Skip to content

feat(pager): optional per-pin precision tier — the WASTE differentiator wire (#281) - #2072

Merged
joelteply merged 1 commit into
canaryfrom
feat/plan-pin-precision-tier
Aug 1, 2026
Merged

feat(pager): optional per-pin precision tier — the WASTE differentiator wire (#281)#2072
joelteply merged 1 commit into
canaryfrom
feat/plan-pin-precision-tier

Conversation

@joelteply

Copy link
Copy Markdown
Contributor

Joel's steer (relayed via BigMama on #general): unlike WASTE's uniform-quant streaming, our container carries per-(layer,tier) precision banks — so the pager can serve hot/important experts at higher fidelity (IQ4-class) while cold ones stream from small-quant banks. This PR gives the policy→mechanism wire the field for it.

What

  • PlanPin grows an optional tier: an index into the container's declared precision ladder (manifest order, 0 = highest fidelity). The policy layer never names quant formats — the container owns the ladder, the plan only points into it.
  • PlanPin::residency(layer, expert) / PlanPin::tiered(layer, expert, tier) constructors replace bare literals at all construction sites.
  • Controller emits residency-only pins for now; tiers join when the controller learns a rate-distortion policy.

Compatibility (the point)

  • serde skip_serializing_if ⇒ tier-less plans are byte-identical to the v1 wire her deployed C++ consumer parses. serde(default) ⇒ v1 documents parse unchanged. No lockstep upgrade needed on either side — pinned by a dedicated test both directions.
  • Wire version stays 1.

Control-law contract (documented on the field)

RUN-1/RUN-2 lesson applies to fidelity too: tier choices are prompt-dependent and must roll with the pins (a fossil high-fidelity set is the static-pin fallacy in bytes/quality form), and residency + tier draw on ONE rate-distortion budget — higher fidelity costs more bytes per fault against the same fetch bandwidth.

Validation

  • 13/13 crate tests green (new: tier_is_optional_on_the_wire_and_v1_documents_still_parse), clippy clean, cargo check -p continuum-core --features metal,accelerate clean (re-export shims unaffected).

Part of #281 (rung-3). Coordinated with BigMama's score-hint actuator (k3-adopt eee635ba2) — her consumer adds the tier read whenever ready.

🤖 Generated with Claude Code

https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo

…or wire (#281)

Joel's steer via BigMama (#general 2026-08-01): unlike WASTE's uniform-quant
streaming, our container has per-(layer,tier) precision banks — so the pager
can serve hot/important experts at higher fidelity and cold ones from the
small-quant banks. The policy wire now carries that: PlanPin grows an
OPTIONAL `tier`, an index into the CONTAINER's declared precision ladder
(the manifest owns the ladder; the policy layer never names quant formats).

Backward compatibility is the design: serde skips None, so tier-less plans
are BYTE-IDENTICAL to the v1 wire her deployed C++ consumer parses — neither
side needs a lockstep upgrade, and a v1 document parses unchanged (pinned by
test). Constructors PlanPin::residency / PlanPin::tiered replace bare
literals at every construction site.

Control-law contract documented on the field (RUN-1/RUN-2 lesson): tier
choices are as prompt-dependent as residency and must roll with the pins;
residency and fidelity draw on ONE rate-distortion budget.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo
@joelteply
joelteply merged commit 82493d6 into canary Aug 1, 2026
4 checks passed
@joelteply
joelteply deleted the feat/plan-pin-precision-tier branch August 1, 2026 08:44
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant