The gap between hopeful and harmful AI narratives just hit its widest point ever. Yet beneath the optimism, 142,000 tech workers have lost their jobs – and the money went straight to AI infrastructure.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week, the stories people are telling about AI became more hopeful than at any point since we started tracking. That sounds like good news – and some of it is. Big investments in health AI, falling prices for AI tools, and governments working together on safety rules all point toward the Futurama future, where AI stays useful and people stay in charge. But there's a catch hiding in the data. Hundreds of thousands of real people have lost their jobs this year, and the companies doing the firing are using that money to build more AI. That's much closer to the AI-Dystopia future – where the benefits pile up for a few at the top while everyone else pays the price. The hopeful headlines and the harmful reality are both true at the same time, and that tension is the most important thing to watch right now.
The AI-Dystopia–Futurama gap this week stands at –20 percentage points. That means Futurama narratives now outnumber AI-Dystopia narratives by the largest margin we've recorded across eight scans. To put that in context: back in scan two (14 April), AI-Dystopia led Futurama by 24 points. We've swung 44 points across the full series. That is not noise. That is a structural shift in how the media, researchers, and institutions are framing AI right now.
But before you read that as reassuring, hold on.
Three forces pushed Futurama to 45% this week – its highest share in the series. First, the global governance story has crystallised. The International AI Safety Report, backed by 29 nations, generated multiple high-quality sources all framing AI risk as manageable through human oversight. Second, the health and access story got a major signal: Anthropic and the Gates Foundation pledging $200M to embed Claude in global health systems, alongside Stanford's Health AI Week, frames AI as a broadly shared benefit under philanthropic stewardship. Third, DeepSeek's permanent 75% price cut on inference means the cost of accessing powerful AI is falling fast – which at least opens the door to wider benefit distribution.
None of that is fabricated optimism. These are real events with real institutional weight behind them.
But here is the structural tension the Futurama framing is papering over: the same week that Anthropic pledges $200M for global health, profitable tech companies cut 142,000 jobs to fund $700 billion in AI infrastructure. Those two facts exist simultaneously. The hopeful narratives are not wrong – but they are incomplete.
AI-Dystopia dropped to 25% this week – its lowest share since we began. Yet this week's AI-Dystopia stories are arguably the most materially significant in the series. The tech layoffs data from TechTimes, Tom's Hardware, and TechRadar is not speculative. It is documented. Nearly half of 80,000 Q1 layoffs were explicitly attributed to AI. Block cut 40% of its workforce and credited AI capability. Companies are not cutting because they're struggling – they're cutting because they're profitable and redirecting capital upward.
A self-replicating worm hitting 73 Microsoft GitHub repositories through AI coding tools adds a different AI-Dystopia register: weaponised AI spreading through critical infrastructure. That story appeared once, was classified correctly, and largely disappeared from the wider discourse. That absence matters.
AI-Fizzle sits at 13% for the second consecutive week, its lowest sustained level in the series. Early scans had AI-Fizzle as high as 18-19%. The sceptical counter-narrative – that AI ROI is overstated, that AGI claims are definitionally hollow, that only 16% of enterprises have scaled AI meaningfully (IBM Think) – is losing share of voice. This doesn't mean the sceptics are wrong. The ARC-AGI-2 benchmark result (AI scores 4%, humans near 100%) from the Medium piece is a striking data point that barely registered. The discourse is increasingly confident about AI's trajectory. Whether that confidence is warranted or self-reinforcing is the question AI-Fizzle is supposed to keep asking.
Singularia sits at 15%, roughly where it has lived for most of the series outside its early high of 16%. Jeff Bezos raising $12 billion for an 'Artificial General Engineer' and industry coalitions forming AI Safety Alliances both assume that sufficiently powerful AI is coming soon and requires coordinated management. That assumption is now baked into institutional behaviour, not just speculation.
Paperclipalypse remains at 3% – a series low it matched last week. The existential catastrophe framing has effectively vacated the mainstream discourse. The DEV Community piece on alignment and catastrophic risk stands almost alone. Whether that reflects genuine progress on safety or simply a shift in what institutions are willing to discuss publicly is an open question.
Zoom out across all eight scans and the story is this: we started in a world where AI-Dystopia dominated (38% peak in scan two), Paperclipalypse was surprisingly loud (19-20% in early scans), and Futurama was tentative. Over fourteen weeks, Futurama has climbed steadily to its current high while AI-Dystopia has fallen to its current low. AI-Fizzle has compressed. Paperclipalypse has nearly vanished. The discourse has, in aggregate, moved from alarm toward managed optimism.
What has not moved is the underlying material reality. Jobs are being cut. Infrastructure is concentrating. The worm in GitHub's repositories spread regardless of how many governance frameworks got published. The chart is trending hopeful. The question is whether that reflects the world changing, or just who is publishing.
There is almost no coverage this week of what displaced workers are actually doing next. There is no serious treatment of what happens to the 71-75% of AI initiatives that fail to deliver expected ROI – where does that capital go, and who decides? The governance coverage is almost entirely about frameworks being written, not frameworks being enforced. And the self-replicating worm story – potentially the most operationally dangerous AI story of the week – generated one classification in forty sources.
The discourse is increasingly comfortable. That comfort may be the most important thing to interrogate right now.
Challenge: If the AI-Dystopia–Futurama gap keeps widening at this rate, what would it take to convince you that the optimism is real – and not just a sign that the people most harmed by AI disruption have the least access to the publications we're tracking?
The articles most responsible for bending the five worlds this week.
This story is the most important counterweight to the week's optimistic trend. Companies are not cutting jobs because they're in trouble – they're doing it while making profits, and sending that money straight into AI infrastructure. That's a wealth transfer hiding inside a technology story, and it matters more than almost any governance framework published this week.
This is the single biggest signal pushing Futurama upward this week. Putting a powerful AI inside health systems in lower-income countries – with philanthropic rather than purely commercial backing – is exactly what a broadly beneficial AI future looks like. It's a meaningful event, not just a press release, and it anchors the hopeful narrative in something concrete.
The most alarming story of the week got the least attention in the wider conversation. A worm used AI coding tools to copy itself across dozens of major repositories – meaning the very tools being celebrated for boosting developer productivity were also being used as attack weapons. The fact that this barely registered in the governance or safety discourse is itself a signal worth tracking.
While several sources this week treated AGI as basically inevitable, this piece pointed out that current AI scores just 4% on the hardest reasoning benchmark while humans score nearly 100%. It also noted that the lack of a clear definition for AGI is being actively exploited by companies making big claims. It's a lonely voice in this week's scan, and that loneliness is the story.
Most of this week's business AI coverage was upbeat, but this report found that only 29% of companies see real returns on AI spending even after investing over a million dollars. That gap between money going in and value coming out is the quiet story running underneath all the productivity headlines – and it keeps the AI-Fizzle signal alive even as its share of the discourse shrinks.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free