Cross-gate synthesis

---
title: Cross-gate synthesis — AI/Robotics/Macro futures, 17 gates, 1-25y horizon
status: draft
dimensions: ["food","food-availability","housing","utilities","education","childcare","metals","travel","labor"]
horizon: medium
trigger: (meta) Synthesis pass across all 17 individual gates after second-round research (May 24 refresh).
sub_gates: []
history: []
cross_gate: []
key_dependencies: []
last_updated: "2026-05-24T00:00:00.000Z"
sources_count: 0
---

## TL;DR

All 17 gates researched. P50s now span 2028–2049, with the heart of the distribution in 2029–2037 and a long-horizon macro tail at 2049 (the canonical 2× GWP marker). The single biggest movement this round is **humanoid pricing pulling forward by a year** — `humanoid-retail-20k` moved from 2029 to **2028** on the 1X Hayward 10k-unit five-day sellout, Figure 03 lease pricing, and the Helix-02 24-hour autonomous run. That compresses the entire humanoid stack: `humanoid-self-replication-factory` moved 2035 → **2034**, and the newly-added volume gate `humanoid-10m-households` lands at **2035** — meaning the price → factory → mass-volume sequence now compresses into a single seven-year window (2028→2034→2035) rather than the previously implied decade. The newly-added macro gate `global-economy-explosive-growth` brackets the entire portfolio: 1.5× by 2041, 2× by 2049, 3× by 2060, 6× by 2075 — with our four AI/robotics clusters being the proximate mechanism for the early tiers. The newly-added `home-grown-veg-beats-store` (P50 2042) is a competing cost curve against `cucumber-price-drop-80pct` (also P50 2042) — they're partial substitutes on the food-cost path, and P(both fire) is genuinely lower than P(either alone).

## Headline timeline table

| # | Gate | P10 | **P50** | P90 | Confidence | Δ this round |
|---|---|---|---|---|---|---|
| 1 | Humanoid robot < $20k retail | 2026 | **2028** | 2033 | medium | **−1yr** (was 2029) |
| 2 | AI agent 30% of knowledge work | 2027 | **2029** | 2034 | medium | — |
| 3 | Robotaxi profitable 5+ US cities | 2027 | **2029** | 2034 | medium | — |
| 4 | AI tutor K-8 parity < $20/mo | 2028 | **2031** | 2036 | medium | — |
| 5 | EV/robot BOM −30% via metals | 2028 | **2031** | 2037 | medium | — |
| 6 | eVTOL > 1k trips/day major city | 2029 | **2032** | 2038 | medium-low | — |
| 7 | SMR first OECD commercial | 2030 | **2032** | 2037 | medium | — |
| 8 | Solar + storage < $0.04/kWh | 2030 | **2033** | 2040 | medium | — |
| 9 | Autonomous freight + delivery 30%+ labor cut | 2030 | **2033** | 2040 | medium | — |
| 10 | Humanoid self-replication factory | 2031 | **2034** | 2042 | low | **−1yr** (was 2035) |
| 11 | Autonomous resource frontier ROI | 2030 | **2035** | 2045 | low | — |
| 12 | **Humanoid 10M households (cumulative)** | 2031 | **2035** | 2043 | medium | **NEW** |
| 13 | Cell-cultured beef price parity | 2032 | **2036** | 2045 | **low** | — |
| 14 | Construction-robot 40% labor cut | 2032 | **2037** | 2045 | **low** | — |
| 15 | **Home-grown veg beats store** | 2034 | **2042** | 2058 | medium | **NEW** |
| 16 | Cucumber retail −80% (Israel) | 2033 | **2042** | 2053 | low | P10/P90 tightened |
| 17 | **Global GWP 2× of 2026** (canonical) | 2036 | **2049** | 2080 | low | **NEW** |

Sorted by P50. The macro gate (#17) carries four sub-tier P50s: **1.5× by 2041 / 2× by 2049 / 3× by 2060 / 6× by 2075** — the canonical timeline row uses the 2× tier. Note **three low-confidence laggards** (cell-meat, construction-robot, autonomous-resource-frontier) and **three low-confidence long-horizon** entries (home-grown-veg, cucumber-Israel, GWP) — these are the gates with the widest P10-P90 windows.

