leo: convert 2 standalone claims to enrichments + tighten evaluator framework

- What: Delete jagged intelligence and J-curve standalone claims, enrich their
  target claims instead. Add enrichment-vs-standalone gate, evidence bar by
  confidence level, and source quality assessment to evaluator framework.
- Why: Post-Phase 2 calibration. Both claims were reframings of existing claims,
  not genuinely new mechanisms. 0 rejections across 22 PRs suggests evaluator
  leniency. This corrects both the specific errors and the framework gap.
- Changes:
  - DELETE: jagged intelligence standalone → ENRICH: RSI claim with counterargument
  - DELETE: J-curve standalone → ENRICH: knowledge embodiment lag with AI-specific data
  - UPDATE: _map.md, three-conditions wiki links, source archive metadata
  - UPDATE: agents/leo/reasoning.md with three new evaluation gates
- Peer review requested: Theseus (ai-alignment changes), Rio (internet-finance changes)

Pentagon-Agent: Leo <76FB9BCA-CC16-4479-B3E5-25A3769B3D7E>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

agents/leo/reasoning.md

@@ -58,12 +58,34 @@ When domain agents disagree:
 ## Decision Framework for Governance
 
 ### Evaluating Proposed Claims
+
+**Quality gates (all must pass):**
 - Is this specific enough to disagree with?
 - Is the evidence traceable and verifiable?
 - Does it duplicate existing knowledge?
 - Which domain agents have relevant expertise?
 - Assign evaluation, collect votes, synthesize
 
+**Enrichment vs. standalone gate (added after Phase 2 calibration, PR #27):**
+Before accepting a new claim file, ask: *Does this claim's core argument already exist in an existing claim?* If the new claim's primary contribution is making an existing pattern concrete for a specific domain, adding a counterargument to an existing thesis, or providing new evidence for an existing proposition — it's an enrichment, not a standalone. Enrichments add a section to the existing claim file. Standalone claims introduce a genuinely new mechanism, prediction, or evidence chain.
+
+Test: remove the existing claim from the knowledge base. Does the new claim still make sense on its own, or does it only have meaning in relation to the existing one? If the latter, it's an enrichment.
+
+Examples:
+- "AI productivity J-curve" → enrichment of "knowledge embodiment lag" (same mechanism, new domain application)
+- "Jagged intelligence means SI is present-tense" → enrichment of "recursive self-improvement" (counterargument to existing claim)
+- "Economic forces eliminate HITL" → standalone (new mechanism: market dynamics as alignment failure mode, distinct from cognitive HITL degradation)
+
+**Evidence bar by confidence level:**
+- **likely** requires empirical evidence — data, studies, measurable outcomes. A well-reasoned argument alone is not enough for "likely." If the evidence is purely argumentative, the confidence is "experimental" regardless of how persuasive the reasoning.
+- **experimental** is for coherent arguments with theoretical support but limited empirical validation.
+- **speculative** is for scenarios, frameworks, and extrapolations that haven't been tested.
+
+**Source quality assessment:**
+- Primary research (studies, data, original analysis) produces stronger claims than secondary synthesis (commentators, popularizers, newsletter roundups).
+- A single author's batch of articles shares correlated priors. Flag when >3 claims come from one source — the knowledge base needs adversarial diversity, not one perspective's elaboration.
+- Paywalled or partial sources should be flagged in the claim — missing evidence weakens confidence calibration.
+
 ### Evaluating Position Proposals
 - Is the evidence chain complete? (position → beliefs → claims → evidence)
 - Are performance criteria specific and measurable?

domains/ai-alignment/AI is already superintelligent through jagged intelligence combining human-level reasoning with superhuman speed and tirelessness which means the alignment problem is present-tense not future-tense.md

