rewrites return in under two seconds per topic
- Users set rewrite instructions in plain language, with no configuration UI required
- Privacy-first pipeline: no user content used for model training
Have a system like this?
Tell us where it hurts. Discovery is two weeks and you keep the plan either way.