← Cockpit
238_034predictionAIAI-scaling

Software developers should build for AI agents, not humans — agents are the new consumers

Predictor: Andrej Karpathy · ep#238 "Meta Buys Moltbook, GPT 5.4, and Fruitfly Brain Upload | Moonshots Live at The Abundance Summit 238" · source

Prior probability
50.0%
Current probability
38.1%
evolves via intake + LBP
Conviction
3/5
Signal quality
D
Resolution
pending
Window
2026-04-30 – 2027-09-30
Edges in / out
6 / 0
Tickers exposed
37

Prediction text

Software developers should build for AI agents, not humans — agents are the new consumers | if you're building new software, you should target the agents. The agents are the new consumers... The agents are the new users of the social networks. If you're building something, don't build for humans. Build for the AI.

Verbatim quote

From episode "Meta Buys Moltbook, GPT 5.4, and Fruitfly Brain Upload | Moonshots Live at The Abundance Summit 238"
if you're building new software, you should target the agents. The agents are the new consumers... The agents are the new users of the social networks. If you're building something, don't build for humans. Build for the AI.

Predictor: Andrej Karpathy

κ + Brier as of 2026-05-22
κ (discount)
0.688
Brier
0.0067
excellent
Hits / Misses
3 / 0
of 3 resolved
Hit rate
100.0%
Calibration plot (stated vs observed)

Evidence about this node from Andrej Karpathy 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

Not linked

This node isn't linked to a reference class. The Bayesian update applies without outside-view blending.

Probability over time

4 prob_history rows
0%25%50%75%100%prior 50%2026-04-302026-05-032026-05-10
intake v2milestone miss sweeplbp propagationreference class assignedlegacy v1prior_prob (analyst seed)current = 38.1%

Milestone chain

Pre-event signals (upstream prereqs + window checkpoints) → resolution event → downstream cascades. Status/dates update from linked nodes; re-derive nightly via scripts/ops/derive_milestones.py.
Leading chain: 3 fired ✓ · 6 pending
  1. 2025-10-15hitKarpathy publicly articulates 'agents are the new consumers' thesis
    How: Karpathy publishes public statement/talk advising founders to build APIs, machine-readable docs, and target agents over humans (X posts, podcasts, conference talks)
    Source: https://www.news.aakashg.com/p/master-ai-agent-distribution-channelconf 90%
  2. 2025-12-15hitMCP reaches 97M+ monthly SDK downloads with 10,000+ active servers
    How: MCP downloads exceed 50M/month across Python+TypeScript SDKs and active MCP server count exceeds 5,000, per Anthropic/Linux Foundation Agentic AI Foundation reports
    Source: https://www.pento.ai/blog/a-year-of-mcp-2025-reviewconf 95%
  3. 2025-12-15hitAnthropic donates MCP to Linux Foundation Agentic AI Foundation; OpenAI/Google/MSFT/AWS/Cloudflare join
    How: Public press release confirms MCP donated to Linux Foundation as open standard with major frontier-lab adoption (OpenAI, Google DeepMind, Microsoft, Cloudflare)
    Source: https://www.pento.ai/blog/a-year-of-mcp-2025-reviewconf 90%
  4. 2026-07-27pendingQ1 window check-in (25%)
  5. 2026-10-24pendingQ2 window check-in (50%)
  6. 2026-09-01 → 2026-12-31pendingGartner: 40% of enterprise apps embed AI agents by end of 2026
    How: Gartner / IDC / Forrester publishes report confirming >=30% of enterprise applications have embedded task-specific AI agents (up from <5% in 2025)
    Source: https://finance.yahoo.com/news/40-enterprise-apps-embed-ai-181310288.htmlconf 70%
  7. 2026-06-01 → 2027-06-30pendingCascade: Major SaaS vendor publishes 'agent-first' API spec / agent.md standard
    How: Top-10 SaaS vendor (Salesforce, Stripe, Shopify, HubSpot, Notion, etc.) publishes dedicated agent-facing API documentation explicitly designed for autonomous AI consumption; press release or developer blog post
    Source: https://dev.to/blackgirlbytes/my-predictions-for-mcp-and-ai-assisted-coding-in-2026-16bmconf 65%
  8. 2027-01-21pendingQ3 window check-in (75%)
  9. 2026-06-01 → 2027-09-30pendingCascade: Agent-driven traffic exceeds 10% of API volume on a major dev platform
    How: Stripe, Cloudflare, Vercel, or similar publish public stats showing >=10% of inbound API requests originate from autonomous AI agents (User-Agent or auth-token classification)
    Source: https://www.ruh.ai/blogs/ai-agent-protocols-2026-complete-guideconf 40%

