Exponential collapse in cost of intelligence shifts the primary economic bottleneck of human progress away from human cognition and toward computational throughput.

Exponential collapse in cost of intelligence shifts the primary economic bottleneck of human progress away from human cognition and toward computational throughput. | API pricing curves; cost-per-capability metrics

Key catalyst: API pricing curves; cost-per-capability metrics

Watch events: Frontier model API pricing per million tokens; cost-per-capability benchmarks.

API pricing for frontier models has dropped 10-100x in 2023-2026 per capability unit; inference-per-token cost at 1/100th GPT-4 2023 levels.

No probability history yet. The first evidence will arrive via /api/intake or the daily milestone sweep / weekly LBP run.

No propagation data yet. Run inference/.venv/bin/python scripts/ops/run_loopy_belief_propagation.py on the droplet, or wait for the Sunday 02:00 UTC weekly cron.

Adverse (5)

{
  "nia": false,
  "mode": "THESIS",
  "role": "Host",
  "context": "Implies cheap synthetic intelligence becomes the primary input factor in most knowledge-work industries within decade.",
  "to_year": 2030,
  "conv_cues": "framework claim; bottleneck shift",
  "direction": "HAPPEN",
  "from_year": 2026,
  "timeframe": "this decade",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "GPT-4-equivalent inference cost falls to $0.40/M tokens (1000x decline)",
      "notes": "HIT — direct empirical evidence of 1000x cost collapse in 3 years. Wissner-Gross's 'exponential collapse' thesis confirmed quantitatively.",
      "source": "GPUnex — AI Inference Economics: The 1,000× Cost Collapse Reshaping GPUs",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -8,
      "source_id": null,
      "confidence": 0.99,
      "source_url": "https://www.gpunex.com/blog/ai-inference-economics-2026/",
      "expected_date": "2026-02-01",
      "observed_date": "2026-02-01",
      "research_origin": "deep_research",
      "measurement_criterion": "GPT-4-class inference pricing drops to ~$0.40 per million tokens, ~1000x cheaper than late-2022 baseline of $20"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -7,
      "source_id": null,
      "expected_date": "2026-11-16",
      "observed_date": null
    },
    {
      "kind": "llm_pre_event",
      "label": "Inference cost declines 10x annually sustained",
      "source": "Silicon Data — Understanding LLM Cost Per Token: A 2026 Practical Guide",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -6,
      "source_id": null,
      "confidence": 0.75,
      "source_url": "https://www.silicondata.com/blog/llm-cost-per-token",
      "expected_date": "2026-12-31",
      "research_origin": "deep_research",
      "expected_date_range": {
        "to": "2027-12-31",
        "from": "2026-01-01"
      },
      "measurement_criterion": "OpenAI / Anthropic / Google publish API pricing showing sustained 10x year-over-year cost reduction for top-tier intelligence"
    },
    {
      "kind": "llm_pre_event",
      "label": "Compute throughput, not cognition, becomes named bottleneck in major economic forecast",
      "notes": "Required for the broader economic-bottleneck thesis (not just consumer pricing) to validate.",
      "source": "Anticipated — IMF WEO, OECD Economic Outlook, NBER working papers",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -5,
      "source_id": null,
      "confidence": 0.55,
      "expected_date": "2027-02-14",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2027-12-31",
        "from": "2026-04-01"
      },
      "measurement_criterion": "IMF, OECD, Goldman Sachs, or Brookings report explicitly identifies AI compute throughput (not human cognitive labor) as primary economic constraint on growth"
    },
    {
      "kind": "llm_post_event",
      "label": "Annual GPU shipment growth exceeds 50% YoY",
      "notes": "Cascade — confirms compute throughput is the binding constraint by showing the market signaling for it.",
      "source": "NVIDIA, AMD quarterly earnings; SemiAnalysis tracking",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -4,
      "source_id": null,
      "confidence": 0.65,
      "expected_date": "2027-02-14",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2027-12-31",
        "from": "2026-04-01"
      },
      "measurement_criterion": "NVIDIA + AMD combined data-center GPU shipments grow ≥50% YoY for at least 4 consecutive quarters"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q2 window check-in (50%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -3,
      "source_id": null,
      "expected_date": "2027-10-01",
      "observed_date": null

... (truncated)