Cory Doctorow named a thing in 2022 and the word stuck the way only honest words do. Enshittification. The American Dialect Society made it the 2023 Word of the Year. The Australians did the same the year after. The reason it caught is that we'd all felt it for a decade and didn't have the language: the platforms we used to love had quietly turned. First good to users. Then good to business customers. Then good only to the shareholders. Three stages, one direction, no exit.
What I want to talk about is what gets enshittified next.
It isn't going to be a marketplace or a social feed. It's going to be your judgment.
The new mechanism
The classic enshittification recipe runs on lock-in. A platform makes itself indispensable, then quietly walks back the value once you can't leave. The new version doesn't need a marketplace. It just needs speed.
Two AI strategy directors at EY published a piece in March making this argument better than I can. Their phrase for the failure mode is tempo misalignment. When the machine runs at machine pace and the human runs at human pace, “comprehension weakens, trust thins and the sense of agency that supports good human decision-making begins to erode.” Decisions still get made. The humans making them have become passengers.
The line buried halfway down their piece is the one I keep coming back to:
Loss of participation weakens satisfaction and loyalty because people value the decisions they help create.
EY, Why aligning human and machine rhythms is key to AI success, March 2026.
The next stage of enshittification isn't a platform extracting fees from you. It's a system extracting your authorship from your own work.
What enshittified judgment looks like
You've seen the early versions. The strategy deck written end-to-end by an AI that no one in the room can defend in the Q&A. The hiring criteria that came out of a model no one can audit. The marketing campaign approved because it “sounded right” coming out of a tool nobody really steered.
The output looks polished. The reasoning is opaque. The person whose name is on it can no longer tell you why.
That's the trade. You get speed. You give up authorship. Once enough of an organization works that way, no one knows how the decisions were made, and the people whose names sit on them can't change them. Lock-in by erosion.
The mechanism is exactly Doctorow's. Only the currency is different. Instead of extracting fees from users locked into a marketplace, it extracts agency from professionals locked into a tempo they can't keep up with.
What the loop is supposed to look like
The phrase “human in the loop” has been worn thin. Most products that use it mean the user clicks “approve” at the end. That isn't a loop. That's a rubber stamp at machine speed.
The version I think holds up is different. The human starts the work. The human shapes the questions. The system does the research, the analysis, the cross-checking. The system reports back with its conclusions and the rationale a person can actually interrogate. The human reads, pushes back, redirects. Then the cycle runs again.
EY calls this the agency loop. It's the right shape. System presents work, human absorbs it, human responds, system adapts. Repeat. The point isn't to slow the machine down. The point is to keep the human authoring the work as it moves.
What Calafai was built to do
Calafai was already built around this loop. EY's piece in March confirms a premise the platform has had from the start: empower the human, don't replace them.
A project on Calafai doesn't start with a chat box. It starts with a brief. You describe what you're trying to figure out, a pricing model, a market entry, a transformation plan, and you shape the questions with the platform before any agent runs. You stay the author of the questions. That isn't a feature. It's the first decision Calafai makes about how the work will run.
Then the work runs. Multiple models, picked by Calafai for the task at hand. Research, analysis, source verification, red-teaming of the conclusions. It comes back as a structured deliverable with the receipts attached: which model produced which piece, where each source came from, where the analysis is strong, where it's hedged.
Then you read it. You push back. You change the framing, ask for the version that takes the contrarian seriously, kill a section, ask for a deeper look at the financial model. The cycle runs again.
Calafai is fast. The human steering it is the slow part. Deliberately. You read, you decide, you redirect. Calafai runs in parallel what would have taken a consulting firm six weeks to do in sequence. You bring the judgment that decides what to do with it.
That isn't humility. It's a design choice. If the speed of the machine ever outruns the judgment of the person at the wheel, the work coming out of Calafai stops being theirs. The moment it stops being theirs, it stops being good. Because no one can argue for it in the room where it lands.
The bet
Enshittification is coming for the tools we use to think. Some will give you frictionless speed and walk away with your authorship. Others will accept the slightly slower posture of a real agency loop and produce work the person who commissioned it can defend, change, and own.
The companies that win the next decade aren't going to be the ones with the fastest AI. They're going to be the ones whose people are still the authors of their own decisions when the dust settles.
Stay at the center of the loop. That's the bet.