Friction is information. The question is which kind.
What this doc does: it helps stride's maintainers decide what to do when a user complains that something in stride is annoying. Two possibilities — either the user is doing the wrong kind of work (stride is right to push back), or the gate is firing on a use case nobody designed for (stride should be calibrated). The doc names the question and gives a way to tell them apart.
Stride deliberately introduces friction —
/commit, the trace-back gate, the Vision-confirm step. Each one fires for a reason. When a user also hits that friction, two readings are possible, and they produce opposite actions.
The two readings
| Reading | When it's true | What to do |
|---|---|---|
| Friction-as-signal | The gate is correctly placed; the user's work doesn't fit the project | Push back on the work — it's out of scope, or VISION.md needs to evolve to include it |
| Friction-as-context-mismatch | The gate fires on a context it wasn't designed for; the user's work is legitimate | Calibrate the gate — add a mode, scope a condition, build an off-ramp |
Both look the same from the user's seat: "this isn't working, something's wrong." The maintainer's job is to tell them apart before deciding whether to weaken the gate or push back on the work.
How to tell them apart
The load-bearing question:
"Does this friction recur for users doing this kind of work, OR is this user doing the wrong kind of work?"
Two heuristics:
- Recurrence reveals reality — if the same friction surfaces across multiple users on legitimate work, it's context-mismatch. Calibrate.
- Singular drift is signal — if one user is hitting friction because their work doesn't trace back to Vision, the work is the problem, not the gate. Push back.
Default to signal. Earn context-mismatch with evidence. The asymmetric default keeps anchors honest — it's much cheaper to revisit a calibration that didn't earn its place than to recover an anchor that quietly eroded.
Worked example: the cross-project case
A user wanted to file an issue from repo foo into the Linear project for repo bar. /linear:plan-work ran its Vision check against foo's VISION.md — the work didn't trace, the gate fired, and the user was told to "add a criterion or drop as out of scope."
The first read was signal: "the gate is doing its job — your work doesn't fit foo." Push the user through.
But the user's work was legitimate — it just belonged in bar's project, not foo's. The gate fired on a context it wasn't designed for: cross-project quick-add. foo's Vision was correctly anchoring foo's work; it just wasn't the right Vision to check against for an issue filed into bar.
The maintainer-side move:
- Test for recurrence — would another user, doing legitimate cross-project work, hit the same friction? Yes — anyone with two stride-managed repos hits this on day one.
- Reading flips — the friction was context-mismatch, not signal. The gate needed calibration, not the user.
- Calibration, not removal —
/linear:plan-work --project <name>now skips the Vision check explicitly when filing cross-project, surfaces a one-line note explaining why, and defers the Vision anchoring to/linear:startrunning inside the target repo. The discipline is preserved (the work will be Vision-anchored when picked up); the gate is just no longer firing on the wrong surface.
The work did fit — into a different project. The Distinction made the call land cleanly: not a position to defend, a question to answer.
What this Distinction is not
- Not licence to weaken every gate at the first sign of friction. The default reading is signal. Context-mismatch requires evidence — recurrence + legitimate work — not just one user pushing back.
- Not a way to skip the trace-back rule. When friction-as-signal fires, the answer is still "push back, or add a criterion via
/vision" — not "call it context-mismatch and carve an exception." If you can't show recurrence and legitimate work, it's signal. - Not a substitute for
VISION.md. The Distinction tells maintainers which Principle to reach for; it doesn't replace the anchor itself. The Vision is what defines legitimate work. Without it, neither reading is grounded.