Introducing EdgeRock · Open weights. Fixed-capacity history.
Meet the model ↗A place to study the architecture, build with the model and ask what comes next. Bring your notes, your code and your difficult questions.
From a first-principles research discussion to a practical runtime lab and a look at the machine underneath.
What should a model remember? A focused discussion of bounded history, learned retrieval and the experiments that tell us whether a memory mechanism is useful.
Go from a checkpoint to a useful evaluation. Work through the runtime choices, run a representative prompt and learn how to measure a workload without losing the conditions.
What changes when weights, history state and execution are co-designed? Bring model and hardware perspectives into the same room, and test the assumptions between them.
Start with the report for the mechanism and evidence. Use the whitepaper for the broader research program, or the developer guide if you want to bring a running example.
We welcome ideas for technical deep dives, university reading groups and practical workshops. Send a short description of the audience and the question you want to explore.
Propose a sessionHelp shape the next Rock Matrix conversation.
