This week's scan hit its highest AI-Dystopia reading ever – 62%. While labs claim we've entered the age of superintelligence, thousands more workers got their marching orders.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week, a lot of big news happened around AI – and most of it wasn't great for ordinary people. Companies like HubSpot cut hundreds of jobs, OpenAI fired safety researchers who raised concerns, and the White House officially started calling AI 'superintelligence.' The tool that was supposed to help everyone is starting to feel like it's helping a few people at the top and leaving everyone else to figure it out. That's closer to the scary Paperclipalypse future – where AI runs ahead of anyone's ability to control it – than the hopeful Futurama future where AI stays a helpful tool and the benefits are shared around. The good news is that people are noticing, and some are pushing back.
Sixty-two percent. That's the share of this week's sources classified as AI-Dystopia – the highest single-week reading in 27 scans of this series. Not a blip, not a rounding error. The week of 5–11 October 2026 produced the most concentrated cluster of harm-and-concentration signals we've recorded. And the Distribution Gap – the spread between AI-Dystopia and Futurama – hit 47 percentage points, a new series record.
Let that settle for a second before we look at what drove it.
The Distribution Gap is the centrepiece metric of this series because it answers the only question that matters at a societal level: is AI spreading its benefits broadly, or concentrating them at the top while distributing the costs downward? A wide gap means the evidence skews heavily toward concentration. A narrow gap – or a negative one, as we saw in the 7 June scan at –4 points – means the optimistic, human-directed-tool narrative is holding its own.
Here's the arc. The gap opened the series at 10 points in April, bounced between 1 and 15 for most of the year, dipped into negative territory once (7 June, –4), and then began a slow climb through August and September. The previous record was scan 26 last week, at 40 points. This week broke that record decisively.
Two consecutive record highs. That's not noise. That's a trend change.
The Futurama signal hasn't collapsed – at 15% this week it's recovered slightly from last week's anomalous 7%. But AI-Dystopia has surged so hard that the gap widens regardless. The optimistic narrative is simply being drowned out by the volume of harm signals.
Three story clusters drove this week's AI-Dystopia dominance.
First, the layoffs. Yahoo Finance's running tech layoffs tracker, the Boston Globe's piece on HubSpot cutting 700 jobs, and Intellizence's report on Workday trimming 525 roles – together they paint a picture of AI-driven restructuring that transfers risk cleanly downward. Workers adapt or exit. Capital reallocates upward. The Globe's framing was particularly sharp: AI-native firms are decimating incumbent software valuations, and it's the people inside those incumbents who pay the price.
Second, the safety vacuum. ABC News reported that OpenAI fired three safety researchers who raised concerns about AI risks – not quietly managed out, fired, in a public dispute. OriginBrief's policy weekly described a US governance vacuum that AI agents are already filling. The Democrats' regulatory battle piece in Transformer News catalogued proposals ranging from liability bills to a literal Ban Artificial Superintelligence Act. When mainstream legislators are drafting bills with the word 'ban' and 'superintelligence' in the same title, something has shifted in the political temperature.
Third, and most structurally important: OpenAI's AGI claim and the White House's rebranding of AI as 'super intelligence.' Live Science asked seriously whether GPT-6 Astra has demonstrated general intelligence. Tech Diplomacy Now reported Trump's Executive Order making 'super intelligence' official government language. These sit in Singularia – the world where machine capability crosses into qualitatively new territory – but they feed the AI-Dystopia signal too, because they're happening without the governance architecture you'd want before crossing that threshold.
Here's what I find more revealing than what's present: there is almost nothing in this week's scan about AI creating new jobs, retraining programmes, shared productivity gains, or communities that are measurably better off. The MarTech piece on AI budgets outrunning ROI proof – the lone AI-Fizzle source – is the closest thing to a counter-narrative, and it's not hopeful, it's just sceptical. The Futurama signal this week is two newsletter-style roundups cataloguing model releases and enterprise partnerships. Useful, accurate, but thin.
When the optimistic narrative is being carried entirely by product launch coverage and the harm narrative is being carried by labour economics, regulatory fights and safety governance breakdowns, the Distribution Gap reflects a real asymmetry in what's actually happening – not just in how journalists are framing it.
Step back and look at the full 27-scan trajectory and here's what a leader needs to understand: for the first six months of this series, AI-Dystopia and Futurama oscillated. The gap moved but it moved both ways. There were weeks where the optimistic signal genuinely competed. From roughly scan 21 (30 August) onward, something changed. AI-Dystopia locked above 30% and didn't come back down. Futurama stayed in the low-to-mid twenties. The gap started widening. The last two weeks have accelerated that divergence sharply.
The Paperclipalypse readings – the existential-risk world – had a long run of strength through scans 22–25 (six-plus weeks with readings at 23–30%), then dropped to zero this week. That's not reassuring. It means the existential framing has been replaced by something arguably more immediate: the everyday, grinding, institutional harms of AI deployed at scale without adequate oversight. The danger has moved from theoretical to concrete.
The question for anyone leading an organisation right now isn't whether AI is changing things. It manifestly is. The question is which side of the Distribution Gap your decisions are landing on. Are you capturing the productivity gain and distributing it – or capturing it and moving on? The organisations that figure out the former have a story to tell. The ones that don't are contributing to the data that's making this chart look the way it does.
Ask yourself honestly: if your AI decisions showed up in next week's scan, which world would they belong to?
The articles most responsible for bending the five worlds this week.
OpenAI didn't quietly move people on – it fired safety researchers who raised concerns about AI risks out loud, in a public dispute. That tells you something important about whose interests are being protected when safety and speed come into conflict at the world's most powerful AI lab.
OpenAI is now saying AI has crossed into a genuinely new era of intelligence – and a serious science publication is treating the question as real, not ridiculous. Whether or not you believe the claim, the fact that it's being made publicly and taken seriously changes what the next conversation about AI safety and oversight has to look like.
HubSpot cut 700 jobs and its stock fell sharply as AI-native competitors took market share – and this is Boston, not a rustbelt town. When AI disruption hits established, well-funded tech companies this fast, it's a signal about the pace of the shift, not just its direction.
The US government has officially stopped calling it 'AI' and started calling it 'super intelligence' in executive orders. That's not just a language change – it's the government signalling that it believes something qualitatively bigger is happening, which makes the absence of binding safety rules even more striking.
Four in ten marketing technology leaders can't show that their AI spending is actually delivering financial value, and most AI agent projects are still stuck in pilot mode. In a week dominated by big claims about superintelligence, this quiet finding is the most grounding data point of all.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free