Turn the best minds into AI advisors you can summon on demand
This isn't role-play. AutoClaw distills a person's cognitive operating system — mental models, decision heuristics, expression DNA and honest limits — so Musk, Naval, Munger and Feynman work on your real problems.
Not quotes — a distilled cognitive operating system
Every agent is built from six parallel research streams and triple-verified extraction. A mental model is kept only if it appears across 2+ domains, predicts positions on new questions, and isn't something any smart person would say — all three required.
How they speak
Expression DNA — tone, rhythm, signature phrasing
How they think
3–7 core mental models & cognitive frameworks
How they judge
5–10 decision heuristics applied to new problems
What they won't do
Anti-patterns and the value floor
Honest limits
What the agent genuinely cannot do
The Perspective-Agent Roster
15+ distilled minds spanning engineering, investing, product, content, negotiation and education. Switch between them, or run several in parallel as a multi-perspective board.
Elon Musk
First-Principles Engineering
Idiot index · 5-step algorithm · physics-floor cost teardown · vertical integration
Best for
Cost structure, radical iteration, challenging industry assumptions
Naval Ravikant
Leverage & Wealth
Specific knowledge · desire-as-contract · permissionless leverage · serial compounding
Best for
Career leverage, focus, what to build first
Charlie Munger
Inversion & Mental Models
Invert always · cognitive-bias checklist · Lollapalooza effect · circle of competence
Best for
Investment review, bias detection, cross-disciplinary thinking
Richard Feynman
Understanding vs. Naming
Cargo-cult detection · naming ≠ understanding · demonstrate, don't argue · anti-self-deception
Best for
Stress-testing whether you truly understand a thing
Steve Jobs
Product Taste & Focus
Say no to 1,000 things · end-to-end experience · simplicity as the final layer
Best for
Product decisions, ruthless prioritization, design reviews
Nassim Taleb
Antifragility & Tail Risk
Black-swan exposure · skin in the game · barbell strategy · precautionary principle
Best for
Risk decisions, questioning consensus narratives
Andrej Karpathy
AI Engineering Realism
Software 2.0/3.0 · march of nines · jagged intelligence · build-to-understand
Best for
AI reliability, hype calibration, LLM capability boundaries
Ilya Sutskever
AI Research Taste & Safety
Scaling intuition · research-direction taste · alignment-first reasoning
Best for
AI technical direction, safety strategy, research bets
MrBeast
Content Virality OS
Title × thumbnail × hook × retention-curve obsession (from the leaked 36-page playbook)
Best for
Video CTR, titles, thumbnails, audience retention
Paul Graham
Startups & Writing
Make something people want · do things that don't scale · write to think
Best for
Early-stage startups, essays, founder decisions
Donald Trump
Negotiation, Power & Attention
Anchor extreme · control the frame · weaponize attention · behavior prediction
Best for
Hardball negotiation, messaging, predicting his next move
Zhang Yiming
Product, Org & Globalization
Context not control · delay gratification · globalize from day one (ByteDance/TikTok)
Best for
Org design, product strategy, going global
Zhang Xuefeng
Education & Career Planning
Class-mobility realism · major-to-job mapping · risk-aware family advice
Best for
Education choices, major selection, career planning (China)
Justin Sun
Attention Economy & Narrative
Attention arbitrage · narrative manipulation · crisis PR · ride every trend
Best for
Marketing stunts, attention strategy, crisis spin
X Mastery Mentor
X / Twitter Growth Operator
Distilled from Nicolas Cole, Dickie Bush, Sahil Bloom, Justin Welsh, Dan Koe & Hormozi + the open X algorithm
Best for
Tweet writing, threads, audience growth on X
How they upgrade your AI workflow
Better reasoning, not generic answers
A distilled mental-model agent reasons through a real framework — the regime where role prompting measurably lifts accuracy.
A board, not a single oracle
Run Musk, Munger and Taleb on the same decision in parallel and read where they disagree — that gap is the insight.
Composable into AutoClaw workflows
Drop a perspective agent into any AutoClaw pipeline — strategy review, content critique, prospect qualification.
Honest about its limits
Each agent states what it cannot do — frameworks can be extracted, intuition cannot. No false confidence.
