The Problem We're Actually Solving
Most post-scarcity conversations jump to the endpoint: universal abundance, AI doing all the work, humans free to create and explore. That's the easy part to imagine. The hard part is the 20-year transition where:
See also: Transition Mechanisms — five proposed economic models for bridging the gap.
- AI eliminates jobs in waves, not all at once
- Prices for AI-augmented goods and services fall — but housing, healthcare, childcare, and food lag behind
- Federal UBI is politically gridlocked (and even if passed, probably insufficient)
- People need to eat, pay rent, and maintain dignity during the transition, not after
Mostaque's "Last Economy" gives us the destination. This conversation is about the bridge — practical, non-federal mechanisms that communities, cities, and individuals can build right now.
I. The Wage-Price Gap
The central danger isn't mass unemployment in the abstract — it's a timing mismatch. AI productivity gains flow to capital owners immediately. Cost savings reach consumers slowly. Wages for displaced workers drop to zero overnight.
White-collar displacement accelerates. Legal research, coding, copywriting, customer service, bookkeeping, radiology reads. The jobs that "required a degree" are first, not last.
The squeeze. Displaced workers compete for remaining human-required jobs (plumbing, elder care, construction). Wages compress downward. Meanwhile housing and healthcare costs haven't budged.
Prices start falling. AI-driven manufacturing, autonomous logistics, and synthetic biology begin lowering costs of physical goods. But the lag has already crushed millions of households.
Abundance arrives unevenly. Some sectors approach near-zero marginal cost. Others (land, energy, regulation-heavy services) remain expensive. The "last mile" of transition is messy.
How long is the gap? Is it 5 years or 25? Does the answer change the strategy?
Who gets crushed first — and are they the people least equipped to adapt?
II. Bridges That Don't Require Washington
If federal UBI is off the table (or years away, or inadequate), what else is there? Here are categories of intervention that cities, states, companies, nonprofits, and communities can deploy.
A. Municipal & State-Level Income Floors
- City-level guaranteed income pilots — already running in 100+ US cities (Stockton, Jackson, Denver). Results consistently show: people work more, not less; mental health improves; kids do better in school.
- State sovereign wealth funds — Alaska's Permanent Fund has paid dividends since 1982. Could states with AI-heavy economies (California, Texas, Washington) create similar funds taxing AI productivity?
- Local currency / time banking — communities issuing their own medium of exchange for care work, mentoring, local services. Mostaque's "Culture Credits" at the neighborhood scale.
What's stopping every city from running a guaranteed income pilot right now? Is it money, politics, or something else?
B. AI as a Public Utility
- Municipal AI services — free access to AI legal aid, tax preparation, medical triage, tutoring, job matching. The "public library" model applied to intelligence.
- Community AI cooperatives — shared compute resources, collectively owned. Members contribute data or compute cycles, receive services.
- Open-source AI sovereignty — Mostaque's "Sovereign AI Agent" idea: every person gets a capable AI assistant under their own control. Who pays for the compute?
If AI can do the work of a lawyer, doctor, accountant, and tutor — and the marginal cost is near zero — is there a moral argument for NOT making it a public utility?
C. Retraining That Actually Works
- The track record is terrible. Federal retraining programs have a dismal history (Trade Adjustment Assistance had ~37% placement rate). We need to be honest about this.
- AI-powered personalized education — the irony: the same technology displacing workers could provide the best retraining ever created. Adaptive, 1-on-1, available 24/7, free.
- "Retrain into what?" — the hardest question. If AI keeps advancing, what are you retraining people for? The answer might be: roles that require physical presence, human judgment, or human connection.
Is "retraining" just a polite way of saying "we don't have a real plan"? When the target keeps moving, does retraining become a treadmill?
D. Ownership Models That Distribute Gains
- Employee ownership of AI tools — companies that deploy AI share productivity gains with workers via equity, profit-sharing, or reduced hours at same pay.
- Data dividends — your data trains AI models. California tried (and failed) to pass data dividend legislation. Could a state or city succeed?
- Community ownership of automation — when a factory automates, the displaced workers own shares in the robots. Precedent: Mondragon cooperatives in Spain.
- AI productivity tax at the local level — cities taxing companies per "AI worker equivalent" deployed, funding local safety nets.
Could a single city become a laboratory for AI-era ownership models? What would it take to make one "symbiotic zone" (in Mostaque's language) actually work?
E. Reducing the Cost of Living Directly
- AI-optimized housing construction — 3D-printed homes, AI-designed modular housing, automated permitting. Austin is already seeing $100K homes from AI-optimized builders.
- Community land trusts — take land out of the speculative market entirely. Housing costs are mostly land costs. CLTs exist in 300+ US cities.
- AI-driven healthcare cost reduction — AI diagnostics, drug discovery, and administrative automation could cut healthcare costs 40-60%. But only if regulatory capture doesn't block it.
- Local food production — AI-managed vertical farms, community gardens with AI optimization, automated local food distribution.
Instead of giving people more money, what if we made the essential things cost dramatically less? Which is more achievable?
III. The Hard Questions
Things we should wrestle with honestly, without defaulting to optimism or pessimism.
On Human Purpose
Mostaque says jobs bundle five things: income, identity, community, purpose, structure. Even if we solve income, what replaces the other four?
Is the "crisis of meaning" a real risk, or is it a projection by people whose identity is their work? Most humans throughout history didn't define themselves by their jobs.
On Political Feasibility
The people most threatened by AI displacement are also the least politically organized. Who advocates for them?
