# Futures — AI/Robotics → Cost of Living & QoL gated predictions

## Purpose
Tamir's structured forecasting project. 11 trigger-event gates (1–10y horizon) covering AI/robotics impact on consumer cost of living, family planning, global commodities, and travel. Each gate gets a 2–4h deep-research markdown writeup with P10/P50/P90 timeline, evidence synthesis, cross-gate dependencies, and decision implications for Tamir.

## Current State (2026-05-13)
- Design spec approved → `docs/superpowers/specs/2026-05-13-futures-gate-tree-design.md`
- Project skeleton + spec committed
- 11 gates selected (humanoid robot, AI agent knowledge work, solar+storage, robotaxi, AI tutor, cell-cultured meat, construction robot, metals/REE supply, eVTOL, SMR, autonomous trucks)
- Research not yet started

## Output Format
- **Canonical**: one `.md` file per gate in `src/gates/<slug>.md`. YAML frontmatter + structured body sections.
- **Derived**: Astro static site (V2) rendering each `.md` to an HTML page with a "Full markdown" raw view and a direct download link to the `.md`. Site deploys to Vercel.

## Key Decisions
| Decision | Reasoning | Date |
|---|---|---|
| Markdown (`.md`) as canonical, not MDX | Plain markdown is portable — pasteable into Notion, GitHub, slides; HTML is a *derived* view that links back | 2026-05-13 |
| 11 gates total, ~3h each | Tamir picked Deep over Sketch — quality > coverage | 2026-05-13 |
| Approach A — scenario-driven trigger events | Each gate is a measurable threshold; downstream impact essays connect to outcomes | 2026-05-13 |
| Astro over Next.js | Mostly prose with embedded interactive widgets; Astro's MDX-as-data is the sweet spot, ships less JS | 2026-05-13 |
| Defer visual widgets to V2 | Tamir: "research depth is most important. we can then work on presentation." | 2026-05-13 |

## Tamir's Preferences
- Research depth > polish at this stage
- Output must always have a text-only / raw markdown section available
- Iterations OK — start lean, deepen later

## Inter-gate Dependency Layers
1. **Sub-gate dependencies** (causal) — upstream prerequisites for each gate. Recorded in frontmatter `sub_gates:`.
2. **Cross-gate links inside the 11** — `enables`, `correlates`, `substitutes`. Recorded in frontmatter `cross_gate:`.
3. **Outcome co-dependence** — multiple gates feed the same outcome dimension. Aggregated at outcome-index pages.
4. **Synthesis pass** — after all 11 are done, write `cross-gate-synthesis.md` surfacing patterns.

## Known Issues / TODOs
- [ ] Astro project scaffold (deferred to V2)
- [ ] Vercel deployment config (deferred to V2)
- [ ] Research all 11 gates
- [ ] Write cross-gate synthesis
- [ ] Stale-source monthly check script

## Context for Claude
- Source data lives in `src/gates/*.md`. Treat the frontmatter as structured fields.
- Research uses these MCPs: `exa` (semantic web), `firecrawl` (structured scrape), `semantic-scholar` (papers + citation graph), `arxiv` (preprints), `reddit` (sentiment), plus WebFetch for Metaculus/Manifold.
- Frontmatter schema is in the design spec; follow it exactly — the eventual Astro build will validate it.
- When researching a new gate, dispatch one research subagent that runs all 5 evidence streams, then synthesize the .md.