Rex Turns Your Camera Roll Into a Social Recommendation Feed

01 /The Concept

Rex is a recommendation app with an unusual content source: your camera roll. The platform's AI watches photos as they're taken, identifies what's in them, and prompts users to convert the relevant ones into recommendations. Photograph food at a restaurant, and Rex extracts the geolocation, identifies the venue, and offers to post it as a review with a short caption. If you took ten photos at dinner, the AI selects the strongest one rather than asking you to decide.

Published recommendations are visible only to friends and followers within the Rex network – not to the general public. There are two views: a map that shows pins for each recommendation, and a "playlist" that organizes recommendations by city and then by specific location. Each pin opens the photo and the attached note.

The founder traces the idea to a personal obsession with travel planning. The platform's Instagram presence, however, skews heavily toward food photography – which makes practical sense. Food is what people photograph most, and restaurants are one of the most useful things to get honest recommendations about.

Rex spent its first 18 months building the image recognition system before entering beta. It just exited beta – with the announcement of a $3.96M seed round – and has not yet launched any monetization tools. The investors appear to be betting that if the core behavior takes hold, revenue models will follow.

02 /Why It Matters

Social influence shapes purchasing decisions more than any other signal: industry data suggests around 74% of purchase decisions are affected by recommendations from social networks. Visual recommendations amplify this further – posts with images get roughly 1.5x more engagement and nearly twice as many link clicks as text-only posts.

Set against that demand sits a supply problem. In 2022, people took an estimated 1.72 trillion photos – roughly 4.7 billion per day. The overwhelming majority were taken on smartphones (92.5%), and most were shared in some form on social platforms. Yet query any location for recommendations and you'll find a few dozen entries, many of them months old, with none of the organic character of a friend's actual opinion.

The gap is enormous: billions of photos created daily, and a relative handful of published recommendations. Rex's thesis is that a meaningful fraction of those photos are latent, unspoken recommendations – and that removing the friction of converting them (geolocation is automatic, photo selection is automated, posting takes one confirmation) will unlock a qualitatively different volume of genuine local intelligence.

The trust layer reinforces this. Most people distrust review aggregators precisely because they aggregate everyone's opinion indiscriminately. Rex shows you only what people you actually follow have recommended, which makes the signal more relevant and harder to game than SEO-optimized Yelp listings.

A [related review](/review/rekomendacii-rabotajut-esli-sdelat-vot-tak) covered Atly, which takes a similar approach but routes recommendations through interest-based communities rather than personal social graphs. The two mechanisms point in the same direction: people don't want more reviews, they want better-filtered ones.

03 /Opportunities

The interesting principle behind Rex is not the recommendation format but the supply strategy. Instead of asking users to create new content, it converts behavior they're already engaged in – taking photos – into a content format that has separate value. The marginal effort is close to zero; the output is a recommendation.

BlaBlaCar used the same underlying logic: drivers were already making long-distance trips; the platform just gave them a way to convert unused car seats into revenue. The behavior existed; BlaBlaCar added structure around it.

The question worth sitting with: in what other domains is there a large existing behavior that could, with minimal friction added, be converted into something valuable – a recommendation, a dataset, a contribution to a shared resource? The answer probably isn't in places where the behavior is already well-documented, but in the gaps where valuable information is being generated and immediately discarded.

Rex as a template is replicable. The camera roll as an input layer for surfacing location-based social content is a defensible product angle – the behavior (photographing everything) is embedded in modern life, and even capturing a small fraction of it as structured recommendations would shift the scale of what's currently available.