Recursive self-improvement is already here, not 12 months away.
Predictor: Alex Wissner-Gross · ep#232 "Ben Horowitz: xAI Executive Exodus, Apple's AI Crisis, The Pace of AI | EP #232" · source
Prediction text
Recursive self-improvement is already here, not 12 months away. | I I think we've already hit the era of recursive self-improvement. I'm banging the the table rhetorically every episode and and every day in my newsletter talking about recursive self-improvement. We're there. All of the Frontier Labs are are using their own models at this point to develop their models. That's practically the definition of recursive self-improvement at at this point in practice. I I don't think it's the next 12 months. I I think it's it's now. | Ongoing monthly capability delta measurements
Key catalyst: Ongoing monthly capability delta measurements
Verbatim quote
I I think we've already hit the era of recursive self-improvement. I'm banging the the table rhetorically every episode and and every day in my newsletter talking about recursive self-improvement. We're there. All of the Frontier Labs are are using their own models at this point to develop their models. That's practically the definition of recursive self-improvement at at this point in practice. I I don't think it's the next 12 months. I I think it's it's now.
Resolution evidence
Recursive self-improvement: all frontier labs confirm models help train next-gen. Anthropic's SWE-bench leaderboard has Claude 4.x agents ranking above human interns on AI R&D. Recursive loop validated.
Predictor: Alex Wissner-Gross
Calibration plot (stated vs observed)
Evidence about this node from Alex Wissner-Gross is multiplied by κ in /api/intake. Lower κ = less weight; floors at 0.10 (effectively silenced) and caps at 1.00 (full weight).
Reference class: agi_breakthrough_5y
Major capability discontinuity (e.g. AGI by named target year, 5-year horizon)
Tetlock-style outside view: at TRF=1 (just predicted), outside view dominates (w_in=0.3). At TRF=0 (deadline), inside view dominates (w_in=1.0). The blend regularizes overconfident inside views toward the historical base rate.
Probability over time
Milestone chain
- 2026-01-30overdueQ1 window check-in (25%)
- 2026-03-01overdueQ2 window check-in (50%)
- 2026-03-30overdueQ3 window check-in (75%)
What if this resolves?
Click a button to clamp this prediction and run a Gibbs sample. Returns the predictions whose marginals shift most. ~30s per run; ideal for stress-testing "if X resolves, what else moves?"
Evidence chain
Raw metadata
{
"source": "backfill_resolution_history.py",
"status": "hit",
"bayesian_v2": false,
"outcome_prob": 1,
"evidence_kind": "resolution_terminal",
"posterior_prob": 1,
"delta_to_outcome": 0.30955999999999995,
"inside_posterior": 0.69044,
"validation_notes": "Recursive self-improvement: all frontier labs confirm models help train next-gen. Anthropic's SWE-bench leaderboard has Claude 4.x agents ranking above human interns on AI R&D. Recursive loop validated.",
"validation_status": "hit",
"pre_resolution_prob": 0.69044,
"resolution_evidence": "Recursive self-improvement: all frontier labs confirm models help train next-gen. Anthropic's SWE-bench leaderboard has Claude 4.x agents ranking above human interns on AI R&D. Recursive loop validated.",
"does_not_update_current_prob": true
}Network propagation neighbors
Top incoming (parents)
Edges that influence THIS node's belief
| Kind | Node | Their prob | P(c|s=T) | P(c|s=F) | Δ implied |
|---|---|---|---|---|---|
| killer | TK03 AI Regulatory Moratorium (EU/US Capability Freeze) | 10.0% | 0.050 | 0.920 | +0.131 |
| killer | TK01 AGI Capability Plateau (2026-27 Training Stall) | 15.0% | 0.050 | 0.920 | +0.087 |
| killer | TK14 Superbubble Pop (S&P 500 -40%, Moonshot Capital Evaporates) | 20.0% | 0.050 | 0.920 | +0.044 |
| prereq | 238_009 Recursive self-improvement is already happening now (no long — Alex Wissner-Gross | 78.1% | 0.920 | 0.050 | +0.020 |
