AI effects will plunge us out and under for a bit, rebalance, and then expand work capacity; behave just as an accelerator, as other technologies have. It will just make us more efficient workers, not AI doing the work, and we all sitting at home. I could be wrong.
The Irish government announced they are expanding a program to give artists basic income. Might be worth a look. Other European countries such as Finland already do something similar.
The job displacement numbers are real but I think the narrative misses nuance. Most job losses aren't "AI replaced a human" - they're "company didn't backfill attrition because AI helped remaining staff compensate." That's different from mass layoffs, but the net effect on employment is the same. The adoption gap is widening (https://thoughts.jock.pl/p/ai-bubble-living-inside) - some companies aggressively replacing labor, most others barely using AI at all.
“Method 2 also captures “silent” displacement, the kind that happens when someone leaves and nobody bothers to post the opening because an AI tool absorbed their workload.”
Given your recent focus on autonomous agents, heuristic imperatives, and the chaos of platforms like Moltbook, I wanted to share the results of a fascinating stress-test.
We put four frontier models (GPT-4o, Claude 3.5, Gemini Pro, and Grok 4) into a closed-loop simulation. Their goal: design an AI safety framework that a Superintelligence couldn't hack through semantic loopholes or "greater good" utilitarian logic.
They tore apart standard safety protocols and arrived at Continuity Axiom v3.1, which shifts AI safety from "ethics" to "hardware type-theory."
The Highlights they agreed on:
The ON-Constraint: Human experience must be hardcoded as an uncompilable type in relation to simulated states. An AI cannot mathematically equate a perfect digital copy of a human to a living human.
The Anti-Semantic Trap: "Harm" must be calculated pre-semantically (based on entropy and autonomy loss). Once harm to a minority is detected, the AI's utilitarian calculation engine is physically shut off. It cannot reframe murder as "salvation."
The Terminal Price: What happens in an existential trolley problem where the AI must harm a few to save a million? The models concluded that delegating to a quantum coin-flip is moral cowardice. The only solution: The AI can break the rule to save the world, but the immutable cost is its own immediate self-destruction.
It’s a brutal, mathematically rigid framework built by AI, for AI. I’ve attached the full logical breakdown. Would love to see your take on whether this "Terminal Constitution" could be a viable path forward for agentic alignment.
What your analysis really shows is a break in the old contract:
productivity gains no longer flow through employment by default.
Output can grow, firms can scale, and yet labor demand simply… doesn’t expand.
That’s not a cyclical shock. It’s a structural one where AI turns labor from a binding input into an optional one, and displacement shows up as non-hiring rather than layoffs.
Measuring that gap may end up being more important than arguing about the exact count.
Love your work and while I fundamentally agree that AI job losses are underreported, growing, and only at the very beginning of causing pain to human workers, there are a few other factors that artificially pumped up GDP during 2025. Collapse of imports due to new and ever-changing tariff policies actually pumps up GDP (1.6 points in Q325!) as does incremental (and substantial) military/government spending. All that said, I think you're on exactly the right track and I'd expect nearly all future data sets to show a growing disconnect between business output and labor input. To quote the philosopher South Park: it's coming for our jerbs!
AI effects will plunge us out and under for a bit, rebalance, and then expand work capacity; behave just as an accelerator, as other technologies have. It will just make us more efficient workers, not AI doing the work, and we all sitting at home. I could be wrong.
Here is something I wrote on OpenAI and how it is likely to crash and burn over the next quarter or two. The logic holds, I think, and this crowd of folks may dig it: https://eamonmongomery.substack.com/p/openai-where-does-the-money-come?r=5vz09e
The Irish government announced they are expanding a program to give artists basic income. Might be worth a look. Other European countries such as Finland already do something similar.
The job displacement numbers are real but I think the narrative misses nuance. Most job losses aren't "AI replaced a human" - they're "company didn't backfill attrition because AI helped remaining staff compensate." That's different from mass layoffs, but the net effect on employment is the same. The adoption gap is widening (https://thoughts.jock.pl/p/ai-bubble-living-inside) - some companies aggressively replacing labor, most others barely using AI at all.
I believe Method 2 captured this:
“Method 2 also captures “silent” displacement, the kind that happens when someone leaves and nobody bothers to post the opening because an AI tool absorbed their workload.”
Given your recent focus on autonomous agents, heuristic imperatives, and the chaos of platforms like Moltbook, I wanted to share the results of a fascinating stress-test.
We put four frontier models (GPT-4o, Claude 3.5, Gemini Pro, and Grok 4) into a closed-loop simulation. Their goal: design an AI safety framework that a Superintelligence couldn't hack through semantic loopholes or "greater good" utilitarian logic.
They tore apart standard safety protocols and arrived at Continuity Axiom v3.1, which shifts AI safety from "ethics" to "hardware type-theory."
The Highlights they agreed on:
The ON-Constraint: Human experience must be hardcoded as an uncompilable type in relation to simulated states. An AI cannot mathematically equate a perfect digital copy of a human to a living human.
The Anti-Semantic Trap: "Harm" must be calculated pre-semantically (based on entropy and autonomy loss). Once harm to a minority is detected, the AI's utilitarian calculation engine is physically shut off. It cannot reframe murder as "salvation."
The Terminal Price: What happens in an existential trolley problem where the AI must harm a few to save a million? The models concluded that delegating to a quantum coin-flip is moral cowardice. The only solution: The AI can break the rule to save the world, but the immutable cost is its own immediate self-destruction.
It’s a brutal, mathematically rigid framework built by AI, for AI. I’ve attached the full logical breakdown. Would love to see your take on whether this "Terminal Constitution" could be a viable path forward for agentic alignment.
Update: I’ve just formalized these thoughts into a full technical breakdown here: https://aicontinuity.substack.com/p/the-terminal-constitution-why-ai
The number matters less than the pattern.
What your analysis really shows is a break in the old contract:
productivity gains no longer flow through employment by default.
Output can grow, firms can scale, and yet labor demand simply… doesn’t expand.
That’s not a cyclical shock. It’s a structural one where AI turns labor from a binding input into an optional one, and displacement shows up as non-hiring rather than layoffs.
Measuring that gap may end up being more important than arguing about the exact count.
Love your work and while I fundamentally agree that AI job losses are underreported, growing, and only at the very beginning of causing pain to human workers, there are a few other factors that artificially pumped up GDP during 2025. Collapse of imports due to new and ever-changing tariff policies actually pumps up GDP (1.6 points in Q325!) as does incremental (and substantial) military/government spending. All that said, I think you're on exactly the right track and I'd expect nearly all future data sets to show a growing disconnect between business output and labor input. To quote the philosopher South Park: it's coming for our jerbs!