The gap between Domination (AI-Dystopia) and Spread (Futurama) keeps narrowing – but the real shock this week is an Autonomy (Singularia) reading that nearly doubles the previous high. The question of what AI actually is has arrived.
New to Fizzle, Spread, Domination, Autonomy and Chaos? Here’s why I renamed the five worlds – the Aaronson & Barak originals appear in brackets.
This week, more stories than ever before were about whether AI is about to become as smart as – or smarter than – humans. That is the kind of future some people call the Chaos (Paperclipalypse) scenario: AI that outgrows our control. At the same time, the cheerful Spread story – where AI is just a useful tool that people manage sensibly – is still the biggest single storyline. The worrying part is that the two happier stories, 'AI is overhyped and fizzling out' and 'humans are in charge', are both getting quieter. So right now, the loudest voices are split between 'AI is going to change everything' and 'AI is already hurting people' – and the calm middle ground is shrinking.
The Domination–Spread gap this week stands at –16 percentage points – Spread leads, meaning optimism still outweighs alarm when you look at raw source counts. But the gap has narrowed from its widest point. In Scan 8 it sat at –20, the most optimistic reading in the entire series. This week it pulls back to –16. That is not a reversal, but it is a signal worth watching.
To read that gap properly, you need the full series in front of you. Back in Scan 2 – the second week of tracking – Domination led Spread by 24 points. That was the peak of alarm. Since then the series has made a long, uneven journey toward optimism: the gap crossed zero in Scan 5 (–3, Spread first takes the lead), widened sharply through Scans 7 and 8, and now pulls fractionally back. The direction of travel over the whole nine-scan arc is unambiguously toward the Spread world. But the journey has never been smooth, and this week introduces the most significant complication yet.
The headline number this week is not the gap – it is Autonomy at 29%, the highest single reading in the entire series. The previous high was 16% (Scans 2 and 7). This week it nearly doubles that. Six sources – ranging from the AGI-26 Conference to an arXiv paper on economic policy – are all built on the same foundational assumption: that artificial general intelligence is a real, approaching milestone that changes the rules of the game.
This is structurally significant. Autonomy has historically been a background frequency, occasionally spiking when a Musk quote lands or a safety report drops. This week it is not a spike driven by one celebrity prediction. The sources are spread: academic conferences, investment analysis, formal safety governance, economic modelling. The discourse around AGI has broadened and deepened. When TimeTrex calls 2026 "a profound and irreversible inflection point" and 29 nations co-sign an international safety report premised on catastrophic risk from general-purpose AI, these are not fringe voices.
Critically, the International AI Safety Report sources sit in Autonomy precisely because their entire logic depends on AI capability accelerating toward something dangerous. Safety work is Autonomy's shadow. You do not build defence-in-depth governance frameworks for a tool that is fizzling out.
While Autonomy surges, Fizzle drops to 7% – the lowest reading in the series. Three sources, two of them pointing at the same finding: that most AI deployments fail to deliver promised ROI, and developer productivity tools return 1.6–3.5x rather than the vendor-claimed 10x. That story has not gone away. But it is being drowned out. In the early scans Fizzle was regularly at 13–19%. The deflation narrative held real weight. Now it is a footnote.
This matters because Fizzle was doing important corrective work in the discourse. When IBM reports that only 25% of AI initiatives deliver expected ROI and WRITER finds 79% of enterprises facing adoption challenges, those are not minor findings. The fact that they are producing fewer headline-level sources this week does not mean the underlying reality has changed – it may simply mean the media has moved on. The absence of Fizzle-world thinking at scale is a structural gap in the current conversation.
At 24%, Domination is at its series low – but that low is still almost a quarter of all sources. And this week's dystopian sources are unusually concentrated and coherent. Seven of the ten come from a single theme: AI-attributed tech layoffs. The numbers are specific and staggering – 80,000 jobs in Q1 2026, 1,115 layoffs per day, $700 billion in AI infrastructure capex funded, at least partly, by firing people. TechTimes names Meta and Oracle directly. SkillSyncer's tracker puts a number on it: 185,894 workers in 2026 so far, majority AI-attributed, by companies simultaneously posting record profits.
The MIT Media Lab study and the ScienceDaily piece on AI swarms hijacking democratic discourse add a second dystopian thread: cognitive and political manipulation. These two clusters – labour displacement and information control – have been the twin engines of Domination throughout the series. They show no sign of losing fuel.
Step back and read the whole chart as a leader would. Nine weeks ago, Domination dominated and the dominant media frame was alarm. Over the following months, something shifted: enterprise ROI stories multiplied, regulatory frameworks matured, productivity data came in, and Spread became the single largest world in the series. That is a real change in the discourse, not noise. But this week introduces a new variable. The conversation is no longer just 'is AI useful or dangerous?' It is increasingly 'is AI about to become something categorically different?' Autonomy's surge tells us the AGI question has moved from specialist conferences into mainstream business and policy media. That changes the terrain. A tool-world narrative and a threshold-crossing narrative cannot both be true at the same time – and the sources this week have not resolved that tension. They have amplified it.
What is almost entirely missing from this week's scan is any serious engagement with distribution – not just whether AI creates value, but who captures it. The layoff data is there. The productivity data is there. But the sources that would connect those two things – asking why record AI capex and record layoffs are happening simultaneously at the same profitable companies – are almost entirely absent. The Domination sources describe the pattern. The Spread sources describe the gains. Nobody this week is writing the story that sits between them.
The articles most responsible for bending the five worlds this week.
This study found that AI-powered fake personas can quietly slip into online communities and nudge how people think – without anyone realising it is happening. It is one of the most direct challenges this week to the idea that AI is just a neutral tool, because it shows AI being used specifically to take away people's ability to make up their own minds.
This piece put a dollar figure on the trade-off: the same companies making record profits are spending $700 billion on AI while letting tens of thousands of workers go. That connection – between who pays and who gains – is the clearest picture this week of why the Domination world signal is not going away.
Twenty-nine countries and organisations including the UN and OECD jointly signed off on this report, which is built on one core belief: that AI is getting powerful enough to cause catastrophic harm if we are not careful. A report this official, backed by this many governments, is the main reason Autonomy hit its series high this week.
Elon Musk predicting AI will exceed total human intelligence by 2030 is a big claim – but what matters here is that business and investment media are now running it as a serious story rather than a provocation. When mainstream financial outlets treat AGI timelines as market-relevant information, the Autonomy world stops being a niche concern.
IBM's finding that three in four AI projects fail to deliver what was promised is one of the most grounding facts in this week's scan – and yet it barely moved the needle, because Fizzle is at its lowest point in the series. The story is true, the data is solid, but nobody is listening as loudly as they were a few months ago. That gap between evidence and attention is the most important thing this number tells us.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free