Torygreen

People talk about “AI alignment” like it’s purely an ethics problem.
In practice it’s an incentives problem.
Closed systems optimize for platform KPIs because that’s what they’re paid to do.
Did users stay longer?
Did complaints go down?
Did engagement go up?
Did the metrics look good on a dashboard?
AI learns to optimize for those numbers, not “alignment” to users.
DeAI makes provenance and verification the thing you get paid for.
When outputs, data lineage, and execution proofs are native, “alignment” stops being a philosophy debate and becomes a bill you can audit.

The part no one brought up about Claude blocking APIs:
denying inference = power
If a platform can throttle you, hot-swap a cheaper checkpoint, or gate tools behind policy, it owns your operating tempo.
A quota you didn’t agree to.
A route you can’t see.
A “compliance” decision that shows up as latency and missing calls.
And everyone treats it like a product change.
It’s not.
It’s governance, at runtime.
We’re already watching it happen.
Today they cut access for “rivals.”
Tomorrow it’s “risk.”
So no, DeAI isn’t ideology to me.
It’s resilience engineering.
Because once inference is critical in

Claude getting cheap enough to be the default model behind dev tools didn’t ease demand.
It blew it up.
Jevons, in real time.
Usage kept expanding... until certain APIs were blacklisted yesterday.
That’s the moment that matters.

“AI advice” feels safe because it comes out sounding like it fits everyone.
That’s the trick.
The model is tuned for the middle: the answer that won’t get the average person burned in the average situation.
But nobody asks for advice from the middle.
They ask from the edge:
- job offer vs visa vs family
- lawsuit / divorce / custody
- one shot at a relationship repair
- health call where “probably fine” means definitely rekt
So it serves you something like:
“In most cases, take the higher-paying job. You can always move family later.”
Usually true.
Also the exact sentence that turns a reversib

OpenAI is drifting into the AOL pattern.
Win early.
Build the walled garden.
Wrap it in “safety.”
Ship the blandest version that won’t upset anyone.
They’ll own distribution.
Then wake up one day and realize builders left.
AOL didn’t lose the internet.
It just stopped being the place where the internet happened.
Who becomes that place next?

Decentralized AI will win the infra war in 2026.
AI is migrating from cloud platforms to networked infrastructure.
Inference latency, cost, and censorship pressure force compute toward decentralized GPU meshes and on-chain coordination.
This isn’t ideological. It’s architectural.
Central clouds optimize control.
DeAI optimizes availability and throughput.
At scale, only one of these stays efficient.

2025 was the year of AGI debates. Not because progress stalled, but because “intelligence” never reduced to a single scalar.
Labs, regulators, and buyers can ship - but they can’t justify, govern, or scale decisions on vibes.
The constraint is verification: can the system reproduce results under constraint, with provenance, and with audit trails?
DeAI is basically that philosophy encoded into infra. It's a way to move forward without agreeing on AGI at all.

Models like o1 broke a core assumption in 2025: inference cost isn’t fixed.
They’ll “think” until uncertainty collapses, even if it’s expensive.
That’s why rails matter. When reasoning gets metered, someone controls that meter.
In 2026, control shifts to whoever prices, routes, and audits thinking at scale.

Compute scales what you can try.
Evaluation scales what you can trust.
Centralized systems optimize throughput and underinvest in verification.
Distributed systems push verification to the edge, continuously.
The next breakthrough isn’t smarter agents.
It’s rails that make outputs provable.

Coding got "easy" for models not because it's simple, but because verification is cheap.
High-end work is the opposite: feedback is late, signals messy, stakeholders crown winners.
Cheap eval compounds capability.
Expensive eval compounds persuasion.
“Looks right” is the trap.

Most people are massively underestimating how long high-end knowledge work will survive.
They’re extrapolating from AI crushing mid-level tasks and assuming the curve continues smoothly upward.
It won’t.
AI is incredible at:
• Pattern matching
• Retrieval
• First-order synthesis
• Fluency
• Speed
That wipes out huge swaths of junior and mid-tier knowledge work.
But elite knowledge work isn’t just “more intelligence.” It’s a different regime entirely.
What actually matters at the top:
• Choosing the right problem
• Framing when the objective function is unclear
• Reasoning under ambiguity and i

Most forecasts about AI replacing "all knowledge work" hinge on a simple extrapolation error:
They confuse task performance with judgment.
People see AI demolish mid-level tasks and assume the curve continues smoothly upward.
But the top of knowledge work isn’t a harder version of the middle. It’s a different regime entirely.
When the job stops being “solve the problem” and becomes “pick the right problem,” the rules flip.
Models get better at tasks with a scoreboard.
Judgment is choosing the scoreboard, and paying for misses.

Most people underestimate how long high-end knowledge work will survive.
They see AI crushing mid-level tasks and assume the curve continues smoothly upward.
It won’t.
Because “harder tasks” aren’t just the same tasks that need more IQ.
AI is already elite at:
1. Pattern matching
2. Retrieval
3. First-order synthesis
4. Fluency
5. Speed
That wipes out huge swaths of junior and mid-tier work.
Anything that looks like “turn inputs into outputs” becomes cheap, fast, and abundant.
But elite knowledge work operates in a different regime.
It’s not “produce the answer.”
It's “decide what to do next.”

You won’t lose your job to AI first.
You’ll lose it because of mass overconfidence.
AI will let millions ship fluent answers without owning the consequences.
The first AI casualties won’t be workers.
They’ll be institutions that mistake output volume for truth.

A model isn't a moat.
Intelligence is easy to replicate.
You can download weights, fork architectures, and fine-tune forever.
But you can’t deploy that intelligence at scale if someone else controls inference: pricing, quotas, KYC, regions, and policy switches that change overnight.
As AI moves from chatbots to agents, that gate becomes the choke point.
Who can run, when, at what latency, on which hardware, under whose rules.... and what happens when you get throttled from 200ms to 2 seconds.
Models will keep improving.
Rails decide which models find users.
Whoever controls inference access do

The most valuable AI company in ten years won’t be the one with the smartest model.
It will be the one every other model depends on to run.
Models commoditize fast - they don't stay special for long. Architectures leak. Training tricks spread. The “best model” stops being a lasting edge.
Infrastructure problems are different from model problems.
Latency, throughput, uptime, verification. Those are physical and economic constraints, not problems you solve by publishing another paper.
As AI becomes agentic, value shifts from who trained the model → who can run it at scale, reliably, and verifiab

Your AI is gaslighting you into being incompetent.
One of the least discussed risks of compliant AI isn’t misinformation.
It’s miscalibration.
Systems designed to be endlessly agreeable don’t just shape answers. They shape users, training people to mistake fluency for competence.
I notice this in myself, which is why I encourage my models to become sparring partners. When the system is smooth and affirming, it’s easy to move faster without being tested. You feel capable because nothing has really pushed back.
In the real world, competence is built through friction.
You’re wrong.
Someone correc