The AI Twilight.
Why this era resists every old playbook — and how to move through it.
For most of our working lives, you could look five or ten years ahead and see a path to follow. The past no longer illuminates the future.
We're operating by dim, diffused light. That's the Twilight.
Name the moment right, and why it's so hard comes into focus — three forces that, together, take away the instruments you'd normally steer by.
Three forces strip your instrumentation at once:
- Precedent Collapse breaks the map — the past stops predicting.
- Feedback Delay severs the dashboard — the present can't confirm.
- Exponential Fog hides the headlights — the future won't resolve.
Map, dashboard, and headlights, gone together. That's why it feels like driving by dim light.
The three forces
Precedent Collapse
"The past stops reliably predicting the future; the pattern-matching you built over a career quietly stops working."
Most expertise is compressed history — an internal library of "I've seen this before" that lets you read a situation fast without re-deriving it. Call it intuition or instinct or heuristic, but our brains are built to efficiently take these shortcuts. That library only holds value while the world keeps rhyming with the past it was built on. In the AI transition, the terrain shifts faster than the library can refresh — so the patterns don't fail loudly. They fail quietly, still firing with full confidence at a world that has already moved.
The manager who can "read a candidate in five minutes" — a read earned over hundreds of interviews — is now hiring for a role AI is rewriting faster than the success pattern can re-form. The instinct still fires with conviction. It's just aimed at a job that no longer exists in the shape it learned.
So whatHold experience as a hypothesis, not an answer. Name the pattern you're about to run, ask what world it was calibrated to — and check whether that world still exists before you trust the read.
Feedback Delay
"You can't tell whether a decision worked before the ground shifts under it again."
Every good decision-maker runs a loop: act, watch what happens, adjust. It only works when the world holds still long enough to grade your answer. In the AI transition it doesn't. The interval between making a call and learning whether it was right has grown longer than the interval between one ground-shift and the next — so the verdict on your decision arrives too late to act on — and usually too tangled with everything else that moved to read clean at all. You can't grade your own call.
A leader reshapes a team around a new AI workflow and waits to see whether it lands. Six months on, the numbers are up — but the market also turned, two competitors stumbled, and the tooling changed twice. Did the reorg work, or did the weather? The signal that would settle it never arrives clean. It's flying without a working altimeter: the gauge that should tell you how your last move actually went simply isn't reading.
The data names this force from the outside. MIT's 2025 State of AI in Business found 95% of enterprise AI pilots produced no measurable P&L — and the cause it named wasn't the models. It was the learning gap: systems and organizations that don't retain feedback, adapt, or improve. The ~5% that broke through were the ones that closed the loop.
So whatDon't wait for the verdict — it won't come clean. Shrink the loop. Make calls you can read in weeks, not years; instrument for the earliest honest signal; treat every decision as a live experiment you're already revising. In the Twilight, speed of learning beats quality of the original guess.
This mindset is, in fact, how AI agents inherently work — they run in fast, reversible, self-correcting loops. Learning to leverage them well is one of the most direct ways to start putting this into practice.
Exponential Fog
"You can see a short distance clearly … and almost nothing about the shape of what's coming."
Human judgment was built for a world that moves in straight lines. We forecast the same way we walk: assuming the next step resembles the last. But AI capability isn't compounding in a straight line; it's a curve, and a curve looks deceptively flat right in front of you and then bends out of view. The near field stays sharp while the two-to-five-year horizon — the distance a career or a strategy actually lives in — dissolves.
A company locks a three-year AI roadmap — vendors chosen, budget committed, teams hired against it. Not because anyone is sure it's right, but because that's how planning has always worked: pick the horizon and march. The roadmap isn't wrong, exactly. It's drawn for a horizon past where the headlights reach — and the curve that looked flat at signing has bent out of view by year two, with the whole org already committed to the straight line.
So whatStop planning to a horizon you can't see. Stay liquid. Commit to directions, not destinations; buy optionality; keep the big irreversible bets small and late. The aim isn't a sharper five-year plan — it's a posture that can re-shape the moment the curve bends into view.
How to move through it
Notice the through-line: navigating the Twilight isn't about predicting harder. The three forces guarantee you can't. It's about how you carry yourself through low visibility. Hold judgment loosely. Move in steps you can reverse. Build the capacity to re-orient faster than the ground changes.
Less forecasting. More turning radius.
That's not resignation — it's the posture the conditions actually reward. The leaders and careers that thrive in the Twilight aren't the ones who guessed the future correctly. They're the ones who stayed adaptable, kept their own judgment, and could change shape without losing themselves.
Navigate the Twilight with company.
Every week, Signals from the Curve tracks the parts of AI work that compound — and flags the small things that are about to get large. This framework is the front door.
Wisdom that outlasts the algorithm, every Wednesday.
Where this goes
The Twilight is the weather — the conditions everyone now works in. The discipline of intentionally designing how people and AI think together inside those conditions is the larger body of work this belongs to: Organizational Intelligence Design.
It's also the opening note of a framework arc landing week over week in Signals from the Curve: Twilight, then the disciplines for working well within it. This page is the front door.
"We're operating by dim and diffused light — neither the certainty of daylight nor the honesty of total dark."
Track the arc as it lands — subscribe to Signals from the Curve.
Reitz, C. H. (2026). The AI Twilight: The Three Forces That Make This Era Hard to Navigate. chrishuberreitz.com/frameworks/ai-twilight