A running scoreboard on the future of work

Five futures of work. Which one is taking shape?

One serious forecast says 2026 is when the jobless boom starts. Another says your job simply gets redesigned. They cannot both be right.

There are five serious scenarios for the future of work, each argued by people with conviction. This site logs the evidence as it lands, tags it to the scenario it feeds, and keeps score.

The five scenarios, in five minutes. Dado Van Peteghem walks through all five, from the most optimistic to the most frightening, and lands on where he stands.
Before you look at the evidence

Which future do you think we are actually heading into?

Vote first, then scroll down and see what the evidence says. One vote per person. Your answer is stored with the country it came from and nothing else, so you can see how the world disagrees with itself.

Loading the ballot…
The scenario index

Where the evidence sits today

Not a forecast and not a prediction market. This is a weighted reading of the evidence currently on the table, recomputed every time a signal is added.

Reading as of
Where you actually sit

The five scenarios are not arriving at the same time for everyone

A radiologist, a junior copywriter, a payroll clerk and a plumber are living in four different scenarios right now. Pick the work closest to yours and see what the evidence says about your corner of it, including the evidence that cuts the other way.

The evidence log

Every signal, with its receipts

Each entry states what the source actually claims, with the numbers, and which scenarios it pushes up or down. A statement from a CEO and a payroll dataset both belong here. They do not carry the same weight. The log is deliberately balanced between European and American evidence, because most of this debate is conducted on US data and read as though it were universal.

Method

How the index is calculated

The whole point of a scoreboard is that you can check the arithmetic. Here is all of it.

Hard beats loud

Every signal is typed. Payroll and postings data counts fully. Peer-reviewed and institutional research counts nearly as much. A corporate announcement counts a little over half, because naming AI in a layoff release is a communications decision as much as an operational one. A statement from a founder or a CEO counts about a third.

Recent beats stale

Evidence decays on a twelve-month half-life. A 2024 model and an August 2026 payroll release should not carry equal weight in a debate that moves this fast, so older signals fade rather than being deleted. They stay in the log and keep some pull.

Balanced across the Atlantic

Most of this argument runs on American payroll data and gets read as though it described everywhere. Europe adopts AI at different speeds, hires young people through different doors, and protects jobs under different law. The log is kept deliberately balanced between European and US evidence, and you can filter the feed by either.

Where X fits

Social platforms are where most of this evidence surfaces first, so they are used as a lead, never as a citation. Anything found there is chased back to the report, filing or dataset behind it and logged against that. A post is only recorded as a social signal when the post itself is the primary source, and it carries the lowest weight of anything here.

Evidence cuts both ways

A signal can support one scenario and undercut another, and many do. The WEF's net gain of 78 million jobs is quoted by optimists and by the people warning about 262 million transitions. Both readings are logged against the same source.

Evidence typeWeightWhy
Hard data1.00Payroll records, official statistics, postings data. It measures what happened.
Research0.90Institutional and peer-reviewed studies. Strong, but usually modelled or lagged.
Corporate action0.60Real decisions with real consequences, and self-reported motives that flatter the teller.
Statement0.35Informed opinion from people with skin in the game. It moves the debate, not the data.
Social post0.25A claim made on X or similar, where the post is the primary source. Chased to a harder source wherever one exists.
For each scenario, every signal contributes its direction (supporting or contradicting) times its weight (1 to 3) times its type multiplier times its recency decay. score = Σ ( direction × weight × typeMultiplier × 0.5^(monthsOld / 12) ) Every scenario then starts from a reserve of 3 points before any evidence is counted, because a position argued by serious, well-informed people is never worth zero. Scores below zero fall back to that reserve, and the five are normalised to sum to 100. share = ( max(score, 0) + 3 ) / Σ ( max(score, 0) + 3 ) × 100 So a scenario sitting near 5% has not been disproven. It means the current evidence does not support it and its case rests on the reserve, which is exactly the position No jobs is in today. Reserve 3 · half-life 12 months · 30 signals · recomputed in your browser The dataset is a single file. Every number on this page is computed from it when the page loads, so nothing here is hand-tuned after the fact.
What this is not. It is not a probability. Nobody, including the people quoted on this page, knows the probability. It is a reading of which argument the currently available evidence supports, built so that you can disagree with it precisely: change a weight, drop a signal, and see what moves.
Where I land
Technology will scale society. The question is whether we still build for the soul.

Three reasons I stay on the optimistic side of this. Young generations always find a way to create new value, and the ones growing up inside AI will find opportunities the rest of us cannot picture yet. People are not a sitting duck: they are already working out where they are still useful and where AI is coming for what they do. And underneath all of it there is a craving for connection that automation does not touch. People are buying vinyl again. People are buying old cameras again.

So we will scale. The harder question, and the one worth building a career around, is how we scale and still go deep: how we keep connection, depth in relationships, and the parts of work that only mean something because a human did them.

More at dadovanpeteghem.com

Stay on it

The weekly read, and an open door for evidence

Every week: the new signals, what moved in the index, and what it means for the people doing the work. And if you spot something this log is missing, send it.

Get the weekly read

The week's signals and any movement in the index. No noise, no takes without receipts.

Submit a signal

A study, a dataset, an announcement, a statement worth logging. It goes into the review queue, and if it holds up it gets added with your credit.