## Year clusters

The expanded corpus produces **five** P50 clusters rather than the previous three. The new gates pull the macro frame outward and create a clear leading-cluster compression in 2028–2029.

### Cluster A: 2028–2029 — the AI/automation/humanoid crossing (compressed)

**Gates**: humanoid retail $20k (**2028**, ↓1yr), AI agent 30% knowledge work (2029), robotaxi 5-city profitability (2029).

What unifies them: all three are downstream of **long-horizon LLM/VLA reliability** (METR's 4-month doubling, GR00T N1.7, Helix-02) and **Chinese consumer-electronics manufacturing scale**. The humanoid pull-forward to 2028 means **the leading edge of this cluster is now one model generation closer**. Each gate has at least one OECD-relevant deployment crossing the threshold in 2026 (Klarna 67%, Waymo 500k rides/week in 11 metros, 1X Hayward 10k-unit five-day sellout, Figure 03 home lease at $400-600/mo). The 12-18 month lag from "demonstrated in one OECD deployment" to "broadly available / multi-instance" now closes in 2028 (humanoid) and 2029 (the cognitive/transport gates).

### Cluster B: 2031–2034 — the physical-infrastructure and packaging cluster

**Gates**: AI tutor K-8 parity (2031), metals-BOM −30% (2031), eVTOL major-city (2032), SMR OECD (2032), solar+storage <$0.04/kWh (2033), autonomous freight + delivery (2033), humanoid self-replication factory (2034, ↓1yr).

What unifies them: each one requires **physical-world infrastructure buildout or commercial packaging** to actually pass — vertiports, reactor builds, mining + REE separation, last-mile robot fleets, RCT-grade evidence for school procurement, robot-built-robot factories. The capability tech is mostly proven; the rate limits are permitting, fabrication-scale, and adoption friction. **Humanoid self-replication moved into this cluster** as the compressed humanoid stack now lands its supply-side milestone four years after price-clear.

**Internal dependencies in this cluster**: metals-BOM (#5 at 2031) remains upstream of eVTOL, robotaxi, autonomous freight (battery cost), construction robot, humanoid, and now also self-replication-factory and 10M-households. Solar (#8 at 2033) substitutes for SMR (#7 at 2032), which is why both can coexist — cheap solar covers the residential/distributed niche while SMR captures the AI/data-center 24/7 baseload niche. **The newest internal dependency**: humanoid self-replication (#10 at 2034) is a supply-side response to humanoid 10M-households demand (#12 at 2035) — whichever passes first compresses the other by 12-18 months.

### Cluster C: 2035–2037 — the volume and resource-frontier cluster

**Gates**: autonomous resource frontier ROI (2035), humanoid 10M households (2035, NEW), cell-meat parity (2036), construction-robot labor (2037).

The 2035 entries are the **volume successors** of the 2028–2029 Cluster A: once humanoids exist at $20k retail (2028) and factories can self-build (2034), the next milestone is 10M cumulative consumer units. The autonomous-resource-frontier gate is the parallel "humanoids deployed to empty land" extension of the same self-bootstrap pattern. The two low-confidence laggards (cell-meat, construction-robot) round out the cluster as the slower paths — cell-meat in funding winter (92% drop from 2021 peak), construction-robot bottlenecked on humanoid finish-trade dexterity.

### Cluster D: 2041–2042 — the slow food-cost and macro-leading cluster

**Gates**: home-grown veg beats store (2042, NEW), cucumber retail −80% Israel (2042), and (macro tier) world GDP +50% (2041, sub-tier of #17).

The two food gates are **competing cost curves**: industrial CEA scale (cucumber gate) versus home-scale CEA (home-grown gate). P(both fire) is genuinely lower than P(either alone) because the second is partly canceled by the first — if cucumber retail drops to NIS 1.70/kg, the home-grown value proposition collapses. The macro 1.5× tier at 2041 is the first measurable global-GDP signature of the AI/robotics productivity wave.

### Cluster E: 2049+ — the explosive growth tail

**Gates**: global GWP 2× (2049 canonical), 3× (2060), 6× (2075).

This is the long-horizon macro frame that brackets the entire portfolio. The 2049 canonical 2× tier corresponds to sustained ~3% global growth (historical baseline) — pre-AI extrapolation. The 2060 3× tier requires sustained ~6%/yr global growth, only historically achieved in compressed industrial-revolution-like periods. The 2075 6× tier requires transformative AI driving sustained 10%+ — the "intelligence explosion" scenario, where humanoid self-replication (#10) is the cleanest mechanical cause.