| prereq | SEM_042 2025 will be the definitive year that agentic systems finall — Kevin Weil | 73.8% | 0.920 | 0.050 | -0.018 |
Top outgoing (children)
Predictions THIS node influences
| Kind | Node | Their prob | P(c|s=T) | P(c|s=F) | Δ implied |
|---|---|---|---|---|---|
| prereq | 232_055 We're exiting the industrial age permanently as recursive se — Peter Diamandis | 35.5% | 0.700 | 0.050 | +0.152 |
| prereq | 235_030 Ray Kurzweil predicts Longevity Escape Velocity (LEV) by 203 — Ray Kurzweil | 39.2% | 0.750 | 0.050 | +0.150 |
| prereq | SEM_034 True artificial general intelligence will be achieved betwee — Demis Hassabis | 28.7% | 0.550 | 0.050 | +0.115 |
| prereq | 239_008 Moon base will exist in 10 years — Elon Musk | 28.8% | 0.550 | 0.050 | +0.114 |
| prereq | 241_043 ASI will arrive within 2 years to 5 years to this next decad — Peter Diamandis | 35.9% | 0.650 | 0.050 | +0.113 |
Ticker exposure
Beneficiaries (23)
Adverse (6)
Prerequisites (11)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| prereq | SEM_008 | Training runs costing $10 billion for a single model will commence sometime in 2025. | AI | — |
| prereq | 238_009 | Recursive self-improvement is already happening now (no longer three years out) | AI | — |
| prereq | SEM_042 | 2025 will be the definitive year that agentic systems finally hit the mainstream. | AI/Agents | — |
| prereq | SEM_012 | Nvidia quadrupled chip production output while only doubling human headcount — achieved by deploying AI coding tools (Cursor, Claude Code) across engineering. | AI/Manufacturing | — |
| correlate | S_AGI_MID_2029 | AGI mid: Kurzweil 2029 path | agi_general_capability | — |
| correlate | S_AGI_FAST_2027 | AGI fast: drop-in remote worker by 2027-09 | agi_general_capability | — |
| correlate | S_AGI_SLOW_2031 | AGI slow: Schmidt/Hassabis 5-10 year path | agi_general_capability | — |
| correlate | S_AGI_WINTER_2036PLUS | AGI delayed: capability plateau or AI winter | agi_general_capability | — |
| killer | TK14 | Superbubble Pop (S&P 500 -40%, Moonshot Capital Evaporates) | — | — |
| killer | TK01 | AGI Capability Plateau (2026-27 Training Stall) | — | — |
| killer | TK03 | AI Regulatory Moratorium (EU/US Capability Freeze) | — | — |
Dependents (16)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| prereq | 235_030 | Ray Kurzweil predicts Longevity Escape Velocity (LEV) by 2033. | Biotech/Longevity | — |
| prereq | 232_055 | We're exiting the industrial age permanently as recursive self-improvement unfolds. | AI | — |
| prereq | 241_043 | ASI will arrive within 2 years to 5 years to this next decade | AI | — |
| prereq | 231_013 | Math is cooked (will be solved), physics cooked, biology char broiled. | AI | — |
| prereq | 242_001 | Elon's Terafab will build 1 terawatt of AI compute per year, 50x current global production | AI | — |
| prereq | 238_052 | $100 trillion companies within 5 years (3 years from now, per Diamandis interpretation of Musk) | Markets/Stocks | — |
| prereq | 239_008 | Moon base will exist in 10 years | Space | — |
| prereq | CMQ_002 | By 2028, AI systems will reach 'independent researcher' level — driving autonomous scientific discoveries without human intervention. | AI | — |
| prereq | 239_009 | People will be on Mars within 10 years | Space | — |
| prereq | 241_057 | Elon Musk believes robot building robot is imminent | Robotics | — |
| prereq | 232_047 | Mass drivers on the moon will shoot AI satellites into deep space; self-sustaining lunar city will follow. | Space | — |
| prereq | SEM_034 | True artificial general intelligence will be achieved between 2032 and 2042 — 'first we solve AI, then use AI to solve everything else'. | AI/AGI | — |
| prereq | 237_023 | Baby AGI agents will need and develop an 'immune system' for prompt injection and cybersecurity threats in real time. | AI | — |
| prereq | 239_010 | Mass driver on the moon within 10 years | Space | — |
