First release: recommendation
Recommendation is where behavioral data is densest and where the evaluation is least ambiguous, so it is where we started. The model is a sequence model over user interaction histories. It consumes a user’s ordered actions and predicts what comes next, in the same autoregressive way a language model predicts the next token. Items are represented as semantic IDs rather than as rows in an embedding table, which is what lets the model transfer across catalogs at all.Cold start
A new item is representable the moment it has content. No interaction history required before it can be recommended.
Transfer
Pretrained on interaction data across domains, then adapted to your catalog. Useful before you have accumulated your own signal.
Generative retrieval
The model generates an item identifier directly. Constrained beam search over the codebook replaces approximate nearest neighbor over an index.
Scales with catalog
Parameter count is set by the codebook, not by the number of items. A catalog can grow without the model growing with it.
How you would use it
Tokenize your catalog
Fit a semantic ID tokenizer on your item content. See Semantic IDs.
Adapt on your sequences
Supervised fine-tuning on your user histories, expressed as next-token prediction over semantic IDs.
Align to your objective
A reinforcement learning pass against your actual ranking objective, not a proxy. What counts as a win is your definition, set at onboarding.
Serve
Generate recommendations through the API, or take the weights and serve them yourself.