No downstream cascades — this prediction is a leaf in the dependency graph.

What if this resolves?

Clamp this prediction TRUE or FALSE and run a counterfactual Gibbs sample. Surfaces the predictions whose marginals shift most under that assumption.
(live posterior: 38%)

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

Every probability update with full Bayesian provenance — chronological, latest first
LBP2026-05-10T02:00:02Z38.1%-1.7pp
Network propagation: 39.8% → 38.1%
6-iter LBP, residual 0.00584 · damping 0.5, w_intrinsic 0.5 · method lbp_v3 · run e5c18d29
LBP2026-05-03T02:00:01Z39.8%-3.4pp
Network propagation: 43.2% → 39.8%
6-iter LBP, residual 0.00677 · damping 0.5, w_intrinsic 0.5 · method lbp_v3 · run 1a683ac9
LBP2026-04-30T16:39:51Z43.2%-2.3pp
Network propagation: 45.5% → 43.2%
5-iter LBP, residual 0.00825 · damping 0.5, w_intrinsic 0.5 · method lbp_v2 · run 0c8a4ea3
LBP2026-04-30T02:18:57Z45.5%-4.5pp
Network propagation: 50.0% → 45.5%
5-iter LBP, residual 0.00825 · damping 0.5, w_intrinsic 0.5 · method lbp_v1 · run 592311ef

Network propagation neighbors

Top edges sorted by latest LBP cross-impact
All propagation →

Top incoming (parents)

Edges that influence THIS node's belief

KindNodeTheir probP(c|s=T)P(c|s=F)Δ implied
prereqS_AGI_FAST_2027
AGI fast: drop-in remote worker by 2027-09
30.0%0.5000.050-0.196
killerTK03
AI Regulatory Moratorium (EU/US Capability Freeze)
10.0%0.0500.500+0.074
killerTK02
AI Compute Supply Shock (TSMC/Taiwan Disruption)
12.0%0.0500.500+0.065
killerTK01
AGI Capability Plateau (2026-27 Training Stall)
15.0%0.0500.500+0.052
killerTK09
Energy Grid Cap (Data Center Power Wall)
35.0%0.0500.500-0.038

Top outgoing (children)

Predictions THIS node influences

No outgoing edges.

Ticker exposure

37 ticker(s) linked

Beneficiaries (24)

MUWULFIRENEQIXALABAPLDASMIYASMLPLABNVDANBISCRWVAAPLAMTAMZNDELLGOOGLIRMLNVGYMETAMSFTORCLSFTBYSTX

Adverse (6)

ACNGENCHGGIBMWNSLRN

Prerequisites (6)

Predictions that must hit first
TypePredTitleDomainLag
prereqS_AGI_FAST_2027AGI fast: drop-in remote worker by 2027-09agi_general_capability
killerTK09Energy Grid Cap (Data Center Power Wall)
killerTK05Rate Regime Persistence (10y > 5% through 2028)
killerTK01AGI Capability Plateau (2026-27 Training Stall)
killerTK02AI Compute Supply Shock (TSMC/Taiwan Disruption)
killerTK03AI Regulatory Moratorium (EU/US Capability Freeze)

Dependents (0)

Predictions enabled by this
TypePredTitleDomainLag
No dependents

Linked documents (3)

