Skip to content
Resource library / PDF guides
FIELD NOTES / EDITION 01

The inference field guide

Eight pages of worked serving decisions: KV reservations, prefill deadlines, cancellation, speculation, benchmark qualification and kernel investigation.

8 pages · Revised 2026-10-21 UTC · Worked synthetic examples
Sign in for download access

Online Premium includes the reader below. PDF downloads require active paid Premium or an eligible full-access grant; referral access covers online reading only.

Cover of The inference field guide with its contents and evidence standard

Keep the decision beside the evidence.

A compact reference for the serving path. Read the derivations online, take the PDF into a review, or work through the six-lesson serving course.

  1. Capacity and time
  2. KV ownership
  3. Prefill chunking
  4. Asynchronous lifetimes
  5. Speculative decoding
  6. Benchmark decisions
  7. Execution evidence

Read the guide

Worked decisions, ready to reference.

The online edition includes all seven sections and their derivations. It is included with Premium online access.

Read with Premium