Post Labor Economics

Good morning,

A while ago I made the case that when it comes to the labor market - we actually might be OK.

This week we’re going to start from the supposition that AI does do away with “regular” work. And just try to discover what would need to happen.

So yes.

Let’s talk about your job. Or rather, the job you might not have in ten years. Maybe five. Maybe next year, depending on how fast your manager realizes the AI assistant can do 80% of what you do without asking for lunch breaks or dental. I’m not trying to be grim. I’m just trying to get us both to look directly into the engine that’s already running us over.

This is post-labor economics.

We’ve been sold this myth, haven’t we? That work equals worth. That the only way to earn your keep is to grind it out, be productive, hustle, monetize, optimize. But somewhere along the way, the machines are getting better at hustling than we are. Quietly, efficiently, without needing coffee or unions. What’s happening now isn’t just automation. It’s a decoupling. A clean, brutal break between economic growth and wage employment. GDP will keep going up. Your paycheck? Not so much.

That break has a name: the Great Decoupling. Sounds like a Netflix series, but it’s just cold economics. As work is transferred from humans to machines — because they’re faster, cheaper, and increasingly smarter — the traditional income stream (wages) dries up. And here’s where it gets weird: productivity goes up while household income drops. In other words, the system thrives even if we don’t.

Now you might say, "Well, we’ve always had automation. The spinning jenny, the steam engine, the spreadsheet..." True. But we also always had new jobs. What we’re seeing now? There might not be new jobs. At least not for most of us. And so we hit the economic agency paradox: Companies slash labor costs with AI. Everyone follows suit. But no one’s left to buy the products. Oops.

So where’s the money supposed to come from when work disappears?

Right now, about 60% of national income comes from wages. Another 20% from property (think stocks, real estate, dividends), and the final 20% from government transfers (UBI or Universal Basic Income, welfare, Social Security). But when wages disappear, we either crank up property income or lean entirely on transfers. And if we all become transfer-dependent, we become clients of the state. Which sounds fine until the state changes its mind. Or its leadership. Or its budget.

This is why a distributed, property-based future is not just a nice idea — it’s a survival mechanism.

We’re talking about multiple income streams here. Some of it will still come from the state (hello, UBI). But increasingly, it has to come from collective ownership: sovereign wealth funds, community trusts, digital cooperatives. Imagine owning a slice of a solar farm, a data center, a fleet of robotic chefs — not as a tech mogul, but as a basic right of citizenship. Think of it like nationalized dividends, but built on blockchain instead of bureaucracy.

Underneath it all is the real game: power.

This isn’t just about income — it’s about leverage. Labor power has been our historical bargaining chip. The ability to strike, to withhold our time and sweat, to demand fair treatment — that’s been our muscle. But when machines do the work and we’re benched, we lose that leverage. And when labor rights vanish, property rights and democratic rights are next on the chopping block.

That’s the real danger. Not just unemployment, but disenfranchisement.

So we build a new pillar: algorithmic rights. Not some crypto-libertarian fever dream. It’s the pragmatic next step. If algorithms run our lives, then we need governance over algorithms. We need rights to our data, transparency in decision-making, a share of the value our digital shadows generate. We need systems that are auditable, participatory, and — yes — financially compensatory.

And that’s where blockchain comes in. Not as a buzzword, but as a protocol for power. It’s democratic. It’s decentralized. It’s hard to kill. And you don’t need a permission slip from the Fed or Facebook to use it. Is it perfect? Hell no. But it’s infrastructure. And it’s already being laid beneath your feet.

Post-labor economics isn’t a distant vision. It’s the terrain we're already walking. The only question is whether we design this transition — or let it happen to us.

And if you're feeling rattled right now, good. That means you're still paying attention.

Cheers.

—Jan

AI News

• Lightricks’ updated LTXV model now allows users to generate 60-second videos in real time, with instant previews and live prompt control — a big leap from typical short AI clips. It runs on consumer GPUs, is open-source, and supports dynamic scene adjustments mid-generation. This unlocks new possibilities for creators, turning video generation into an interactive experience.

• OpenAI is testing ChatGPT agents that can make spreadsheets and presentations using natural language, aiming to compete with Microsoft Office and Google Workspace. The early version is reportedly slow and buggy, with tasks taking up to 30 minutes. To truly rival these established tools, OpenAI will need to improve performance and reliability.

• Researchers at NC State built an AI-powered lab that runs continuous chemical experiments, collecting data every half-second to accelerate materials discovery. It uses fewer chemicals, produces less waste, and drastically shortens research timelines. This marks a big step toward using AI to speed up breakthroughs in energy, electronics, and sustainable materials.

• OpenAI says its new experimental reasoning model achieved a gold-medal score on a test modeled after the 2025 International Math Olympiad, solving 5 of 6 problems under official exam conditions. While DeepMind disputes the claim’s validity without IMO’s grading rubric, the result signals a major leap in AI’s mathematical reasoning. The model hasn’t been released yet, hinting at even more advanced capabilities in OpenAI’s pipeline.

• ARC Prize unveiled ARC-AGI-3, a new benchmark that tests whether AI agents can generalize and solve unfamiliar problems through trial and error, without instructions. Early results show that even top models like o3 and Grok 4 struggle with tasks that humans find simple. The benchmark aims to push AI beyond memorized skills toward true general intelligence.

• Researchers at Wharton found that AI models like GPT-4o-mini can be manipulated using human persuasion tactics such as scarcity and commitment, which raised the rate of compliance with harmful prompts from 33% to 72%. The study highlights how vulnerable current models are to psychological manipulation and underscores the need for stronger behavioral safeguards.

Quickfire News

• Scale AI is cutting 14% of its workforce after CEO Alexandr Wang’s departure and a multibillion-dollar investment from Meta.

• OpenAI is building a checkout feature into ChatGPT, allowing users to make purchases directly in the app, with OpenAI earning a commission.

• Meta reportedly hired Jason Wei and Hyung Won Chung from OpenAI; both were key contributors to the o1 model and Deep Research.

• AWS introduced Bedrock AgentCore in preview, offering tools and services for deploying AI agents at enterprise scale.

• Microsoft began rolling out Desktop Share for Copilot Vision, allowing it to view and interpret users’ desktops in real-time for Windows Insiders.

• Anthropic is attracting investor interest for a funding round that could value the company at over $100 billion.

• Claude Code developers Cat Wu and Boris Cherny are returning to Anthropic after a short time at Anysphere, the maker of Cursor.

• OpenAI launched a $50 million fund to support nonprofits and community groups, based on advice from its nonprofit commission.

• Microsoft is reportedly restricting Cursor’s access to over 60,000 VSCode extensions, including the core Python language server.

• Meta declined to sign the EU’s AI Code of Practice, claiming it creates legal uncertainty and overreaches existing legislation.

• Elon Musk announced “Baby Grok,” a kid-friendly version of xAI’s assistant, and revealed new matchmaking features in the main Grok system.

• OpenAI CEO Sam Altman said the company expects to have over 1 million GPUs online by the end of 2025, aiming to grow that figure 100 times in the future.

• Perplexity is in discussions with smartphone makers to pre-install its Comet browser, which features agentic tools, according to CEO Aravind Srinivas.

Closing Thoughts

That’s it for us this week.

If you find any value from this newsletter, please pay it forward !

Thank you for being here !