AI scaling laws have not hit an asymptote yet

AI scaling laws have not hit an asymptote yet | the ultimate essentially scaling laws are not done yet. I keep asking my friends, when does the asymptote arrive, and when does the curve slow down? We've not seen it yes yet

the ultimate essentially scaling laws are not done yet. I keep asking my friends, when does the asymptote arrive, and when does the curve slow down? We've not seen it yes yet

Beneficiaries (23)

Adverse (6)

{
  "nia": false,
  "url": "https://www.youtube.com/watch?v=DpwmmXmzvfo",
  "mode": "THESIS",
  "role": "Guest-CEO",
  "caveats": "there will be a limit eventually",
  "context": "the ultimate essentially scaling laws are not done yet... We've not seen it yes yet",
  "to_year": 2026,
  "verbatim": "the ultimate essentially scaling laws are not done yet. I keep asking my friends, when does the asymptote arrive, and when does the curve slow down? We've not seen it yes yet",
  "direction": "UP",
  "from_year": 2026,
  "timeframe": "present / near future",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "GPT-5/5.2 ships with measurable benchmark gains over GPT-4 generation",
      "notes": "HIT — GPT-5 cleared 94.6% AIME and GPT-5.2 reached 100%, well beyond GPT-4 baseline, supporting Schmidt's no-asymptote claim.",
      "source": "https://www.adwaitx.com/ai-implementation-guide-2026-models-tools/ — GPT-5.2 100% AIME 2025, 40.3% FrontierMath",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -8,
      "source_id": null,
      "confidence": 0.98,
      "source_url": "https://www.adwaitx.com/ai-implementation-guide-2026-models-tools/",
      "expected_date": "2025-12-31",
      "observed_date": "2025-08-07",
      "research_origin": "deep_research",
      "measurement_criterion": "OpenAI releases GPT-5 (or GPT-5.x) achieving ≥90% on AIME 2025 and ≥40% on FrontierMath, demonstrating that scaling + RL-on-thinking is still producing capability gains"
    },
    {
      "kind": "llm_pre_event",
      "label": "Test-time compute / inference scaling becomes dominant capability lever",
      "notes": "HIT — confirms scaling continues but along a different axis (inference compute), not pretraining FLOPs.",
      "source": "Industry roundup of o1/o3/Gemini reasoning models, Epoch AI analysis",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -7,
      "source_id": null,
      "confidence": 0.95,
      "source_url": "https://epoch.ai/blog/can-ai-scaling-continue-through-2030",
      "expected_date": "2025-12-31",
      "observed_date": "2025-09-01",
      "research_origin": "deep_research",
      "measurement_criterion": "OpenAI o1/o3, Gemini 2.0/3, Anthropic Claude reasoning modes all release with explicit inference-compute scaling demonstrating Sharpe-like gains beyond pretraining"
    },
    {
      "kind": "llm_pre_event",
      "label": "Frontier labs report diminishing returns from naive parameter scaling",
      "notes": "PARTIAL — naive parameter scaling is hitting walls, but test-time compute / RL-on-thinking opens new scaling axes. Consistent with Schmidt's nuanced 'not the final asymptote yet' claim.",
      "source": "https://www.hec.edu/en/dare/tech-ai/ai-beyond-scaling-laws — diminishing returns from naive scaling at NeurIPS",
      "status": "partial",
      "weight": 0.4,
      "ordinal": -6,
      "source_id": null,
      "confidence": 0.85,
      "source_url": "https://www.hec.edu/en/dare/tech-ai/ai-beyond-scaling-laws",
      "expected_date": "2026-04-15",
      "observed_date": "2026-04-15",
      "research_origin": "deep_research",
      "measurement_criterion": "Industry analyst (NeurIPS, IEEE Spectrum, Epoch AI) explicitly documents 'scaling wall' for naive parameter scaling, even as test-time compute / inference scaling produces gains"
    },
    {
      "kind": "prereq",
      "label": "Nvidia quadrupled chip production output while only doubling human headcount — achieved by deploying AI coding tools (Cursor, Claude Code) a",
      "status": "hit",
      "weight": 0.5,
      "ordinal": -5,
      "source_id": "SEM_012",
      "expected_date": "2026-04-29",
      "observed_date": "2026-04-29"
    },
    {
      "kind": "prereq",
      "label": "Training runs costing $10 billion for a single model will commence sometime in 2025.",
      "status": "hit",
      "weight": 0.5,
      "ordinal": -4,
      "source_id": "SEM_008",
      "expected_date": "20
... (truncated)