Complete history
Extended streaming history is imported, normalised and reconciled, so a play from 2009 sits in the same table as one from this morning.
Independent studio · Gothenburg
Your listening history, turned into a map of your taste.
SpotRecap is a private analytics platform for music. It holds every play you have ever streamed, grows a model of taste from that record, and answers questions about it in plain language.
Live in private beta · EU data region · 271,712 plays indexed
Plays this month
3,184
+12% against your five-year average
271,712plays indexed end to end
17 yrsof continuous listening history
6model providers behind one interface
9 msp50 read latency, analytics API
Measured against the reference history on 9 November 2026.
The product
Built like infrastructure, not like a dashboard.
Most listening stats stop at a wrapped year. SpotRecap starts with the complete record — every play, track, feature and date — and keeps learning from what you listen to next.
Taste model
Audio features, listening context and explicit feedback build a per-track vector space — the basis for every recommendation the product makes.
Answers, not charts
Ask “what did I play the week I moved?” and get a real answer with the numbers behind it, streamed back as it is computed.
Memory that persists
The assistant keeps durable notes about your taste and past questions, so context carries between sessions instead of resetting.
Phone-first client
The phone client is in build now: a 380 px design width, installable, one-handed. Anything not usable there does not ship.
Provider-independent AI
Six model backends sit behind one abstraction layer, with a fallback chain and per-request routing. No lock-in to a single vendor.
Why it exists
Streaming platforms remember everything. They just never show it to you.
The insight a listener actually wants is not an annual summary. It is the long view: what changed, what stayed, which records keep coming back after a decade.
SpotRecap was built to answer that question for one listener first — a testbed that had to survive seventeen years of messy exports before it was allowed to become a product.
- Your data stays yours
Single-tenant storage, no advertising pipeline, no resale of listening data. Export at any time.
- Readable by default
Every number on screen is traceable to records you own — down to the individual play.
- Explains its reasoning
Recommendations carry the evidence that produced them: features, neighbours and history.
- Runs lean
A cost ladder is a first-class roadmap phase, not an afterthought once the bill arrives.
Roadmap
Where the product is going.
Phases ship in public order. A phase is finished when it is deployed, measured on a phone, and has a test that fails loudly if it breaks.
Ingestion and the data foundation
Full history import, 14-table schema, materialised views and vector indexes, plus the analytics API that everything else reads from.
The assistant
Thirteen question handlers, answers streamed as they are computed, and a provider abstraction with an ordered fallback chain.
Memory and the taste model
Track vectors, listening context and feedback stored so the model improves with use and context survives between sessions.
The phone client
An installable progressive web app at a 380 px design width, with the dashboard, explore and chat surfaces in one shell.
Insight suite, then discovery
Year recaps, streaks and artist trajectories, followed by playlist tools and a discovery engine that explains itself.
The team
Three people, one product.
Erik Lindqvist
Founder · Engineering
Data platforms and backend systems. Built the ingestion pipeline and the query layer the product runs on.
Klara Sjöberg
Product · Design
Product direction and interface design. Responsible for the phone-first shell and the visual language.
Oskar Berglund
ML · Infrastructure
Recommendation models, evaluation and the cloud estate — from cost controls to release automation.
Private beta
Bring seventeen years of listening. We will show you what is in it.
Access is granted in small batches so the model has time to learn each history properly. Tell us what you listen to and we will be in touch.