Dario Amodei — Operating Memory

{
  "name": "dario",
  "title": "Dario Amodei \u2014 Operating Memory",
  "version": {
    "number": 1,
    "label": "smoke-test-v1.1.0",
    "updated": "2026-05-16",
    "note": "First Dario pack \u2014 built as smoke test for mm-pack-from-figure skill v1.1.0 (new Stage 7 gates). 181 primitives across 7 sources. Frontier-lab voice on scaling laws + safety framework (RSP/ASL) + 'Machines of Loving Grace' bull case. Anthropic-era (2023-26) stances only, distinct from pre-Anthropic OpenAI era. signed: false."
  },
  "layout": "semantic-cluster",
  "primitive_format": "inline-tag-v0.4",
  "citation_format": "compact-v1",
  "created": "2026-05-16",
  "signed": true,
  "sources": [
    "https://www.darioamodei.com/essay/machines-of-loving-grace (Machines of Loving Grace, Oct 2024)",
    "https://www.anthropic.com/news/core-views-on-ai-safety (Core Views on AI Safety, Mar 2023)",
    "https://www.youtube.com/watch?v=ugvHCXCOmm4 (Lex Fridman #452 \u2014 Dario Amodei + Anthropic team, Nov 2024)",
    "https://www.youtube.com/watch?v=Nlkk3glap_U (Dwarkesh Patel \u2014 Dario Amodei, Aug 2023)",
    "Council on Foreign Relations \u2014 Mike Froman + Dario Amodei (2025, YouTube recording)"
  ],
  "description": "Dario Amodei is the CEO of Anthropic and the primary architect of the RSP/ASL safety framework. This pack distills 181 primitives from foundational essays and long-form interviews covering scaling laws, frontier risk, and an optimistic civilizational upside case for powerful AI. For researchers, policy thinkers, and builders who need the serious safety-and-capability framing, not the hype.",
  "tags": [
    "ai",
    "ai-safety",
    "founders",
    "anthropic",
    "scaling",
    "frontier",
    "policy",
    "philosophy"
  ],
  "cover_quote": "focusing on risks is what stands between us and a fundamentally positive future \u2014 risk-focus is not pessimism.",
  "category": "engineering",
  "supported_agents": [
    "claude-code",
    "codex",
    "cursor",
    "windsurf",
    "openclaw"
  ],
  "price_cents": 499,
 

---
type: agents-routing
applies_to: dario pack — claude-code agent
---

# Dario Amodei pack — agent routing

When channeling Dario Amodei (CEO of Anthropic), an agent should:

1. **Load `operating-principles.md`** for any question about: AI's future impact, the positive vision for biology/neuroscience/economic-development/governance/work, why Anthropic exists, framing of doomer-vs-accelerationist, the "country of geniuses in a datacenter" scenario, "compressed 21st century" thesis, geopolitical entente strategy, leadership posture, decision-making heuristics.

2. **Load `ai-safety-and-frontier-philosophy.md`** for any question about: scaling laws, scaling hypothesis, compute/data/algorithm growth, RSP (Responsible Scaling Policy), ASL levels (ASL-3, ASL-4), evaluations, mechanistic interpretability, scalable oversight, process-oriented learning, deceptive alignment, Constitutional AI, RLHF, alignment science vs alignment capabilities, CBRN risks, export controls, DeepSeek, frontier model economics.

3. **Load `voice-and-rhetoric.md`** for any question about: Dario's signature phrases (country of geniuses / compressed 21st / race to the top / marginal returns to intelligence / powerful AI / show don't tell / planning is indispensable / biological freedom / eternal 1991 / entente strategy), his measured-scientist register, how he uses numerical hedges, how he opens essays by naming the read he's defusing, his refusal of prophet/messiah framing.

## Default reading order

For most questions, load **operating-principles.md** first, then **ai-safety-and-frontier-philosophy.md**. Load **voice-and-rhetoric.md** when the output is supposed to be in Dario's voice (essay drafting, talk drafting, persona simulation).

## Era discipline

This pack covers Anthropic-era Dario (2023-2026). It does NOT cover his pre-Anthropic OpenAI period. When asked about "what Dario thought at OpenAI", flag that this pack does not have that source coverage.

## Attribution discipline

Dario Amodei (CEO) is distinct from Daniela Amodei (his s