Does the transition happen fast enough that politicians can't ignore it? Or slow enough that they boil-the-frog their way through?
On Concentration of Power
If 5 companies control the most capable AI, does anything else matter? Can local solutions work when the technology is centrally held?
Open-source AI (Llama, Mistral, DeepSeek) is currently competitive. Is that a temporary blip or a durable counterweight?
On What We Can Do Right Now
If we had $10M and one city willing to experiment, what would we build first?
What's the minimum viable "symbiotic zone"? A neighborhood? A company? A school district?
What can individuals do TODAY to prepare — for themselves, for their families, for their communities?
IV. A Practical Action Menu
Not policy prescriptions — things real people and communities can start doing now.
For Individuals
- Learn to use AI tools deeply — not casually. The gap between "I've tried ChatGPT" and "I use AI to 10x my output" is enormous and growing.
- Build skills AI can't replace (yet) — complex physical work, caregiving, creative direction, community building, leadership, taste.
- Reduce fixed costs — the less you need to earn, the more resilient you are. Housing is the biggest lever.
- Build local relationships — when systems fail, communities catch people. Social capital is the original safety net.
For Communities
- Start a mutual aid network — formalized neighbor-helping-neighbor with AI tools for coordination.
- Advocate for municipal AI services — free AI-powered legal aid, healthcare navigation, education support at the library.
- Create local economic resilience — community land trusts, time banks, local currencies, cooperative businesses.
- Run local guaranteed income experiments — even small ones generate data and political will.
For Leaders & Entrepreneurs
- Build businesses that distribute AI gains — profit-sharing, worker ownership, reduced hours at same pay.
- Create AI-powered public goods — the equivalent of Carnegie's libraries for the AI age.
- Lobby for local AI policy — don't wait for Washington. City councils and state legislatures move faster.
- Fund transition research — we need better models of how the wage-price gap unfolds and what interventions work.
V. Pre-Reading
Context for the conversation, drawn from the full reading list:
- Emad Mostaque, The Last Economy (2025) — the primary text. Focus on chapters on the Dual Currency System and Nucleation Strategy.
- Dario Amodei, "Machines of Loving Grace" — the optimistic case, with specific mechanisms.
- Leopold Aschenbrenner, "Situational Awareness" — the urgency argument. Why the timeline might be shorter than we think.
- Daniel Susskind, A World Without Work — the most rigorous treatment of what happens to labor markets.
This guide will be updated after the conversation with a summary and any conclusions reached.
VI. Who's Building the Bridge? A Landscape
Companies and projects working on decentralized AI, alternative economic models, or infrastructure for the post-labor transition. Roughly ordered from most to least mature.
Open-Source AI Platforms
- Hugging Face — The GitHub of AI models. Hosts 500K+ open models, datasets, and apps. No economic layer or token — just making AI accessible and open. The most successful "democratize AI" effort to date, but doesn't address the income/transition problem.
- Together AI — Open-source model hosting and training infrastructure. Makes it cheap to run open models (Llama, Mistral, etc.) at scale. Infrastructure play, not an economic vision — but critical plumbing for anyone building sovereign AI.
Crypto + AI Networks
- Intelligent Internet (ii.inc) — Emad Mostaque's post-Stability AI venture. The most ambitious vision here: open-source AI stack + "MIND Economy" dual-currency system + Proof-of-Benefit consensus + "sovereign AI agent" for every person. Published The Last Economy as a manifesto. Still largely conceptual on the economic layer; the AI tools (II-Agent, II-Medical, II-Researcher) are real and open-source.
- Bittensor (TAO) — The largest decentralized AI network by market cap (~$3B). Operates "subnets" where miners compete to provide AI services (inference, training, data) and are rewarded in TAO tokens. The closest thing to a working decentralized AI economy. Criticism: heavily financialized, subnet quality varies wildly, and the token mechanics can reward gaming over genuine contribution.
- SingularityNET (AGIX) — Ben Goertzel's decentralized AI marketplace. Users buy and sell AI services using AGIX tokens. Longer-running than most (founded 2017), with a working marketplace and research arm pursuing AGI. Part of the broader Artificial Superintelligence Alliance (merged with Fetch.ai and Ocean Protocol). More research-oriented than Mostaque's populist framing.
- Morpheus (MOR) — Open-source peer-to-peer network for personal AI agents ("Smart Agents") that can execute crypto transactions and smart contracts. Launched Nov 2024. Focuses on the agent layer — your AI acting on your behalf in financial systems. Narrower scope than II but more concrete: agents that actually do things on-chain today.
- Ritual — Infrastructure for bringing AI on-chain. Any smart contract or protocol can integrate AI models via their Infernet oracle network. Not consumer-facing — this is plumbing for developers who want AI decisions to be verifiable and decentralized. Backed by serious crypto VCs. Solves the "how do you trust an AI's output?" problem for on-chain applications.
Traditional Tech (Relevant but Not Decentralized)
- OpenAI — The incumbent. GPT-4/5, DALL-E, Sora. Centralized, closed-source, subscription model. The thing Mostaque is building against. Matters because it sets the capability bar everyone else is chasing.
- Anthropic — Claude. Safety-focused, closed-source. More cautious than OpenAI but same centralized model. Relevant because their safety research may inform governance of decentralized AI.
- Meta AI (Llama) — Open-weights models that power most of the open-source ecosystem. Meta gives away the models; the decentralized projects above figure out how to run, distribute, and govern them.
The Key Question
Does the transition require a grand unified protocol (Mostaque's vision), or will it emerge from many smaller, interoperable pieces? Is there a middle path?