Skip to main content
Past couple of months (Oct - Dec, 2025), we focused on strengthening the context layer, improving performance and developer experience across the platform.

​
The milestones we reached

  • Ingestion Speed and reliability
    • We toiled at improving our ingestion speed as much as possible, and will continue doing so. We improved our upload times by around 37% for large documents, and by >70% over short conversations. Exact numbers may vary, and so do the speedups; but they are definitely there!
  • UI Revamp
    • We simplified navigation and workflows to make core features easier and faster to use. Onboarding experience is shorter and smoother, whether you’re a user trying to make ChatGPT remember stuff, or an experienced developer incorporating context to your AI agent!
  • Retrieval speed
    • Reduced ranked retrieval time (radius: 10k data points) by ~71% (from ~4s to ~1.14s)
    • Improved latency for fast mode by ~57% (from p50 ~400ms to p50 ~170ms)
  • Introduced Namespaces
    • Group and scope data under named namespaces for precise, controlled context retrieval.