Our autocomplete is aware of your recent changes, has full-codebase context, and is exceptionally fast (200ms P50 latency). We’ve deployed this across regions in us-west, us-east, and eu!
Because we have access to the IDE, we can also use the IDE’s code graph/index to automatically lookup definitions and references across your project. This helps the LLM see the exact definition rather than having to guess.
We’ve also shipped a strong coding agent that runs faster than Claude Code with minimal performance loss.
We’d love to get your thoughts and feedback!