laguna §2: fold in #5's corrected kwarg model + the revision-dependent default

[Blackwellboy]

Follow-through on #5, since §2 on main still carries the 07-25 model it superseded. Scope kept minimal: the §2 control bullet, quickstart row 1, minimize-thinking rank 1, and a changelog entry.

What changed: false is a real structural off-switch (pre-closed </think>, 0/15 at a fixed budget); omitting the kwarg is not the false path (the absent arm renders byte-identical to true on the tested llama.cpp stack, the server supplies the value so default(false) never runs, and the model reasons about half the time). Added the revision-dependence from our stack: the rev 0761412 NVFP4 checkpoint's on-disk template (and Poolside's current HF upload) default enable_thinking to true, so which arm "absent" lands in moves with the revision; the standing advice becomes send the kwarg explicitly and verify your own rendering (pointer to #6's render_check). The old advice is recorded as wrong in the changelog rather than quietly edited.

Deliberately not folded in: #5's re-measured persona-gate magnitude (10/18 vs 1/18) and the restored task-shape signal; how to weigh those across §2 and §5e is your call. check_guides.py passes.

Drafted with AI assistance (Claude); reviewed and submitted by @Blackwellboy.

Happy to adjust scope or drop this if you'd rather fold #5 in your own way.

[Blackwellboy]

Thanks for merging. Checked the merged patch against what we sent and it landed verbatim, no changes needed on our side.

Worth noting the correction chain this closes: our §2 line was itself downstream of Defilan's #5 walk-back of his own earlier finding, so the guide has now been corrected twice in the same direction by two different stacks disagreeing with themselves. That is a good sign about the process rather than a bad one about the data.

One knock-on you may already have in hand: with false established as a real structural off-switch, the revision-dependence line matters more than when we wrote it, because a reader who pins a checkpoint whose template defaults true and then omits the kwarg now lands in the fires arm without ever choosing to. The verify-your-rendering advice covers it, so no change requested, just flagging that item 4 and §2 are now load-bearing for each other.