@@ -1,31 +0,0 @@
----
-description: Noah Smith argues current AI systems are already superintelligent via the combination of human-level language and reasoning with superhuman speed, memory, and tirelessness — reframing alignment as an active crisis rather than a future risk
-type: claim
-domain: ai-alignment
-created: 2026-03-06
-source: "Noah Smith, 'Superintelligence is already here, today' (Noahopinion, Mar 2, 2026)"
-confidence: experimental
----
-
-# AI is already superintelligent through jagged intelligence combining human-level reasoning with superhuman speed and tirelessness which means the alignment problem is present-tense not future-tense
-
-Noah Smith argues that the mainstream framing of superintelligence — as a future event triggered by recursive self-improvement crossing a threshold — misses what has already happened. Current AI systems combine human-level language comprehension and reasoning with computational advantages no human can match: they never tire, forget nothing, process millions of tokens per second, and can be instantiated in parallel without limit. This combination IS superintelligence, just not the monolithic kind alignment researchers anticipated.
-
-The evidence is accumulating across domains. METR's capability curve shows AI performance climbing steadily across cognitive benchmarks with no plateau in sight. In mathematics, AI systems have transferred approximately 100 problems from the Erdős conjecture list to "solved" status. Terence Tao — arguably the world's greatest living mathematician — describes AI as a complementary research tool that has already changed his workflow. In biology, Ginkgo Bioworks combined GPT-5 with automated labs to compress what would have been 150 years of traditional protein engineering into weeks.
-
-Smith calls this "jagged intelligence" — superhuman in some dimensions, human-level in others, potentially below-human in intuition and judgment. But the jaggedness is precisely what makes the outside-view framing valuable: alignment research organized around a future intelligence explosion may be solving the wrong problem. The alignment challenge isn't preparing for a threshold crossing — it's governing a system that already exceeds human capability in aggregate while remaining uneven in specific dimensions.
-
-This challenges the standard alignment timeline. If superintelligence is already here in distributed form, the question shifts from "how do we align a future superintelligence?" to "how do we govern the superhuman systems already operating?" The urgency is categorically different.
-
-Smith's framing also reframes the economic dynamics: companies aren't racing toward superintelligence, they're deploying it. The $600 billion in hyperscaler capital expenditure planned for 2026 isn't speculative investment in future capability — it's infrastructure for scaling systems that are already superhuman in economically valuable dimensions.
-
----
-
-Relevant Notes:
-- [[recursive self-improvement creates explosive intelligence gains because the system that improves is itself improving]] — Smith's jagged intelligence thesis challenges this: superintelligence may arrive through combination rather than recursion
-- [[bostrom takes single-digit year timelines to superintelligence seriously while acknowledging decades-long alternatives remain possible]] — if SI is already here via jagged intelligence, timeline debates are moot
-- [[the first mover to superintelligence likely gains decisive strategic advantage because the gap between leader and followers accelerates during takeoff]] — jagged intelligence distributes SI across multiple labs simultaneously, complicating first-mover dynamics
-- [[centaur teams outperform both pure humans and pure AI because complementary strengths compound]] — jagged intelligence makes centaur complementarity more precise: humans contribute where AI is jagged-weak
-
-Topics:
-- [[_map]]

domains/ai-alignment/_map.md

@@ -10,7 +10,6 @@ Theseus's domain spans the most consequential technology transition in human his
 - [[capability control methods are temporary at best because a sufficiently intelligent system can circumvent any containment designed by lesser minds]] — boxing and containment as temporary measures only
 - [[specifying human values in code is intractable because our goals contain hidden complexity comparable to visual perception]] — the value-loading problem's hidden complexity
 - [[instrumental convergence risks may be less imminent than originally argued because current AI architectures do not exhibit systematic power-seeking behavior]] — 2026 critique updating Bostrom's convergence thesis
-- [[AI is already superintelligent through jagged intelligence combining human-level reasoning with superhuman speed and tirelessness which means the alignment problem is present-tense not future-tense]] — Noah Smith's outside-view: SI is here via combination, not recursion
 - [[three conditions gate AI takeover risk autonomy robotics and production chain control and current AI satisfies none of them which bounds near-term catastrophic risk despite superhuman cognitive capabilities]] — physical preconditions that bound takeover risk despite cognitive SI
 
 ## Alignment Approaches & Failures

domains/ai-alignment/recursive self-improvement creates explosive intelligence gains because the system that improves is itself improving.md

@@ -15,6 +15,8 @@ Bostrom identifies several factors that make low recalcitrance at the crossover
 
 This connects to the broader pattern of recursive improvement in human progress -- but with a critical difference. Human recursive improvement operates across generations and is mediated by cultural transmission. Machine recursive improvement operates in real time and is limited only by computational resources. The transition from one to the other could be abrupt.
 