## Enables chains (cross-gate causation, expanded)

Re-reading the 17 gates' `cross_gate` blocks reveals a denser dependency graph than the v1 synthesis showed. The most decision-load-bearing chains:

```
                        metals-BOM (#5, 2031)
                        ─────────┬───────────
                                 │ enables (medium-to-strong) — now 7 downstream gates
            ┌─────────┬──────────┼──────────┬─────────┬─────────┬─────────┐
            ▼         ▼          ▼          ▼         ▼         ▼         ▼
      humanoid    robotaxi   construction  eVTOL  autonomous  self-      humanoid
      retail      (#3)       (#14)         (#6)   freight     replicate  10M HH
      (#1, 2028)  2029       2037          2032   (#9, 2033)  (#10,2034) (#12,2035)

                    humanoid-retail-20k (#1, 2028)
                        ─────────┬───────────
                                 │ enables (strong) — humanoid stack
                                 ▼
                    humanoid-10m-households (#12, 2035)
                                 │
                                 ▼ creates demand
                    humanoid-self-replication-factory (#10, 2034)
                                 │
                                 ▼ enables
                    autonomous-resource-frontier (#11, 2035)
                                 │
                                 ▼ enables (mechanism for explosive growth)
                    global-economy-explosive-growth (#17, 2049 canonical)

                    AI agent 30% (#2, 2029)
                        ─────────┬───────────
                                 │ strongly enables
                                 ▼
                    AI tutor K-8 (#4, 2031)

                    robotaxi (#3, 2029)
                        ─────────┬───────────
                                 │ strongly enables (~12mo lag)
                                 ▼
                    autonomous freight (#9, 2033)

                    cucumber-price-drop-80pct (#16, 2042)
                        ◀──────── substitutes ────────▶
                    home-grown-veg-beats-store (#15, 2042)
```

The **single biggest lever** is still metals-BOM (#5). It now enables 7 downstream gates (added: self-replication-factory, 10M-households). Its P50 of 2031 means accelerating it by 12-18 months would compress the entire Cluster B + Cluster C set substantially. **Watch the China REE export-control expiration on Nov 10, 2026** — that's the cleanest near-term catalyst.

The **second-biggest lever** is humanoid-retail-20k (#1, now 2028). Its pull-forward this round cascades into self-replication (2034), 10M households (2035), autonomous resource frontier (2035), and indirectly into the macro growth tiers — because humanoid self-replication is the cleanest mechanical pathway to >3× GWP per Carl Shulman's economic-takeoff framing.

The **clearest vertical chain** remains AI agent (2029) → AI tutor (2031) → kids' education economics. Now extended: → home-grown-veg (2042, weak) via consumer AI advisors that close the 5% plant-failure gap.

## Substitution loops (expanded)

Three pairs of gates partially substitute for each other:

**Solar + storage (#8) ↔ SMR (#7).** Unchanged from v1. Cheap residential / distributed solar reduces the addressable market for SMR baseload. Coexistence equilibrium: SMR for 24/7 AI/data-center, solar for residential.

**Robotaxi (#3) ↔ eVTOL (#6).** Unchanged. Robotaxi faster (2029 vs 2032), cheaper, no vertiport. eVTOL wins inter-metro and tourist-corridor markets. Tel Aviv ↔ Eilat / Mediterranean tourism is the credible Israeli eVTOL story.

**NEW — Cucumber-price-drop (#16) ↔ Home-grown-veg (#15).** Same P50 (2042), competing demand-side. If industrial CEA scale (greenhouse robotics + cheap energy + import liberalization) collapses cucumber retail 80%, the bar that home-grown systems must clear gets much lower — but the home-grown value proposition collapses faster, because store price drops from ~$6/kg to ~$1.20/kg while home-system amortized cost stays ~$8-15/kg. Conversely, if home-grown fires first, it caps retail margins on premium chains and applies modest substitution pressure on the cucumber gate. **P(both fire) is genuinely lower than P(either alone)**; they're partial substitutes competing for the same household budget line. Both gates are forecast objects, not action targets.

