A 5-OOM (100,000x) effective-compute leap will occur between 2024-2027 — bridging GPT-4 high-schooler to fully automated AI researcher.

A 5-OOM (100,000x) effective-compute leap will occur between 2024-2027 — bridging GPT-4 high-schooler to fully automated AI researcher. | Epoch AI FLOP tracking; frontier model releases

Key catalyst: Epoch AI FLOP tracking; frontier model releases

Watch events: Frontier training FLOPs; published algorithmic-efficiency papers; emergent capability benchmarks.

Epoch AI tracking shows frontier training compute doubled ~every 6 months 2022-2025; algorithmic efficiency gains (MoE, distillation) material.

Beneficiaries (11)

{
  "nia": false,
  "qty": "5 OOMs (100,000x)",
  "mode": "FORECAST",
  "role": "Cited-Researcher",
  "caveats": "Unhobbling gains hardest to quantify; algorithmic efficiency may saturate.",
  "context": "Same size jump as GPT-2 → GPT-4 (5 OOM). Derived from three vectors: physical compute (~0.5 OOM/yr), algorithmic efficiency (~0.5 OOM/yr), and unhobbling (RLHF, CoT, tools, memory).",
  "to_year": 2027,
  "conv_cues": "specific quantitative target; derivation provided",
  "direction": "NUMERIC_TARGET",
  "from_year": 2024,
  "timeframe": "2024-2027",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "overdue",
      "weight": 0.05,
      "ordinal": -6,
      "source_id": null,
      "expected_date": "2024-09-11",
      "observed_date": null,
      "miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
      "miss_emitted_by": "metadata_milestone_sweep"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q2 window check-in (50%)",
      "status": "overdue",
      "weight": 0.05,
      "ordinal": -5,
      "source_id": null,
      "expected_date": "2025-05-24",
      "observed_date": null,
      "miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
      "miss_emitted_by": "metadata_milestone_sweep"
    },
    {
      "kind": "llm_pre_event",
      "label": "Test-time compute / reasoning OOM unlocked via o1, o3, R1",
      "notes": "HIT — o1 (Sep 2024), o3 (late 2024/early 2025), R1 (Jan 2026) shipped. Unhobbling axis validated.",
      "source": "Stockalarm Pro / Dwarkesh Patel — Aschenbrenner test-time compute call validated",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -4,
      "source_id": null,
      "confidence": 0.95,
      "source_url": "https://www.dwarkesh.com/p/leopold-aschenbrenner",
      "expected_date": "2026-01-31",
      "observed_date": "2026-01-31",
      "research_origin": "deep_research",
      "measurement_criterion": "Three+ frontier reasoning models shipped (o1, o3, DeepSeek R1) demonstrating Aschenbrenner's predicted unhobbling OOM"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q3 window check-in (75%)",
      "status": "overdue",
      "weight": 0.05,
      "ordinal": -3,
      "source_id": null,
      "expected_date": "2026-02-02",
      "observed_date": null,
      "miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
      "miss_emitted_by": "metadata_milestone_sweep"
    },
    {
      "kind": "llm_pre_event",
      "label": "Aschenbrenner 1GW per cluster prediction validated by 2026",
      "notes": "HIT — 1GW cluster milestone confirmed; 10GW under construction. Compute axis on track for 5-OOM stack.",
      "source": "Stockalarm Pro 'Situational Awareness Two Years Later' — '1 GW per cluster by 2026: hit'",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -2,
      "source_id": null,
      "confidence": 0.95,
      "source_url": "https://pro.stockalarm.io/blog/situational-awareness-two-years-later",
      "expected_date": "2026-04-30",
      "observed_date": "2026-04-30",
      "research_origin": "deep_research",
      "measurement_criterion": "Public confirmation of 1GW-class AI training cluster operational, validating Aschenbrenner's 'compute scaling 0.5 OOM/year' axis"
    },
    {
      "kind": "llm_pre_event",
      "label": "Frontier model demonstrates 1 full OOM effective compute over GPT-4",
      "source": "Epoch AI tracking / OpenAI, Anthropic, Google releases",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -1,
      "source_id": null,
      "confidence": 0.75,
      "source_url": "https://epochai.org/data/notable-ai-models",
      "expected_date": "2026-09-15",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2026-12-31",
        "from": "2026-06-01"
      },
      "measurement_criterion": "Public release of model with ≥10x effective compute vs GPT-4 (per Epoch AI FLOP estima
... (truncated)