Auto-generated by cosine similarity from Polymarket / Manifold / EDGAR / GDELT
SimSourceTitleMarket probPolarityReviewedPublished
0.674arxivCoding with "Enemy": Can Human Developers Detect AI Agent Sabotage?mentionspending2026-06-04
0.670arxivFrom Intent to Execution: Composing Agentic Workflows with Agent Recommendationmentionspending2026-05-05
0.662arxivBUILD-AND-FIND: An Effort-Aware Protocol for Evaluating Agent-Managed Codebasesmentionspending2026-05-07

Raw metadata

From Thesis_Timeline_v1.0_FINAL workbook
{
  "nia": false,
  "url": "https://www.youtube.com/watch?v=d__HRChE2ZE",
  "mode": "CITED_PREDICTION",
  "role": "Cited-Executive",
  "context": "The zeitgeist right now has this idea that increasingly Andre and others speak to this point that if you're building new software, you should target the agents. The agents are the new consumers... If you're building something, don't build for humans. Build for the AI.",
  "cited_by": "Alex Wissner-Gross",
  "verbatim": "if you're building new software, you should target the agents. The agents are the new consumers... The agents are the new users of the social networks. If you're building something, don't build for humans. Build for the AI.",
  "conv_cues": "should target",
  "direction": "HAPPEN",
  "timeframe": "Now",
  "conv_level": "MEDIUM",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "Karpathy publicly articulates 'agents are the new consumers' thesis",
      "source": "https://www.news.aakashg.com/p/master-ai-agent-distribution-channel",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -9,
      "source_id": null,
      "confidence": 0.9,
      "expected_date": "2025-10-15",
      "observed_date": "2025-10-15",
      "research_origin": "deep_research",
      "measurement_criterion": "Karpathy publishes public statement/talk advising founders to build APIs, machine-readable docs, and target agents over humans (X posts, podcasts, conference talks)"
    },
    {
      "kind": "llm_pre_event",
      "label": "MCP reaches 97M+ monthly SDK downloads with 10,000+ active servers",
      "source": "https://www.pento.ai/blog/a-year-of-mcp-2025-review",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -8,
      "source_id": null,
      "confidence": 0.95,
      "expected_date": "2025-12-15",
      "observed_date": "2025-12-15",
      "research_origin": "deep_research",
      "measurement_criterion": "MCP downloads exceed 50M/month across Python+TypeScript SDKs and active MCP server count exceeds 5,000, per Anthropic/Linux Foundation Agentic AI Foundation reports"
    },
    {
      "kind": "llm_pre_event",
      "label": "Anthropic donates MCP to Linux Foundation Agentic AI Foundation; OpenAI/Google/MSFT/AWS/Cloudflare join",
      "source": "https://www.pento.ai/blog/a-year-of-mcp-2025-review",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -7,
      "source_id": null,
      "confidence": 0.9,
      "expected_date": "2025-12-15",
      "observed_date": "2025-12-15",
      "research_origin": "deep_research",
      "measurement_criterion": "Public press release confirms MCP donated to Linux Foundation as open standard with major frontier-lab adoption (OpenAI, Google DeepMind, Microsoft, Cloudflare)"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -6,
      "source_id": null,
      "expected_date": "2026-07-27",
      "observed_date": null
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q2 window check-in (50%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -5,
      "source_id": null,
      "expected_date": "2026-10-24",
      "observed_date": null
    },
    {
      "kind": "llm_pre_event",
      "label": "Gartner: 40% of enterprise apps embed AI agents by end of 2026",
      "source": "https://finance.yahoo.com/news/40-enterprise-apps-embed-ai-181310288.html",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -4,
      "source_id": null,
      "confidence": 0.7,
      "expected_date": "2026-10-31",
      "research_origin": "deep_research",
      "expected_date_range": {
        "to": "2026-12-31",
        "from": "2026-09-01"
      },
      "measurement_criterion": "Gartner / IDC / Forrester publishes report confirming >=30% of enterprise applications have embedded task-specific AI agents (up from <5% in 2025)"
    },
    {
      "kind": "llm_post_event",
      "
... (truncated)