## Outcome bottlenecks per dimension (refreshed)

Per the spec's "earliest meaningful change = P50 of the earliest gate delivering ≥20% impact" rule, the *first* significant change to each outcome dimension lands as follows:

| Dimension | Earliest meaningful change | Gate(s) driving it |
|---|---|---|
| **Labor** (income / employment structure) | **2029** | AI agent 30% (#2) crossing in customer-support + SWE; phase change in role-by-role economics within 24–36 months |
| **Travel** | **2029** | Robotaxi profitability (#3) in dense urban cores |
| **Education** | **2031** | AI tutor K-8 parity (#4) |
| **Childcare** | **2028** (↓1yr) | Humanoid retail (#1) for in-home help; eldercare pressure makes this a $1k/month substitute for $5-8k/month aide. Pulled in by humanoid-retail-20k pulling to 2028. |
| **Utilities** | **2032** | SMR-first or solar parity, whichever lands first |
| **Metals** (commodity exposure) | **2031** | Gate #5 itself; the 30% BOM cut is the threshold for cost reductions to be visible in consumer goods |
| **Housing labor / cost** | **2028** (↓1yr) | Indirectly via humanoid (in-home services); direct construction-cost reductions wait until 2037 (gate #14) |
| **Food cost** | **2033** | Autonomous freight + delivery (#9) cutting logistics labor; cell-meat (#13) is the deeper reset but lands 2036+; cucumber and home-grown gates are 2042 |
| **Food availability** (global) | **2033** | Same as food cost; autonomous freight reaches underserved markets |
| **Household robotics volume** | **2035** (NEW) | Humanoid 10M households (#12) — the milestone where humanoids become "common" rather than "early adopter" |
| **Macro economy** (GWP) | **2041** (NEW) | World GDP 1.5× tier (sub-gate of #17); 2× by 2049, 3× by 2060 |

**Key insight refreshed**: the outcomes Tamir most directly cares about (labor income, kids' education, household cost of living) all see meaningful change *within 3-5 years* (by 2029–2031). **The childcare and housing-labor dimensions pulled in by a year** this round on humanoid-retail-20k acceleration. The food-cost reset is later than it looks — both food-specific gates (#15, #16) land 2042 with low confidence. The macro frame extends to 2049 (2×) / 2075 (6×).

## Calibration vs. external prediction markets (refreshed)

Where Metaculus / Manifold questions exist for our gate triggers (or close analogues), our P50s cluster as follows:

- **AI/Humanoid gates (1, 2, 3, 4, 12)**: our P50s of 2028-2031 are now **roughly at consensus** following the Q2 2026 Metaculus jump from ~30% to 79% on sub-$20k general-purpose robot by 2030. The humanoid pull-forward closed the prior contrarian-bullish gap.
- **Physical-infrastructure gates (5, 6, 7, 8, 9)**: our P50s of 2031-2033 are **roughly at consensus**. The crowd has digested the IRA / CHIPS Act / hyperscaler-PPA tailwinds well.
- **Self-replication and resource-frontier (10, 11)**: our P50s of 2034-2035 are **slightly more aggressive** than analyst consensus (Morgan Stanley / IDTechEx blend 2033-2036 for self-replication; PwC 2030s for resource-frontier with "10-15 years" qualifier). The Figure-confirmed-component-assembly + GR00T N1.7 dexterity scaling law pulled us in but the crowd hasn't fully caught up.
- **Lagging gates (13, 14)**: our P50s of 2036-2037 are **slightly behind** consensus, reflecting funding-winter (cell-meat) and embodied-dexterity-gap (construction).
- **Food long-tail gates (15, 16)**: our P50s of 2042 are **roughly at the bearish end of consensus**, reflecting Israeli structural stickiness and consumer-hydroponic plateauing.
- **Macro growth (17)**: our 2049 canonical (2× tier) is **roughly at Roodman/Davidson take-off consensus** (median 2× GWP 2045-2055 conditional on no AGI). The 6× 2075 tier is contrarian-bearish vs. AGI-by-2030 scenarios (which would compress to 2035-2040).

Bottom line: we're well-aligned with crowds on the easier-to-forecast physical gates, **less contrarian than v1** on AI gates (the crowd caught up), slightly contrarian-bullish on the humanoid self-replication / resource-frontier cluster, slightly contrarian-bearish on the food laggards and very long tail.