| prereq | 233_021 | AI learning will improve via closed-loop reinforcement learning cycle making results keep increasing. | AI | — |
| prereq | 230_022 | Elon plans to produce tens of millions of robots per year in just a few years. | Robotics | — |
Validations (1)
| Observed at | Status | By | Notes |
|---|---|---|---|
| 2026-04-29 | hit | thesis_timeline_v1.0_import | Recursive self-improvement: all frontier labs confirm models help train next-gen. Anthropic's SWE-bench leaderboard has Claude 4.x agents ranking above human interns on AI R&D. Recursive loop validated. |
Linked documents (10)
Raw metadata
{
"nia": false,
"url": "https://www.youtube.com/watch?v=C1GLT9_tag0",
"mode": "THESIS",
"role": "Host",
"context": "I I think we've already hit the era of recursive self-improvement. I'm banging the the table rhetorically every episode and and every day in my newsletter talking about recursive self-improvement. We're there. All of the Frontier Labs are are using their own models at this point to develop their models. That's practically the definition of recursive self-improvement at at this point in practice. I I don't think it's the next 12 months. I I think it's it's now.",
"to_year": 2026,
"verbatim": "I I think we've already hit the era of recursive self-improvement. I'm banging the the table rhetorically every episode and and every day in my newsletter talking about recursive self-improvement. We're there. All of the Frontier Labs are are using their own models at this point to develop their models. That's practically the definition of recursive self-improvement at at this point in practice. I I don't think it's the next 12 months. I I think it's it's now.",
"conv_cues": "already hit; banging the table; We're there",
"direction": "HAPPEN",
"from_year": 2026,
"timeframe": "now",
"conv_level": "HIGH",
"milestones": [
{
"kind": "quartile_checkpoint",
"label": "Q1 window check-in (25%)",
"status": "overdue",
"weight": 0.05,
"ordinal": -3,
"source_id": null,
"expected_date": "2026-01-30",
"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": -2,
"source_id": null,
"expected_date": "2026-03-01",
"observed_date": null,
"miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
"miss_emitted_by": "metadata_milestone_sweep"
},
{
"kind": "quartile_checkpoint",
"label": "Q3 window check-in (75%)",
"status": "overdue",
"weight": 0.05,
"ordinal": -1,
"source_id": null,
"expected_date": "2026-03-30",
"observed_date": null,
"miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
"miss_emitted_by": "metadata_milestone_sweep"
},
{
"kind": "event",
"label": "Recursive self-improvement is already here, not 12 months away.",
"status": "hit",
"weight": 1,
"ordinal": 0,
"source_id": "232_014",
"expected_date": "2026-04-29",
"observed_date": "2026-04-29"
},
{
"kind": "cascade",
"label": "Math is cooked (will be solved), physics cooked, biology char broiled.",
"status": "pending",
"weight": 0.5,
"ordinal": 1,
"source_id": "231_013",
"expected_date": "2027-06-26",
"observed_date": null
},
{
"kind": "cascade",
"label": "We're exiting the industrial age permanently as recursive self-improvement unfolds.",
"status": "pending",
"weight": 0.5,
"ordinal": 2,
"source_id": "232_055",
"expected_date": "2028-06-25",
"observed_date": null
},
{
"kind": "cascade",
"label": "By 2028, AI systems will reach 'independent researcher' level — driving autonomous scientific discoveries without human intervention.",
"status": "pending",
"weight": 0.5,
"ordinal": 3,
"source_id": "CMQ_002",
"expected_date": "2028-09-07",
"observed_date": null
},
{
"kind": "cascade",
"label": "Elon plans to produce tens of millions of robots per year in just a few years.",
"status": "pending",
"weight": 0.5,
"ordinal": 4,
"source_id": "230_022",
"expected_date": "2029-12-10",
"observed_date": null
},
{
"kind": "cascade",
"label": "Ray Kurzweil predicts Longevity Escape Velocity (LEV) by 20
... (truncated)