v4 — original AI-led framing. AI/robotics central; macro, regulation, capital, sectoral structure as response variables. The framing the project started from. Anchored against documented June 2026 data.

Outcome present
Translucent = fewer runs (lower convergence)
Past
Not yet
▏ April 2026 — NOW

Stress test — sociopolitical primary. Political legitimacy, demographics, wealth distribution, institutional trust, geopolitics as primary drivers. AI is one variable, not the lens.

Outcome present
Translucent = fewer runs (lower convergence)
Past
Not yet
▏ April 2026 — NOW

Labour market projections, stress-tested across 5 framings. Each row is an outcome tested under v4 (AI-led), Sociopolitical, Macroeconomic, Cultural/values, and Systems-fragility framings — same 1,000 runs each. The convergence number is the MINIMUM across all 5 framings (the worst-case showing). Rows are grouped by tier: Stress-tested Robust (MIN ≥ 80%), High (MIN ≥ 60%), Framing-dependent (MIN ≥ 40%), Framing-fragile (MIN < 40%). Hover any row for per-framing breakdown.

Outcome present
Translucent = fewer runs (lower convergence)
Past
Not yet
▏ April 2026 — NOW

Stress test — macroeconomic primary. Interest rates, fiscal stance, capital concentration, inflation, trade as primary drivers. Tests v4 headlines against a macro-finance framing.

Outcome present
Translucent = fewer runs (lower convergence)
Past
Not yet
▏ April 2026 — NOW

Stress test — cultural/values primary. Generational AI acceptance, work-meaning, family structure, religious engagement, cultural polarisation as primary drivers.

Outcome present
Translucent = fewer runs (lower convergence)
Past
Not yet
▏ April 2026 — NOW

Stress test — systems-fragility primary. Cascade fragility, supply chain, cyber, pandemic, infrastructure as primary drivers.

Outcome present
Translucent = fewer runs (lower convergence)
Past
Not yet
▏ April 2026 — NOW

Calibrated for Denmark, Sweden, Norway, Finland. EU AI Act as given constant. Six Nordic-specific axes: Flexicurity response · Union negotiation · EU regulatory stance · Consumption feedback · Demographic buffer · Tech sector positioning. In 30/50 runs the flexicurity system holds but strains; in 6/50 runs it breaks entirely. The consumption spiral — Copenhagen/Aarhus housing markets, household debt, municipal kommuneskat shortfall — is the consistent second-order risk. Novo Nordisk and Maersk appear as the primary leading-edge cases across most runs. The "missed loop" per run identifies a blind spot the model's six axes cannot see.

Outcome present
Translucent = fewer runs (lower convergence)
Past
Not yet
▏ April 2026 — NOW
Speculative fiction — labour market focus. 5-year intervals 2026–2060. 1,000 runs. The subject is economic and social consequences: UBI, wealth distribution, where displaced workers go, post-work identity, and political response to displacement. AGI development is background context, not the main story. Worlds explored include The Managed Decline, The Species Split, The Affluent Grief, The Benign Panopticon, The Long Exhale, and 45 others.
Outcome present
Translucent = fewer scenarios
▏ April 2026 — NOW
Speculative fiction — not a forecast. This tab runs the cascade game at 5-year intervals from 2026 to 2060 across 1,000 scenarios, covering AGI, brain-computer interfaces, post-scarcity economics, biological enhancement, climate AI, and the far edge of human identity. It is structured imagination, not analysis. Worlds explored include The Abundance Paradox, The Velvet Enclosure, The Scorched Transition, The Long Opening, and 46 others. The serious projections are in the other tabs.
Outcome present
Translucent = fewer scenarios
▏ April 2026 — NOW

How to read this

Plain-language reference for the methodology. Read this first if the grid is confusing.

What you're actually looking at
A labour-market projection 2025–2030 tested under 5 different framings of the same question. Each framing was run 1,000 times. The default Aggregate view shows only outcomes that survive across all 5 framings — the project's most stress-tested output. This is a thinking instrument, not a forecast.
How to read the grid
Each row is an outcome. Each column is a six-month period — H1 2025, H2 2025, H1 2026 (the orange "NOW" line), H2 2026, and so on out to H2 2030. The 5-year-resolution long-horizon Civilisation grids use 2026–2060 instead. Filled cells mean the outcome shows up by that period in the runs where it fires at all. Empty means it hasn't arrived yet.

