Siri AI's Skimmer: regex-only fast path calls tools up to 20× faster than LLMs
SumitGup · x · 2026-11-11
A teardown of Siri AI's on-device first stage, Skimmer: it uses regular expressions to match requests to deterministic tool flows, filling tool arguments without running any LLM. Measured: 'play' music tool calls in under 120ms (up to 6× faster than on-device/PCC models), simple tools like 'flip coin' under 60ms (up to 20× faster). Since no model runs, latency is consistent across devices and works offline; unmatched queries fall back to on-device/PCC models. The rules are browsable on macOS in Enigma.json under IntelligenceFlowPlannerSupport.framework.
More from Models
- OpenAI docs still mention "models 6.2 and later", hinting at unreleased GPT-6.2 — Angaisb_ · 2026-11-11
- Collaborator model Ampersand lands on Vercel AI Gateway, free for a week at 744B params — ericzelikman · 2026-11-11
- Same API price, 8x cheaper: Muse Ember builds voxel scene for $3.09 vs Haiku 5.5's $24.96 — alexandr_wang · 2026-11-11
- Anthropic's Fable 5.2 shows major robot-control gains, beats purpose-built systems on RoboDojo — DJiafei · 2026-11-11
- NVIDIA launches Alpamayo 2 Edge, first real-time Thor-ready reasoning model for vehicles — drmapavone · 2026-11-11
- OpenRouter officially joins Stripe, vowing to connect money rails with AI intelligence rails — jeff_weinstein · 2026-11-11