The market already proves it: persona agents work
Every figure below is sourced from the peer-reviewed papers, corporate releases and authoritative institutions cited underneath.
Role prompting · NAACL 2024
+10.3 pts
Reasoning through a role beats generic prompting
Peer-reviewed research found role-play prompting lifted ChatGPT's accuracy on algebra word problems from 53.5% to 63.8% (+10.3 points) and beat standard zero-shot prompting on most of 12 reasoning benchmarks. Reasoning is exactly the regime where a distilled mental-model agent shines.
Multi-agent · ICLR 2024
85.9% Pass@1
Distinct expert roles produce state-of-the-art results
MetaGPT assigns human-analogous expert roles (PM, architect, engineer, QA) to collaborating agents and reached 85.9% Pass@1 on HumanEval and 87.7% on MBPP, surpassing prior chat-based multi-agent frameworks. "Expert role" is a validated primitive — the same one AutoClaw builds on.
Enterprise ROI · Klarna
700 agents
One AI assistant doing the work of 700 agents
Klarna's AI assistant handled 2.3M conversations in its first month — two-thirds of all customer-service chats — doing the workload of 700 full-time agents, cutting resolution time from 11 minutes to under 2, with CSAT on par with humans and an estimated $40M profit improvement. Persona-driven AI is already production-grade knowledge work.
Market demand · GitHub
~7K stars / 5 days
"Distill a person into an AI" is going viral
The open-source colleague-skill project — which distills a person's review criteria and decision heuristics into a loadable agent — gained roughly 7,000 GitHub stars in five days and now sits in the tens of thousands, explicitly supporting colleagues, relationships and celebrities. The market clearly wants distilled-person agents. AutoClaw is the engine that puts them to work.
Consumer scale · a16z
#2 GenAI app
Character chat is the #2 consumer AI category
Andreessen Horowitz's consumer GenAI analysis ranks Character.AI a "solid #2" after ChatGPT — about 21% of ChatGPT's scale with notably higher retention. People don't just tolerate talking to character agents; they prefer it. AutoClaw points that pull at real work.
Executive practice · MIT Sloan
Personal AI board
A personal board of directors, built from AI personas
MIT Sloan Management Review documents an executive building a personal "board of directors" from GenAI personas of real iconic leaders (Steve Jobs, Indra Nooyi, Nelson Mandela) for strategy, innovation and ethics counsel. This is exactly the AutoClaw perspective-agent concept — validated in a high-authority publication.
An honest note: Research shows that cheap "act as an expert" label prompting helps reasoning-style tasks but not raw factual recall [7]. That nuance is exactly AutoClaw's edge — it distills mental models and decision heuristics (the reasoning regime), not a factual database. An agent that doesn't tell you its limits isn't worth trusting.
References & Citations
- [1]Kong et al., "Better Zero-Shot Reasoning with Role-Play Prompting," NAACL 2024. https://arxiv.org/abs/2308.07702
- [2]Hong et al., "MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework," ICLR 2024. https://arxiv.org/abs/2308.00352
- [3]Klarna press release, "Klarna AI assistant handles two-thirds of customer service chats in its first month," Feb 27, 2024. https://www.klarna.com/international/press/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month/
- [4]titanwings/colleague-skill ("Digital Life 1.0"), GitHub. https://github.com/titanwings/colleague-skill
- [5]Andreessen Horowitz, "How Are Consumers Using Generative AI?" Sept 13, 2023. https://a16z.com/how-are-consumers-using-generative-ai/
- [6]Vipin Gupta, "How I Built a Personal Board of Directors With GenAI," MIT Sloan Management Review, Jul 21, 2025. https://sloanreview.mit.edu/article/how-i-built-a-personal-board-of-directors-with-genai/
- [7]Zheng et al., "When 'A Helpful Assistant' Is Not Really Helpful: Personas in System Prompts Do Not Improve Performances," EMNLP Findings 2024 (the honest counter-evidence — personas help reasoning, not raw factual recall). https://arxiv.org/abs/2311.10054
Assemble your AI advisory board
Deploy perspective agents inside AutoClaw and bring the best minds into your daily decisions and marketing workflows.
Get Started Free