Row opacity does the heavy lifting: a row that fires in 950/1000 runs looks fully solid; 500/1000 looks distinctly translucent; 250/1000 looks ghostly. The fill colour just marks presence; the opacity tells you how often the outcome fired at all. The earlier Fast/Baseline/Slow track shading was dropped — it carried little signal once we moved from one-axis pace projections to nine-axis verticals where most outcomes are driven by domain-specific axes, not by AI pace alone.
What "470/1000" means
For each projection labelled "1000 runs", the number on the right tells you how many of the 1,000 simulated worlds produced that outcome. 470/1000 means 470 of 1,000 ran into it. The smaller range below (e.g. [44–50%]) is the 95% confidence interval — given a sample of 1,000, the underlying probability is most likely somewhere in that band. Outcomes in 850+ of 1,000 runs are treated as near-certain; 55–85% as high-probability; 20–55% as context-dependent.
Where the 1,000 runs come from
Each run picks a combination of nine variables — AI capability pace, robotics pace, macro environment, regulatory tone, sector concentration, energy supply, capital markets, sectoral structure, and decomposed China-parity status — plus a possible AI-incident profile. There are roughly 20,000 unique combinations; the model samples 1,000 strategically: 700 random (uniform across each variable), 200 corner-stress (extreme combinations like recession + tight capital + spiked energy), and 100 plausibility-weighted (April 2026 priors). For each run, a calibrated scoring function computes which outcomes follow.
Why it's called "cascade"
Each year's projection becomes the fixed ground for the next year's projection. You're not predicting 2028 directly — you're asking: given that 2026 played out this way, and 2027 followed from it, what does 2028 look like? This builds discipline. You can't retrofit later years to make a story work, because earlier years have already been locked. In v4 the cascade is also confidence-weighted — uncertain year-1 outcomes don't pretend to be certain when they roll into year 2.
What the model can't see
Each run produces two honesty checks. The named uncertainty is the single thing most likely to break that run's projection — collected across 1,000 runs, it becomes a register of recurring fragility points. The missed loop is a category of consequence the model's variables architecturally cannot surface — corporate tax base erosion, the informal care economy, the lived texture of middle-income displacement, things like that. Pooled, missed loops become the project's blind-spot register. The model is a frame; the missed-loop register is the model admitting which other frames it isn't.
Axes shape what's findable
Every projection in this site uses a chosen set of axes (typically 8–10) to vary across simulated worlds. That choice determines what the model can surface — and what it can't. Outcomes driven by dynamics outside the axis design simply don't appear, no matter how many runs we do. The missed-loop register partially addresses this by forcing each run to name one such blind spot, but the choice of axes is itself a judgment call that shapes every output. Treat the convergence map as "what's structurally robust under this framing" — not as a complete picture of what's coming.
What the convergence numbers mean (and don't)
"470/1000" is honest about its sample size, but it is not the same as "47% probability in the world." It means: under the chosen distribution of scenarios, with the chosen scoring function, 470 of 1,000 simulated worlds produced this outcome. The numbers are internally consistent, not independently validated. There is no fitted statistical model behind them — the scoring function is calibrated on judgment plus documented April–June 2026 anchors. Treat high-convergence outcomes (say 800+ of 1,000) as robust under the framing; treat low-convergence outcomes as conditional on specific combinations. The independence check (avg Jaccard ≈ 0.6) suggests runs vary meaningfully under identical parameters — so convergence reflects something real about the model, not just generator coherence. Whether it reflects something real about the world is a different and harder question.
Why these axes (and not others)
The axes determine what the model can find. With nine axes per vertical (plus decomposed sub-fields for China parity and AI incident), the parameter space is wide enough to surface interesting combinations — but every axis is itself a judgment about what dimensions matter. Different axes would produce different convergence maps over the same underlying reality. The missed-loop register is the partial corrective: each run explicitly names a category of consequence the model architecturally cannot surface, and those get pooled into a blind-spot register. Read the missed loops as the model admitting which other framings it isn't. Read the convergence map as what this framing produces. The framing is the unit of analysis, not the truth.

Glossary

Methodology terms are always shown. Domain-specific terms below update based on the selected projection.

Methodology · always shown
Convergence point
An outcome that appears across all three pace tracks in a multi-run projection. These are the most pace-independent findings — they occur regardless of how fast or slow the overall AI adoption curve moves.
95% Wilson CI
The confidence band on the convergence count. Given a sample of 1,000 runs, the true underlying probability is most likely somewhere inside the bracketed range (e.g. [44–50%]). Tighter than at n=100 by roughly √10.
Archetype
A named scenario type. Each run is assigned to the archetype it most closely fits based on its axis parameters. Runs that don't strongly match any are tagged Mixed. The archetype distribution is the project's "what kinds of futures does this parameter space produce" output.
Named uncertainty
Each run names one factor most likely to break the projection. Pooled across 1,000 runs, named uncertainties cluster into a register of recurring fragility points. The point isn't that the model is right — it's that it's honest about where it isn't.
Missed loop
Each run also names a category of consequence the model's axes architecturally cannot surface. Pooled, missed loops become the project's blind-spot register. The model admits which frames it isn't.
Cell opacity
Cells in a row fade based on convergence count. A row that fires in 950/1000 runs looks fully solid; a row at 500/1000 looks translucent; 250/1000 looks ghostly. The colour-state tells you when across pace tracks; the opacity tells you how often the outcome fired at all.
Calibration tier
Some verticals (Labour Market, Environment, Media as of May 2026) have outcome base probabilities anchored against documented April 2026 data. Others are calibrated to informed extrapolation. The further the vertical sits from the AI/robotics frontier itself, the less data exists to anchor the coefficients.

