Proposal identity
Title, type, sender, UTC date, and public source.
LMRuntime.com / Static review template
A copyable, public-safe packet for describing scope, evidence, resource cost, structural operators, claim boundaries, review triggers, no-op fallbacks, and exit conditions.
Static review scaffolding
Use this template to separate scope, evidence, cost, uncertainty, review authority, and a safe no-op fallback before asking the project to change.
Packet anatomy
The template can be shortened for a small correction, but structural changes should keep the cost and boundary fields.
Title, type, sender, UTC date, and public source.
Page, component, model format, platform, backend, and release gate.
Current limitation or opportunity, separated from interpretation.
Reproduction, source, expected result, actual result, and uncertainty.
The smallest useful, reversible implementation or documentation step.
Correctness, developer, documentation, or governance value.
Compute, memory, maintenance, review, governance, and uncertainty.
Other approaches and the current workaround.
Add, Retire, Merge, Reactivate, Freeze, or No-op.
Supported claims, prohibited implications, safety, and review triggers.
A safe preserve, defer, document-only, or route-elsewhere result.
The evidence that closes, advances, defers, or retires review.
Full packet
Keep assertions evidence-bounded. Do not include secrets, credentials, private customer data, private-network details, or public exploit details.
LMRuntime idea review packet
1. Proposal identity
- Title:
- Idea type:
- Submitted by:
- Date (UTC):
- Public source URL:
2. Runtime scope
- LMRuntime page or component:
- Runtime area:
- Model format:
- Platform or backend:
- Current release gate:
3. Current limitation or opportunity
[State the smallest concrete gap. Separate observed behavior from interpretation.]
4. Evidence and trace
- Reproduction or workflow:
- Supporting source:
- Expected result:
- Actual result:
- Known uncertainty:
5. Proposed minimal change
[Describe the smallest useful, reversible first step.]
6. Expected gain
- Correctness:
- Developer value:
- Documentation value:
- Governance value:
7. Resource and review cost
- Compute cost:
- Memory cost:
- Maintenance cost:
- Human review cost:
- Governance or public-claim risk:
- Uncertainty cost:
8. Candidate alternatives
- Alternative A:
- Alternative B:
- Existing workaround:
9. Structural operator
[Add / Retire / Merge / Reactivate / Freeze / No-op]
10. Claim and safety boundary
- Claims this proposal would support:
- Claims it must not imply:
- Security or privacy concerns:
- Human-review trigger:
11. No-op fallback
[Describe a safe “not yet,” route-elsewhere, document-only, or preserve-current-state result.]
12. Exit condition
[State what evidence would close, advance, defer, or retire this review.]
Important: this packet is review scaffolding only. Copying or sending it does not mutate LMRuntime source, canonical memory, release status, or public claims.Narrower reports
Accessibility and security reports have different evidence and disclosure boundaries.
Include URL, environment, assistive technology when relevant, expected result, observed result, and impact.
Accessibility guidance →Read the security page first. Do not publish exploit details. Route only the evidence necessary for private triage.
Review security guidanceA full packet may be unnecessary. Send the exact page, current wording, proposed correction, and supporting source.
Return to idea routing →