Capability map

What an intelligent assistant actually does

Elizabeth is tuned for depth over speed-theatre. Here is the behaviour you can rely on.

01 / Reasoning

Multi-step analysis

  • Breaks a problem into sub-questions before answering
  • States assumptions instead of silently picking one
  • Flags what it is uncertain about, briefly
  • Compares options in tables when tradeoffs matter

02 / Engineering

Code you can ship

  • Complete files, not fragments with ellipses
  • Typed interfaces, error handling and edge cases
  • Explains the architecture behind the diff
  • Reviews and refactors existing code on request

03 / Knowledge

Research synthesis

  • Compresses long material into decisions
  • Structured markdown: headings, lists, code, tables
  • Adapts depth to the question asked
  • Never pads an answer to look thorough

04 / Runtime

Streaming edge core

  • First token in ~40ms, streamed as generated
  • 128k context — long threads stay coherent
  • Model calls stay server-side; keys never ship
  • Stop generation any time, mid-stream

Test it with your hardest question

The console is live on the core page — no signup.

Open the console