AI Landscape — Models & Tools

The competitive map as of April 2026. Market share estimates are approximate and shift quarter to quarter — Q1 2026 was the densest model release window the industry has seen.

Recent developments · April–June 2026

These developments triggered the June 14 anchor refresh: priors shifted toward Fast AI pace, Ahead robotics, Permissive regulation, Loose capital, and higher China-parity profiles.

Frontier language models
GPT-5.4 / 5.5 Instant
OpenAI
Leader
5.4: 1M token context, autonomous multi-step workflows. 5.5 Instant (May 2026): speed-optimised tier for high-throughput agentic workloads. Largest tracked enterprise API share (~50–55%). Microsoft partnership across Office, Azure, GitHub. Strongest agentic developer ecosystem. Confidential IPO filing June 8 at $852B last valuation.
Claude Fable 5 / Mythos 5 / Opus 4.8
Anthropic
Capability frontier
Mythos 5 is the internal capability frontier (limited availability, Project Glasswing). Fable 5 (June 9, 2026) is the safeguarded public release — SOTA on nearly all benchmarks, with hard blocks on cyber/bio/chem fallback to Opus 4.8. Opus 4.8 (May 28) added dynamic multi-agent workflows. ~22–26% enterprise API share. Anthropic confidential IPO filing at $965B valuation on $47B annualised revenue. Reference product for agentic dev (Claude Code, computer-use).
Gemini 3.1 Ultra / 3.5 Flash
Google DeepMind
Major
3.1 Ultra: 2M token context, native multimodal reasoning. 3.5 Flash (June 2026): refreshed cost-efficient tier with stronger reasoning per dollar. Tightly integrated with Workspace and Google Cloud. Strongest position in Google-ecosystem enterprises and consumer search. Default high-volume choice.
DeepSeek V4-Pro / V4-Flash
DeepSeek AI
Disruptor
V4-Pro / V4-Flash released April 24, 2026. Trails Western frontier "by 3–6 months" per their own admission — a structurally narrower gap than at V3 launch. Deep Huawei chip integration is the geopolitically critical signal: chip-side parity advancing. World-leading long context with drastically reduced compute/memory cost. Open weights. Near-free inference dominates cost-sensitive and SME use.
Llama 4 / 5
Meta AI
Open default
Open weights — self-hostable, fine-tuneable, GDPR-friendly. No direct revenue model but massive adoption. Powers a large share of white-label AI products and on-premise deployments for data-sensitive enterprises. Llama 5 approaching GPT-5-class on most benchmarks; the foundation of the open-source AI ecosystem.
Grok 4
xAI
Niche
Real-time X (Twitter) and web data access. Strong reasoning benchmarks. Integrated into X Premium. Differentiated by recency — knows what happened today. Smaller enterprise footprint but growing in finance and trading workflows where currency of information matters most.
Mistral Large 3 / Le Chat
Mistral AI
European
French company, GDPR-native, EU data residency. Open/closed hybrid model portfolio. Strong multilingual performance. The default compliance-safe option for European enterprises operating under strict AI Act enforcement. Strategic importance as the only significant Western European frontier model.
Phi-4 / Gemma 3
Microsoft / Google
Small models
Efficient small models for edge deployment, mobile, and on-device inference. Not competing with frontier models — competing with the cost and latency constraints that make frontier models impractical. Growing importance as inference moves from cloud to device. Phi runs well on consumer laptops; Gemma on Android.
Robotics & physical AI
NVIDIA GR00T
NVIDIA
Foundation
Open foundation models for humanoid robotics, launched 2026. Industry-standard physical-AI stack. Targets 30,000-units/year humanoid robot production. The Llama-equivalent for robotics — every humanoid manufacturer either uses it or trains against it. Jensen Huang: "AI has moved from experimental infrastructure to a core operating layer."
Figure / Optimus 3 / Atlas
Figure AI / Tesla / Boston Dynamics
Mass production starting
Tesla Optimus 3: production-intent prototype Q2 2026, mass-production start late 2026. Fremont line (former Model S/X) repurposed for 1M units/year. Gigafactory Texas Gen 2 line designed for 10M units/year long-term. Above the v3 anchor (30k units/yr humanoid output) by orders of magnitude — drove the robotics-axis prior shift to Ahead. Figure deployed in BMW; Atlas (electric) in Boston Dynamics commercial pilots. Cost $30k–$1M depending on capability.
Unitree / UBTech / 26 brands
Chinese humanoid cohort
Volume play