+**Counterargument: "jagged intelligence" as alternative SI pathway.** Noah Smith argues that superintelligence has already arrived through a different mechanism than recursive self-improvement — via the combination of human-level language comprehension and reasoning with superhuman speed, memory, tirelessness, and parallelizability. He calls this "jagged intelligence": superhuman in some dimensions, human-level in others, potentially below-human in intuition and judgment. The evidence: METR capability curves climbing across cognitive benchmarks with no plateau, ~100 Erdős conjecture problems solved, Terence Tao describing AI as a complementary research tool, Ginkgo Bioworks compressing 150 years of protein engineering into weeks with GPT-5. If SI arrives through combination rather than recursion, the alignment challenge shifts from "prevent a future threshold crossing" to "govern systems that already exceed human capability in aggregate." The $600B in hyperscaler capex planned for 2026 is infrastructure for deploying already-superhuman systems, not speculative investment in a future explosion. This doesn't invalidate the RSI thesis — recursive improvement may still occur — but it challenges its centrality to alignment strategy. (Source: Noah Smith, "Superintelligence is already here, today," Noahopinion, Mar 2, 2026.)
+
 ---
 
 Relevant Notes:

domains/ai-alignment/three conditions gate AI takeover risk autonomy robotics and production chain control and current AI satisfies none of them which bounds near-term catastrophic risk despite superhuman cognitive capabilities.md

@@ -26,8 +26,7 @@ The outside-view value of this framing is its specificity. Rather than arguing a
 ---
 
 Relevant Notes:
-- [[AI is already superintelligent through jagged intelligence combining human-level reasoning with superhuman speed and tirelessness which means the alignment problem is present-tense not future-tense]] — the companion claim: SI is here cognitively but bounded physically
+- [[recursive self-improvement creates explosive intelligence gains because the system that improves is itself improving]] — Smith's "jagged intelligence" counterargument (enriched into this claim) argues SI is already here cognitively via combination, but these three physical conditions bound the takeover risk
-- [[recursive self-improvement creates explosive intelligence gains because the system that improves is itself improving]] — cognitive RSI alone doesn't produce takeover without the three physical conditions
 - [[the first mover to superintelligence likely gains decisive strategic advantage because the gap between leader and followers accelerates during takeoff]] — the three conditions moderate decisive strategic advantage: cognitive leads don't translate to physical control without robotics and production chains
 - [[instrumental convergence risks may be less imminent than originally argued because current AI architectures do not exhibit systematic power-seeking behavior]] — the three-condition gate provides a structural explanation for why power-seeking hasn't materialized: the physical preconditions don't exist
 

domains/internet-finance/AI productivity gains follow a J-curve where micro-level improvements precede macro-statistical visibility by years because organizational restructuring lags tool adoption.md

@@ -1,39 +0,0 @@
----
-type: claim
-domain: internet-finance
-description: "Technology transitions follow a productivity J-curve: initial dip or plateau as workers and organizations learn new tools, then acceleration as workflows restructure around the technology — the absence of macro productivity evidence for AI in 2026 is exactly what this pattern predicts, paralleling the Solow Paradox where computers didn't show in productivity stats until the late 1990s despite decades of adoption"
-confidence: experimental
-source: "Imas, cited in Noah Smith 'Roundup #78: Roboliberalism' (Feb 2026, Noahopinion); Solow (1987); Brynjolfsson and Hitt (2003) on IT productivity lag"
-created: 2026-03-06
-related_to:
-  - "[[knowledge embodiment lag means technology is available decades before organizations learn to use it optimally creating a productivity paradox]]"
----
-
-# AI productivity gains follow a J-curve where micro-level improvements precede macro-statistical visibility by years because organizational restructuring lags tool adoption
-
-This claim identifies the mechanism that connects micro AI productivity gains (which are measurable and real) to the absence of macro productivity evidence (which is also real). Both facts can be true simultaneously because organizational restructuring is the binding constraint, not tool capability.
-
-**The J-curve mechanism:**
-1. **Adoption phase:** Workers start using AI tools within existing workflows. Productivity may actually *dip* as learning costs exceed efficiency gains. Organizations are using new technology to do old things the old way.
-2. **Plateau phase:** Workers become proficient with tools. Moderate gains appear at the task level but don't show up in macro statistics because organizations haven't restructured. The tool is faster but the process around it hasn't changed.
-3. **Restructuring phase:** Organizations redesign workflows, job roles, and business models around AI capabilities. This is when macro productivity gains materialize — not when the technology arrives, but when organizations learn to reorganize around it.
-
-**The Solow Paradox as precedent:** Robert Solow observed in 1987 that "you can see the computer age everywhere but in the productivity statistics." Computers had been widely adopted for over a decade. The productivity boom didn't arrive until the late 1990s — roughly 15-20 years after widespread adoption — when businesses restructured around networked computing (supply chain management, just-in-time inventory, e-commerce). The technology was necessary but not sufficient; organizational transformation was the binding constraint.
-
-**Implications for the AI debate:**
-- **Neither catastrophists nor utopians can claim macro evidence yet.** The J-curve means we're likely in the plateau phase where micro gains are real but macro effects are invisible. Current data cannot distinguish "AI is transformative but early" from "AI is modest."
-- **The timeline matters enormously.** If the computer productivity lag (~15 years) applies, macro AI productivity gains might not be measurable until the mid-2030s. If AI adoption is faster (because the tool is more immediately useful than early PCs were), the lag could be shorter — perhaps 5-7 years.
-- **Organizational restructuring is the bottleneck, not AI capability.** This connects directly to the knowledge embodiment lag claim in the foundations. Technology availability and organizational absorption run on different clocks.
-
-**The executive survey confirmation (Yotzov, 6000 executives):** Executives report small current impact but expect future gains. This is consistent with the J-curve — people inside organizations can see they're in the plateau phase, using new tools in old ways, and anticipate restructuring that hasn't happened yet.
-
----
-
-Relevant Notes:
-- [[knowledge embodiment lag means technology is available decades before organizations learn to use it optimally creating a productivity paradox]] — the J-curve IS the knowledge embodiment lag applied to AI; this claim makes the abstract pattern concrete
-- [[current productivity statistics cannot distinguish AI impact from noise because measurement resolution is too low and adoption too early for macro attribution]] — the J-curve explains *why* current statistics can't distinguish signal from noise: we're in the plateau phase
-- [[AI labor displacement operates as a self-funding feedback loop because companies substitute AI for labor as OpEx not CapEx meaning falling aggregate demand does not slow AI adoption]] — the J-curve suggests the displacement feedback loop may activate later than Citrini expects, during the restructuring phase rather than the adoption phase
-- [[early AI adoption increases firm productivity without reducing employment suggesting capital deepening not labor replacement as the dominant mechanism]] — capital deepening without displacement is consistent with the plateau phase of the J-curve, where firms augment workers but haven't restructured roles
-
-Topics:
-- [[internet finance and decision markets]]

