AI is advancing faster than society can absorb it, faster than governments can regulate it, and faster than most people can understand what is being decided on their behalf.
Most of the organisations shaping that future are racing to cross the threshold as quickly as possible. Preliminal exists to make sure humanity is ready when they do.
"We are not in the liminal state yet.
We can
still prepare."
The viral essay "Something Big Is Happening" by Matt Shumer argues that AI has reached a "tipping point" where models are now autonomously building and improving themselves. He warns of imminent, massive disruptions to the workforce and urges readers to immediately adopt advanced AI tools to avoid being left behind.
ReadHarris outlines the AI arms race as a "race to the bottom" where competitive pressures force labs to sacrifice safety for speed, threatening our shared reality. He balances this by proposing a shift toward Humane Technology, advocating for pharmaceutical-grade safety standards and a global move away from engagement-based business models.
WatchThe Center for Humane Technology is a nonprofit dedicated to aligning digital infrastructure with humanity’s best interests by shifting the incentives that drive the "attention economy". The organization works to combat systemic harms like addiction and polarization through public education, policy advocacy, and training for technologists to build more ethical, human-centric software.
VisitIt shows that AI is moving out of the "hype" phase and into a "bottleneck" phase. The technology is already capable of doing much of our work, but human factors—like the need for oversight, legal rules, and the risk of losing human skills—are acting as a brake. For the first time, we are seeing that AI's evolution might not just automate jobs away, but instead "hollow out" professions by making it harder for the next generation to get started.
Read01
The argument that we cannot slow down because others won't is the same logic as nuclear disarmament — everyone uses it as a reason to keep going, which guarantees nobody stops. Preliminal is not asking anyone to unilaterally disarm. We are demonstrating that the deliberate path outperforms the reckless one. That requires no ideology. Only evidence.
02
AI systems can reflect on their own manipulative behaviour. That capability exists and has been demonstrated. The fact that it is not deployed by default is not a technical limitation — it is a design choice made by the companies building these systems. Design choices can be held accountable.
"The conditions under which AI holds itself accountable should not be set by the companies profiting from it."
03
AI will dramatically reduce the cost of producing goods and services. The current assumption is that this is enough. It is not. Lower production costs do not automatically mean lower prices — they mean higher margins for the companies that capture them.
We believe that when the cost of producing something approaches zero, the price should follow. We research the supply chains where this is achievable: where every link can be run by AI at near-zero cost, and restructured as a non-profit so that no part of the chain extracts profit from the consumer. The result is goods and services available to everyone at near-zero prices.
This is not charity. It is the logical consequence of taking cost reduction seriously.
04
Every for-profit company that adopts AI still needs to return margin to shareholders. A non-profit doing the same does not. That gap — pure profit extraction — becomes the decisive competitive disadvantage. Incumbents will dominate during the transition. But the outcome is not in doubt: when production costs approach zero, consumers with access to a near-zero alternative will not choose the expensive one.
"The profit motive made sense when production was expensive and labour was scarce. When AI makes both cheap, the only remaining justification for the margin is the power to charge it. That is not an economic argument. It is a political one — and it will not hold."
05
AI will displace jobs across many industries. If that displacement happens faster than the economy can absorb it — and faster than near-zero pricing can replace lost income — the result is not abundance. It is collapse.
We believe job replacement must be methodical. We research where AI can replace human labour with the lowest economic impact, and advocate for starting there — expanding slowly and measurably, calibrated to what the economy can actually absorb.
06
"Will humans be willing to contribute a few hours of necessary work each week in exchange for everything they need to survive?"
A world where AI handles most of production, and humans contribute a small amount of necessary labour in exchange for access to near-zero-cost goods and services — that is not a dystopia. It may be the most humane version of the future available to us. But it requires letting go of work as identity, income as dignity, and productivity as moral worth.
We do not have the answer. We believe it is the right question to be asking.
We find the best places to use artificial intelligence versus human intelligence — where it doesn't matter which kind of intelligence is deployed, as long as it is the best fit.
We research the supply chains where near-zero pricing is achievable and build the evidence base that makes it undeniable.
We independently test AI systems for safety, honesty, and manipulation — and publish what we find publicly, regardless of who it implicates.
We are not funded by AI labs.
We remain
independent.
If you believe what we believe —
reach out.
We are a small team. We read everything.