300+ humanoid robots competed in the Beijing half-marathon (April 19, 2026, 26 brands). The volume cohort — lower capability than Figure/Optimus but at $10–30k price points. Driving the cost curve down faster than Western projections assumed. Geopolitically: Chinese humanoid availability is the parallel to DeepSeek for physical AI.
Digit / Stretch / Spot
Agility / Boston Dynamics
Structured envs
Specialised non-humanoid robots already in production at scale. Digit deployed by Amazon in warehouses. Stretch handles parcels. Spot is the inspection-and-monitoring default for industrial sites. The "boring" robotics — fewer headlines, more deployed units, clearer ROI. Where the structured-environment automation wave actually arrives.
Developer & agentic tools
Claude Code
Anthropic
Agentic dev
Terminal-native agentic coding tool. Reference product for AI-as-junior-developer. Writes, tests, debugs, and deploys autonomously across multi-file changes. Defines what "agentic CLI" means for developer workflows. Fastest-growing AI dev tool of 2025–2026.
Cursor / Windsurf
Cursor AI / Codeium
AI-first IDE
AI-first code editors. Multi-model (Claude, GPT-5.4, Gemini). Deep codebase understanding — references your entire repo in context. Now agentic — Composer/Cascade modes execute multi-file refactors autonomously. Displaced VS Code + Copilot for individual developers and many startups.
GitHub Copilot
Microsoft / OpenAI
Enterprise code
Largest installed base for AI code assistance — enterprise contracts and IDE integrations across VS Code and JetBrains. Workspace and code-review agentic features now in production. Slower to innovate than Cursor/Claude Code but dominant in enterprise via Microsoft sales motion.
Codex CLI
OpenAI
Agentic CLI
OpenAI's terminal-native agentic coding tool — direct competitor to Claude Code. Tighter integration with the OpenAI ecosystem (o-series reasoning models, ChatGPT memory). Strong in teams already standardised on OpenAI APIs.
Knowledge work & creative tools
Perplexity
Perplexity AI
AI search
Citation-backed AI search. Real-time web access with source linking. Strong for research, competitor lookup, technical queries. Under pressure from ChatGPT Search and Google AI Overviews on the consumer end, but retains a loyal professional user base that values citation quality and Comet-class agentic browsing.
M365 Copilot / Notion AI
Microsoft / Notion
Productivity
AI embedded in existing workflows rather than as standalone tools. M365 Copilot integrates into Word, Excel, Teams, Outlook — where enterprise workers already live. Notion AI handles summaries, drafts, and database queries. The strategic importance: AI adoption without users having to change tools. High enterprise attachment rate.
Glean
Glean
Enterprise search
Enterprise search and agent layer for cross-tool retrieval. Indexes everything — Slack, Drive, Notion, Confluence, Salesforce — and surfaces answers with citations. Has become the default knowledge-management AI in enterprises with sprawling tool stacks. Adjacent to but distinct from chat assistants.
Sora 2 / Runway Gen-4 / Veo 3
OpenAI / Runway / Google
Video
Long-form coherent AI video at production quality. Sora 2 is the commercial benchmark. Runway Gen-4 is the production standard for editorial and ad work. Veo 3 leads on cinematic motion. Short-form commercial video production no longer requires a camera crew. Long-form narrative still uneconomic — that window is closing.
Midjourney V7
Midjourney
Image
Dominant image generation tool among creative professionals. V7 output indistinguishable from photography in controlled settings. Bootstrapped, highly profitable. Web-first interface (Discord legacy). No direct API — protects quality perception. Competitors: DALL-E 4 (OpenAI), Stable Diffusion (open), Adobe Firefly (IP-safe).
ElevenLabs
ElevenLabs
Voice
Leading voice cloning and TTS. Indistinguishable from human voice in most contexts. Widely used in content production, dubbing, accessibility, and conversational virtual agents. Real-time synthesis fast enough for live conversation. Deepfake voice misuse is the regulatory pressure point — provenance watermarking now mandatory in EU.
Suno / Udio
Suno / Uncharted Labs
Music
Dominant AI music generation — full vocal tracks, instrumentation, and arrangement from a text prompt. Suno V4+ approaching production-quality output. Used in advertising, social content, indie release pipelines. Active rights-and-training litigation with major labels is the headline regulatory question; commercial use carries unresolved IP risk.