#0003 It mostly works, The Hand-off to bots, Simulating the Post-Human Future, Helpful Llocal Models

Welcome to Constant Flux, a weekly lens on the polycrisis.

AI is taking over execution, governance is chasing its own tail, and computing could be migrating away from the corporate cloud into smaller, local intelligence.

AI is shifting from tool to something more independent, and things could be getting strange.

If humans are no longer the final decision makers, what does that say about control? That is what stuck with me from everything I read this week.

It mostly works

Vibe coding is having its moment. Instead of writing code line by line, people can now act as orchestrators, guiding AI tools to generate code.

I think this trend extends beyond coding. We can talk about AI tools enabling vibe research, where individuals utilize AI to conduct research and enabling "research as leisure activity"

What other disciplines could one "vibe"? Any right? Writing is an obvious contender. Authors already use tools like Sudowrite to generate drafts and shape narratives.

So, what happens when AI takes over execution and users refine outputs instead of mastering principles. It sure won't produce the next Pynchon (but perhaps the next 100 Dan Browns?).

The Hand-Off To Bots

In the article Kelly explores how the decline of human populations intersects with the rise of AI and robotics. This could represent a fundamental shift in how society functions.

• Economic shift: As AI takes on more jobs, it won’t just replace labor, it will reshape markets, influencing both production and consumption.

In the short term I believe this transition will widen economic divides as slower adopters of automation struggle. Job displacement without alternatives will drive unemployment, deepening inequality and fuelling unrest...the usual.

• Social impact: Offloading routine tasks to bots could free people for creative and interpersonal pursuits (MOAR vibe coding!), but I also think it raises questions about purpose and fulfilment in a world where machines handle the essentials. Right?
• Environmental factor: It goes without saying but more automation means more resource consumption. AI needs infrastructure, energy, and materials. What’s the long-term cost of scaling automation?

This may not be just a technological upgrade but a broader shift. Automation, the economy, and society are interconnected, and its impact will depend on how it is managed. I guess last weeks launch of Manus deserves a link here. Kelly talks about the same topic in this interview.

Simulating the Post-Human Future

Continuing the theme of bots. David Mattin discusses Project Sid, where 1,000 autonomous AI agents in Minecraft developed their own economy, culture, and religion. This showcases AI potential to create complex societies, hinting at a future where AI-driven economies operate independently of human intervention.

Autonomous AI agents aren’t just tools; they’re evolving systems that interact, adapt, and create feedback loops we can’t always predict or control. This isn’t just automation—it’s a fundamental shift in how systems operate.

It's obvious, our governance models weren’t built for this. They assume human-driven, hierarchical structures, not networks of AI agents making autonomous decisions, forming economies, and self-organizing in real time. In order to keep check on things we probably need to start thinking about building our own governance AI systems.

Helpful Llocal Models

A five-minute podcast discussing how AI shifting from the cloud to local devices could change how intelligence is distributed.

Once dependent on massive data centers, models are now running on desktops, phones, and modular clusters. This could reshape computing infrastructure.

Framework’s new modular desktop lets users chain units together to scale up compute power at a fraction of traditional costs. If this approach gains traction, it could signal a broader shift in how AI is deployed.

I think what’s clear is that intelligence is no longer tied only to remote servers. The next step may not just be running AI models locally, but making them scalable, modular, and seamlessly integrated.

Tokyo is turning to a 4-day workweek in a desperate attempt to help Japan shed its unwanted title of ‘world’s oldest population’

Maybe Tokyo’s four-day workweek could signal a deeper shift, not just changing work schedules, but transforming broader social patterns. Japan faces a demographic crisis: low birth rates and an aging population tied directly to its intense work culture. Shortening the workweek might improve work-life balance enough to encourage more families and solve these issues. I mean that's the hypothesis.

Uncertainty prevails of course. Cultural resistance is significant, and few companies have embraced the change despite official backing. 

Exclusive: US intel shows Russia and China are attempting to recruit disgruntled federal employees, sources say

File under "Who saw that coming?"

The recent mass layoffs of U.S. federal employees under the Trump administration have led to several problems affecting national security, economic stability, and ultimately geopolitical balance. Critical agencies like the Cybersecurity and Infrastructure Security Agency, which protects against cyberattacks and ensures the security of key infrastructure, (CISA) lost hundreds of staffers, effectively weakening America's ability to defend against cyber threats and safeguard vital infrastructure.

No surprise then that China and Russia are actively targeting and recruiting former U.S. government employees.

By rigorously applying his trademark first principles thinking, Elon Musk has apparently decided that the best way to counter geopolitical threats is to have Russia and China recruit disgruntled federal employees.