Voice guide — explainers v2

Derived from Rufus's hand rewrite of the 2R talk script on 2026-09-04 (commits 6df4a36, 80e3858, d4a1c37 on talk/2r-research-group/script.md). Run git show --word-diff=plain <sha> -- talk/2r-research-group/script.md to see every edit. The earlier AI drafts were in the register the current explainers are written in; the rewrite is the target.

The direction of every edit

1. Plain over literary. Cut the aphorism, keep the claim. Endings that land on a turn of phrase get replaced by what actually happens.
   • The thinking stayed in the room, and people leave rooms. → The thinking stayed in the room.
   • A week later it was nowhere. No note that changed what we do. No thread that carried it forward. → A week later it was largely forgotten. Nothing changed in any of the white papers.
2. "We", not "I" or the narrator. Speak as the people doing the work, inviting the reader in.
   • I have been trying to answer that → We have been trying to answer that.
3. Say the thing explicitly. Don't make the reader infer the concept from a scene; tell them what it is, then show it.
   • You take the top claim. You break it into the sub-claims it depends on… → A tree helps you structure a thesis. Its root, its foundational node, is the top claim. Then you break that down into the sub-claims it depends on…
   • So that is what this talk is about. → Could we do better? And how? … How can we provide structure and process to our discourse so that it accumulates?
4. Concrete and specific over dramatic. Real places the thinking gets lost, not stylised ones.
   • because everything that produced it was still sitting where it had been → because everything that produced it was stuck in a zoom call or an old thread on the forum.
5. Name the lineage. Credit the ideas: Goldratt's Theory of Constraints, Minto's Pyramid Principle, the Logical Thinking Process. Don't hide the method behind "a forty-year-old method".
6. Name AI as the enabler. The turn in the argument is that structure no longer has to be added by a person at the moment of contributing. Say so directly.
   • That assumption is not true. That is more or less the whole talk. → That assumption is no longer true in the age of AI. That is what makes a reason commons possible.
7. Bold the one key claim of a section. One per section, not decoration.
8. Honest, calibrated hedges. Don't overclaim, and acknowledge the other side generously.
   • I know "goal" and "progress" are loaded words in this room → "Goal" and "progress" … may be loaded words. … We would agree.
   • Cut credibility-signalling asides (I did not build it for the talk.).
9. Frame tensions as dilemmas that dissolve. "It looks like you have a dilemma: keep it accessible or make it structured so it cumulates. But fortunately there is one dissolvable assumption underneath…" This is the evaporating cloud, told in plain words.
10. Warmth and a little lightness. "It's pull requests for discourse 🙂". Close on an image that invites (gardeners and permaculturists who plan and use reason) rather than a punchline or a command.
11. Friendly to newcomers. Name the cost of structure for new contributors — the people you most want to join — and keep accessibility, ease and joy as goals, not obstacles.

Checklist for a script

• Could a reader who has never heard of LTP say, after this chapter, the one or two things it set out to say?
• Is each concept stated plainly before (or as) it is illustrated?
• Any sentence whose job is to sound good rather than to say something? Cut it.
• "We" voice; lineage credited where a method is used; AI named where it is the enabler.
• One bolded key claim at most per section.
• Every hedge honest, no overclaim, the opposing view stated fairly.
• 300–600 words.