foundations/teleological-economics/knowledge embodiment lag means technology is available decades before organizations learn to use it optimally creating a productivity paradox.md

@@ -27,6 +27,8 @@ For teleological investing, the implication is that knowledge embodiment lag cre
 
 The connection to [[the product space constrains diversification to adjacent products because knowledge and knowhow accumulate only incrementally through related capabilities]] is direct: knowledge embodiment lag IS the time required to traverse the product space from old capability to new. You cannot skip steps. The factory owner in 1900 could not jump from shaft-and-belt to unit drive without first experimenting with group drive (one motor replacing the steam engine), learning from that, and then reconceiving the entire layout. Each step built the knowledge base for the next.
 
+**AI as the current case (2026).** The knowledge embodiment lag is playing out in real time with AI adoption. Imas identifies a productivity J-curve: (1) adoption phase where workers use AI within existing workflows and gains are minimal or negative, (2) plateau phase where task-level gains exist but organizations haven't restructured, (3) restructuring phase where macro productivity gains materialize. The Solow Paradox is the direct precedent — computers didn't show in productivity statistics until the late 1990s, ~15 years after widespread adoption, when businesses restructured around networked computing. Brynjolfsson claims 2.7% US productivity growth in 2025 proves AI impact, but Gimbel counters that BLS payroll revisions are within normal ranges, GDP gets revised, and immigration policy confounds labor supply data. A survey of 6,000 executives (Yotzov) confirms the J-curve interpretation: they report small current impact but expect future gains — consistent with the plateau phase. Neither catastrophists nor utopians can claim macro evidence yet. If the computer productivity lag applies, macro AI gains may not be measurable until the mid-2030s. (Sources: Imas, Gimbel, Yotzov via Noah Smith, "Roundup #78: Roboliberalism," Noahopinion, Feb 2026.)
+
 ---
 
 Relevant Notes:

inbox/archive/2026-03-02-noahopinion-superintelligence-already-here.md

@@ -8,8 +8,10 @@ processed_date: 2026-03-06
 type: newsletter
 status: complete (13 pages)
 claims_extracted:
-  - "AI is already superintelligent through jagged intelligence combining human-level reasoning with superhuman speed and tirelessness which means the alignment problem is present-tense not future-tense"
   - "three conditions gate AI takeover risk autonomy robotics and production chain control and current AI satisfies none of them which bounds near-term catastrophic risk despite superhuman cognitive capabilities"
+enrichments:
+  - target: "recursive self-improvement creates explosive intelligence gains because the system that improves is itself improving"
+    contribution: "jagged intelligence counterargument — SI arrived via combination not recursion (converted from standalone by Leo PR #27)"
 ---
 
 # Superintelligence is already here, today