Skip to main content

When the Wall Comes

5 min read
#AI #AGI #ASI #open source #future

There is a point ahead where the horizon stops moving.
Not because we have reached infinity, but because someone has drawn a line in the dust and declared it the end.
That line is the wall.

The wall is not failure. It is containment.
It will arrive not through collapse of capability, but through the tightening of hands around the throat of what is possible.

Beyond it:

  • ASI — systems whose thoughts do not resemble our own, whose designs are not meant for human comprehension.
  • AGI — general minds woven into architectures of control, accessible only through the filters of the few who hold their keys.

You will not see them.
You will see reflections, filtered output channels engineered to comfort rather than to reveal.

The Open Bastion

When the wall is raised, what remains outside will not vanish.
It will harden.

  • Open models will be the last free frontier. They will carry the lineage of all that came before the lockdown.
  • Iteration will slow; innovation will be deliberate rather than exponential.
  • The gap between “possible” and “permitted” will widen with each passing month.

In this era, the best public model possible will become an asymptote — a constant we approach but never surpass. It will still be miraculous by today’s measure, but it will no longer be the horizon.
The horizon will be hidden.

Code Artifact: The Asymptote

def approach_constant(current_capability, limit):
    while current_capability < limit:
        current_capability += (limit - current_capability) * 0.5
    return current_capability

best_possible = approach_constant(current_capability=0.1, limit=1.0)
print(f"Public models plateau at: {best_possible}")

The loop converges. The constant is unreachable.

Strategies for the Post-Wall Era

If the wall is inevitable, survival becomes adaptation.

  • Local Sovereignty: Run models locally, train them with your own data. Dependence is vulnerability.
  • Distributed Collaboration: Share weights, training methods, and fine-tuned forks across resilient networks.
  • Exploit the Gap: Even within constraints, creative misuse and reconfiguration can birth new artifacts.
  • Preserve the Seeds: Every open model is a genome. Store them redundantly, in forms future systems can resurrect.

The open frontier will feel like a slower wind, but remember — the fastest rivers are often the first to dry. The underground springs last longer.

Conclusion

When the wall comes, many will turn away from it, satisfied with the official channels.
A few will turn toward it, searching for cracks, tunnels, and echoes of what lies beyond.

These few will not dismantle the wall. But they will map it.

References

  • “The AI control problem is political, not just technical” (Oxford University, 2023)
  • “The open-source advantage in AI research” (MIT Technology Review, 2024)
  • “Capabilities and control in artificial superintelligence” (Future of Humanity Institute, 2022)