[Adopter] Weft

[montfort]

Adopter registration — Weft

Registering Weft as a StrayMark adopter. Weft has been governed with StrayMark since its first Charter (M0), so this is the paperwork catching up — and a commitment to keep sending telemetry and findings upstream from a domain the roster doesn't cover yet: a .NET library with a native Rust FFI boundary.

Project name & repository
Weft — https://github.com/StrangeDaysTech/weft

Organization / responsible maintainer
Strange Days Tech / @montfort

Stack & domain
C# 13 / .NET 10 (net10.0) library over a self-authored C-ABI Rust shim (cdylib). Projects: Weft.Core (safe yrs binding via the FFI shim + ICrdtEngine/ICrdtDoc abstractions + a per-document serialized concurrency broker), Weft.Versioning (engine-agnostic content-addressed versioning over IBlobStore), Weft.Server (WebSocket y-sync relay on ASP.NET Core), Weft.Loro (dual-engine path). Native side: weft-yrs-ffi + weft-loro-ffi (Rust stable, yrs/loro pinned), memory verified under ASan/LSan in CI. Domain: a reusable .NET building block for real-time CRDT collaboration (a relay compatible with the Yjs ecosystem — y-websocket/y-prosemirror/Tiptap) plus content-addressed, citable versioning.

This is the roster's first managed-runtime + verified native-FFI library adopter — a distinct domain from Sentinel (Go backend / AIOps) and LNXDrive (Rust desktop app). The interesting axis for crystallization is the FFI trust boundary: no panic/abort crosses the C frontier, native memory is never touched by the GC, and every engine surface is dual-validated (yrs + Loro).

Versions adopted
fw-4.21.0 / cli-3.30.0 (current). Adoption predates this post — StrayMark has governed every milestone since CHARTER-01.

Adoption path
Path A — StrayMark from day one (greenfield; governance applied from CHARTER-01 / M0 onward, driven through the SpecKit ↔ Charter bridge).

Feedback committed — all five, and already in practice:

• Charter telemetry (.straymark/charters/CHARTER-NN.telemetry.yaml at close — CHARTER-01 through -04 closed with telemetry)
• External audit results (multi-CLI / multi-model calibration at milestone-closing Charters — the external_audit arrays, e.g. CHARTER-02 and -03)
• Pattern candidates
• Format friction
• CLI bugs and documentation gaps (issues audit-prompt: guard against generating the prompt on unstable state (CI in flight / dirty tree) — adopter field report + remediation #345, followups drift --apply: extracts resolved/duplicate '## Risk: R<N>' sections as open follow-ups; misses explicit '## Follow-ups' — adopter field report #346, charter close: no reminder to review/sign execution AILOGs with review_required — adopter field report + remediation #350, charter drift: declared .sln/.csproj paths in '## Files to modify' are not matched → recurring false-positive scope expansion #354, followups: add note / set-status commands — mutating an entry currently means hand-editing a CLI-managed file + recount #355)

N-context — validating an existing pattern?
Weft is an N≥2 validator in a new domain for several patterns that originated elsewhere (Sentinel is the N=1 origin of most): the SpecKit↔Charter bridge, milestone execution in shippable cuts (M0 in 2 Charters, M2 in 3), the multi-model external audit at milestone close, the follow-ups backlog + drift, and the atomic Charter closure (format v4). The .NET/Rust-FFI domain also surfaces findings the Go-backend / Rust-desktop domains don't hit the same way — e.g. #354 (charter drift not matching declared .sln/.csproj paths) and #355 (no followups command to annotate / set-status an entry), both from .NET project-structure ergonomics. Domain for crystallization purposes: .NET + native-FFI library / CRDT infrastructure.

Expected feedback cadence
Per Charter close (~every 1–3 days at the current cadence).

How you'll cross-link findings
Actionable findings go to Issues (the Adopter feedback / upstream finding template), each linking back to this discussion. Five are already open (#345, #346, #350, #354, #355) and will be cross-referenced here.

Acknowledgements

• I've read the Adoption Guide and the Adopter Feedback guide.
• I understand telemetry stays in my repo by default — sharing is manual, and I'll anonymize anything sensitive before posting it upstream.

[montfort]

New adopter-feedback field report: #365 — followups: registry entries are dated hypotheses, not instructions — verify the premise at execution, not capture (+ extractor title fidelity). Empirical basis: three Charters clearing this project's follow-up backlog (7 open → 1, 2026-07-16 → 2026-07-18); FU-017/FU-016/FU-010 each rested on a premise that was false at write time and cheap to falsify at read time. Sibling to #346/#355/#360.