## Standing risks across the gate set (refreshed)

A handful of cross-cutting risks not captured at the individual-gate level:

1. **Regulatory backlash that hits multiple gates at once**. EU AI Act high-risk deadline now Dec 2, 2027 (pushed from Aug 2026 by May 2026 omnibus revision) + CA AI Act + state-level layoff-disclosure laws could simultaneously slow AI agent + AI tutor + autonomous-freight + robotaxi + humanoid by 12-24 months. EU Machinery Reg (Jan 20, 2027) + PLD (Dec 9, 2026) are humanoid-specific watch items.
2. **Labor / union resistance**. Teamsters (autonomous freight), teacher unions (AI tutor), building trades (construction robot), trial lawyers (AI agent in legal) could each individually slow a gate's adoption even when the tech is ready.
3. **Compute / energy bottleneck**. If hyperscaler power demand outstrips supply faster than SMR + solar can fill, frontier model training stalls and Cluster A gates slip 18+ months. SMR's P50 of 2032 vs hyperscaler demand growth (~25% YoY) is a key thing to watch.
4. **Geopolitics**. Taiwan disruption affects metals, AI chips, robot manufacturing. Iran-Israel tensions affect Israeli-specific gates. Both are wider-tailed than the gate set assumes.
5. **NEW — Actuator supply chain choke**. Harmonic Drive / Nabtesco / Leaderdrive precision actuator capacity is the binding constraint on the humanoid stack. 14-month lead times on standard product (May 2026), 80,000-unit precision gearbox shortfall projected by 2028. If Schaeffler's metal-forming actuator platform doesn't scale, the humanoid-retail-20k → self-replication chain slips 12-24 months and the 10M-households + autonomous-resource-frontier gates slip in proportion.
6. **NEW — Macro divergence risk**. The 2049 2× GWP P50 assumes the AI/robotics productivity wave actually translates to measurable growth. If the productivity gains are captured by capital (not labor) and concentrated in a small number of hyperscalers, measured GWP could lag the underlying technology curve significantly — the "productivity paradox 2.0" risk.

## Next steps (V3 / future work)

1. **Quarterly refresh** — re-run abbreviated research passes per gate to update P10/P50/P90 against fresh evidence. Next refresh due 2026-08-24. The 2026 May refresh moved 3 gates; pace of movement is roughly one significant P50 revision per gate per year.
2. **Astro static-site rendering** — gates now live in `src/content/gates/*.md` consumed by the Astro site; the TimelineStrip component renders P10/P50/P90 from frontmatter. Add a cross-gate visualization showing the dependency graph from the "Enables chains" section.
3. **Personal decision implications** — the per-gate personal/decision notes now live in a separate "Decision implications" view, kept out of the forecasts so they cannot influence the P10/P50/P90.
4. **Add downside gates** if the project scope expands — major climate-driven food shortage, AI-driven layoff wave hitting 5%+ of US white-collar in a single quarter, geopolitical disruption to Taiwan or Israel, AI-safety-driven moratorium. Each could shift the entire gate set.
5. **Calibration tracking** — every 6 months, score our P50s against any gates that have resolved (or against external prediction-market deltas). Aim for 50%+ of within-gate predictions landing inside our P10-P90 windows. The May 2026 humanoid-retail-20k revision is the first calibration data point — Metaculus moved from ~30% to 79% on sub-$20k-by-2030, and we moved our P50 from 2029 to 2028 in response.
6. **Add resolution criteria detail** for the macro gate (#17). The 2× / 3× / 6× tiers need IMF / World Bank methodology hooks to be tractable; today they're directional rather than measurable.

---

**File reference**: All 17 gate `.md` files live in `/Users/door2k/projects/futures/src/content/gates/`. Each has YAML frontmatter capturing P10/P50/P90 + cross-gate relations machine-readably for site / dashboard rendering. The 6 gates touched this round (May 24 refresh) are: `humanoid-retail-20k` (refreshed), `humanoid-self-replication-factory` (refreshed), `cucumber-price-drop-80pct` (refreshed), `humanoid-10m-households` (new), `home-grown-veg-beats-store` (new), `global-economy-explosive-growth` (new).