new file mode 100644
index 0000000..9ac402d
@@ -0,0 +1,57 @@
+---
+title: Open Questions — unresolved tensions
+type: concept
+visibility: public
+related:
+ - wiki/north-star-thesis.md
+---
+
+# Open Questions
+
+Unresolved tensions surfaced across the `#explainingideaflow` corpus. These are not bugs. They are where the vision is still negotiating with itself.
+
+## 1. Global stream vs. personal stream
+
+From 2025: Jacob wants *"everything I ever write or produce to end up in my thought stream"* — Slack, email, voice, docs. But the **global stream** is also a promised surface. How are negative filters expressed? How is *"non-explicit" / "raw"* flagged as a first-class node property vs. a list-membership property?
+
+Status: stated as an open architectural decision in 2025. No final call yet.
+
+## 2. Graph-first UX vs. outliner UX
+
+Taylor Mitchell's 2024 analysis: most tools-for-thought people use outliners (Workflowy, Roam) as their mental model, but IdeaFlow wants users to *feel* like they're working in a graph. The UI compromise (workflowy-as-scaffolding with graph semantics) is a tactical choice, not a final one. What does a graph-native UI look like once outliners aren't the onramp?
+
+## 3. Markup burden: still there, just moved
+
+The why-now claim is that LLMs close the markup gap. In practice, LLMs still produce wrong tags and invent nodes. The burden has *moved from the user to the review step.* Who reviews, when, and with what UX?
+
+## 4. Public vs. private at the personal layer
+
+If the long arc federates graphs, what's the default privacy? Node-level? Edge-level? Zone-level? The instinct has been "everything private by default, surface what you choose." But the 2025 alliance-graph framing (query the best intro path across friends' networks) requires aggressive opt-in sharing. Where's the right default?
+
+## 5. Enterprise wedge vs. consumer ambition
+
+Teams pay; individuals don't. But the enterprise wedge has its own gravity — it shapes the product toward Slack harvesters and CEO dashboards and away from *"a notebook for our minds."* How long before the wedge risks absorbing the project?
+
+## 6. The "God data structure" trap
+
+Jacob's maximalist framing (*"our God data structure should be able to contain everything in the world"*) is directionally right but culturally risky: it reads as hubristic. The wiki keeps the phrase because it's Jacob's own, but it's worth flagging that readers will judge it.
+
+## 7. Where does [Noos](https://globalbr.ai) sit?
+
+Jacob's Noos knowledge graph (separate product, separate wiki surface) overlaps significantly with IdeaFlow's personal layer. Are they the same thing at different stages, parallel products, or one being the other's substrate? The `#explainingideaflow` corpus doesn't fully answer this.
+
+## 8. Humanity 3.0 on what timeline?
+
+The long bet is civilization-scale coordination. The wedge is enterprise team-cognition. The gap is huge. No published roadmap ties them together with dates. How much should it?
+
+## 9. Speculative scientific claims
+
+Tagged in `#paperidea`-style messages: *"do this equipping [LLMs] with a hypergraph store."* These are load-bearing for the why-now story. But they're not substantiated in the corpus itself. Do they need their own "research page" in a future wiki?
+
+## 10. Voice vs. keyboard
+
+Thoughtstream started voice-first. The current product is keyboard-first with voice. The "dream-cap" speculation in vision-convo points even further (sleep-capture). What's the input hierarchy, and does it change by layer?
+
+---
+
+When one of these resolves, move it to an entry in [[chronology]] with a dated note.