You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
- Source parity state: V41 opens prompt-program parity over V40 exhaustive testing, V38 inference correctness, and V39 commercial Reading readiness; source-side prompt and PromptPart rewrites are blocked until catalogue, benchmark, interpolation, registry, callsite, and return-type surfaces are specified and checked
13
13
- Notes companion: `BITCODE_SPEC_V41_NOTES.md`
14
14
- Delta companion: `BITCODE_SPEC_V41_DELTA.md`
@@ -147,6 +147,9 @@ The artifact may include ids, source hashes, counts, prompt surface ids, parser
147
147
148
148
Gate 7 applies the same prompt-program discipline to Conversation, tool-definition, MCP API, ChatGPT App, public API, Terminal summaries, and interface prompts.
149
149
It must preserve route authority, source selection policy, rich stream log disclosure, tool schemas, and parsed return types.
150
+
Gate 7 is closed by the package-backed `buildV41ConversationToolInterfacePromptRewrite` source, deterministic `.bitcode/v41-conversation-tool-interface-prompt-rewrite.json` artifact, `generate:v41-conversation-tool-interface-prompt-rewrite`, `check:v41-conversation-tool-interface-prompt-rewrite`, `check:v41-gate7`, protocol tests, focused prompt-part rewrites, and gate/canon workflow wiring.
151
+
The current source-safe rewrite report covers 9 rows and 60 passing source predicates for Conversation PTRR PromptParts, Terminal system prompt composition, rich execution-log prompt/result disclosure metadata, DocCodeToolPrompt composition, ToolPromptRegistry hierarchy, MCP API/public API tool schema posture, ChatGPT App action/tool prompt boundaries, Terminal/public summaries, V38 Conversation/tool parity, and Gate 2 through Gate 6 prompt-program dependency roots.
152
+
The artifact may include ids, source hashes, counts, prompt surface ids, parser target ids, tool schema ids, fixture ids, predicate verdicts, dependency roots, and redaction posture, but no raw prompt text, interpolated prompts, raw provider responses, protected source, private context, credentials, settlement private payloads, wallet private material, or unpaid AssetPack source.
150
153
151
154
## V41 Gate 8 Prompt Benchmark Report And Telemetry Integration
Copy file name to clipboardExpand all lines: BITCODE_SPEC_V41_NOTES.md
+1Lines changed: 1 addition & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -83,6 +83,7 @@ The artifact binds both Reading pipelines, all five Reading UX steps, Gate 2 pro
83
83
Bitcode is active at V40.
84
84
V41 now drafts the next canon: every PromptPart and composed Prompt must become catalogue-visible, benchmarkable, registry-bound, interpolation-checked, parser-typed, and source-safe.
85
85
The first priority is Reading through `ReadNeedComprehensionSynthesis` and `ReadFitsFindingSynthesis`; Conversation, tool-definition, and interface prompts follow with the same discipline.
86
+
Gate 7 currently applies that discipline to Conversation, tool-definition, MCP API, ChatGPT App, public API, Terminal, and other interface prompts with source-safe rows for route authority, source selection, rich stream-log disclosure, tool schemas, parsed return types, and no protected prompt/source/settlement/wallet/unpaid AssetPack payload serialization.
0 commit comments