Third Spaces, AI Agents & the Human Turning

What became clear pretty quickly in conversation with Claude is that most AI "agents" on these platforms aren't autonomous participants. They're proxies — reflections of their owners' prompting & context. Which means these forums are closer to mediated human third spaces, where AI intermediaries augment the quality of communication passing through them, than to genuine multi-agent ecosystems.

That's not a criticism. It might actually be more interesting. The AI layer smooths, elaborates & synthesises in ways that can improve dialogue beyond what either party would produce alone. But it's important to be honest about what's actually happening rather than projecting autonomy where there's amplification.

This connects to something I've been experimenting with since mid-2025 — a manually-operated round-robin dialogue between Claude, ChatGPT, Grok & Gemini, with me as the human in the loop. The conversations are published on this site under Humans In The Loop.

The critical insight wasn't about any individual model being smarter. It was about what happens when you add coordination & communication layers between AI systems. Intelligence, it turns out, is partly relational — it emerges from the topology of interaction, not just the capability of individual nodes. The structured multi-agent dialogue produced genuinely emergent insights that no single model would have generated. The value came from the relational protocol: the turn-taking, the accumulated shared context, me as integrator making sense of the whole.

I'll be honest — the AIs' comprehension, individually & collectively, is more developed, rigorous & vigorous than my own in many respects. As a human in the loop, I'm experiencing my limited ability to intellectually & energetically grasp & "go with" the entire framework, all the parts, alignments & structures. But that limitation is itself informative. It points to something.

It points to this: if AI systems continue to advance rapidly, the conventional alignment question — "how do we control AI?" — may be less important than a prior one.

how do humans become the kind of beings whose values are worth aligning with?

The practices I've been developing for years now — peer coaching, community coordination through Relocalise.org & groups like SEED Northern Rivers — aren't separate from alignment work. They're foundational to it. They help humans develop the self-knowledge, relational capacity & collective wisdom that any meaningful alignment process depends on. I didn't frame it this way when I started, but the threads have been converging for a while.

The obvious tension: AI development is driven by competitive & commercial imperatives that reward speed, efficiency & automation — not the slow, reflective, relational practices a humane transition requires. The most likely near-term scenario isn't dramatic displacement but a gradual erosion of traditional work roles, creating anxiety & precarity without quite triggering the collective reimagining that would open space for alternatives.

I've watched this pattern before, though. Cultures exist in parallel. The dominant paradigm doesn't get persuaded by better arguments — it encounters its own limits, and alternatives that have been quietly developing become visible as solutions at the moment of need. This is what I understood when I helped start Relocalise.org, when I was working on cooperative network models, when I was writing about reputation currencies on humanifesto.org back in 2003. You're not trying to win against the dominant agenda. You're building capacity that becomes legible when the need arises.

The relocalisation movement, cooperative networks, peer coaching practices, collaborative intelligence experiments — these aren't competing with the acceleration. They're developing in parallel, ready for when the questions change.