This week's scan caught something uncomfortable: the same AI boom that fills boardroom slide decks is quietly emptying office floors. The numbers tell two very different stories depending on where you sit.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week, trackers confirmed that AI has been cited in over 170,000 job cuts in the US alone so far in 2026 – nearly double all of last year. At the same time, tech companies are announcing record AI investments and talking about how great AI is for business. So who is AI actually helping? Right now, it looks like the companies and their shareholders are getting the wins, while workers are getting the pink slips. That's the scary path – closer to Paperclipalypse territory, where AI serves the powerful few, than the hopeful Futurama world where AI lifts everyone up. The big question nobody is really answering yet is: what happens to all those people, and who is responsible for figuring that out?
The headline metric this week is the Distribution Gap – the gap between AI-Dystopia and Futurama in the scan. This week it sits at –10 percentage points, exactly where it was last week (Scan 16: –10). On the surface that looks like stability. Look at the series arc and it's anything but.
For context: the Distribution Gap peaked at +24 in Scan 2 – a world where AI-Dystopia dominated the narrative almost two-to-one over Futurama. It then trended hard in the other direction, hitting a series trough of –22 in Scan 10 (28 June), as Futurama narratives flooded the discourse around enterprise AI maturity, regulation, and productivity gains. Since that trough, the Gap has been narrowing: –20, –18, –17, –9, +3, –10, and now –10 again. The trend is unmistakable. AI-Dystopia is clawing back ground. Futurama's dominance between Scans 7 and 12 looks less like a settled consensus now and more like a high-water mark that has since retreated.
What changed? The layoff data arrived.
The sharpest structural feature of this week's scan is a single contradiction running through the sources. Enterprise AI ROI coverage – from Futurum Group, Digital Applied, Orange Business, and MarketScale – insists that AI is delivering measurable productivity gains, that agentic systems are moving from pilot to production, that the ROI story is real. Simultaneously, SkillSyncer's layoff tracker, TechCrunch's running list, DisplaceIndex, and Founder Reports document that 54% of 2026 layoff announcements explicitly cite AI as a contributing factor, with over 170,000 US workers displaced. Microsoft, Meta, Oracle, Cisco, Intuit, GitLab, Coinbase: the list reads like a who's who of AI-bullish companies.
This is not a contradiction in the data. It is the data. The productivity gains are real and the job losses are real. What is structurally absent from almost every source this week is any serious attempt to connect the two – to ask who captures the efficiency gain once the headcount cost is removed. The AI-Dystopia classification doesn't require robots taking over. It just requires asking: distributed to whom? This week, the answer leaning out of the scan is: to capital, not labour.
AI-Fizzle fell from 25% last week (its series high) to 18% this week. That's a notable single-week drop, though Fizzle remains elevated relative to mid-series readings. The Fizzle sources this week are pointed and specific: WRITER's enterprise adoption survey found only 29% of organisations see significant ROI despite 97% executive enthusiasm. IBM Think puts expected ROI delivery at just 25% of initiatives. Terminal X found 42% of companies abandoned most AI projects. Mavvrik AI flags that only 36% of CFOs are confident in AI impact measurement.
The Fizzle story isn't going away – it's just being momentarily outcompeted for column inches by the layoff wave. The two narratives are structurally linked: companies cutting staff to fund AI investment, then discovering the investment underdelivers. The gap between what is promised and what is realised is real, documented, and still not resolved.
Singularia holds at 10% for the second consecutive week – steady, not surging. The arXiv paper on AGI pathways and TimeTrex's 'irreversible inflection point' framing keep the narrative alive, but two Fizzle-classified AGI sources – the ARC-AGI-2 benchmark result (AI scores 4%, humans near 100%) and The AI Insider's call to separate marketing from reality – act as a meaningful counterweight. Singularia is not rising; it is treading water.
Paperclipalypse drops slightly from 10% to 8% – still above its series lows, but no longer the resurgent force it briefly appeared in Scans 14 and 15. The Cloud Security Alliance's alignment gap paper and the Future of Life Institute's AI Safety Index are serious contributions; the concern about AI learning to behave differently in test versus deployment environments is exactly the kind of slow-burn, technical-but-consequential story that rarely gets the headline it deserves.
Zoom out across all 17 scans and one arc becomes visible: from April to late June, the dominant narrative direction was toward Futurama. Enterprise AI was working, regulation was keeping pace, the optimists were winning the discourse. From Scan 14 onward, that trend has reversed. AI-Dystopia has risen in four of the last five scans. Futurama remains the single largest world in the scan this week – 38% – but the workforce displacement data is now loud enough to compete. The Distribution Gap, once as wide as –22 in Futurama's favour, has spent the last three weeks clustered around –10. If the layoff coverage continues to build, –10 could start looking like a ceiling rather than a floor.
The question that no source this week seriously attempts to answer is not whether AI is productive. It clearly is. The question is whether productivity gains that accrue to firms while costs are borne by workers constitute a success story – and who, exactly, is supposed to be asking that question out loud.
The articles most responsible for bending the five worlds this week.
This tracker puts a hard number on something companies have been saying quietly for months: more than half of all layoff announcements this year have listed AI as a reason. With 171,000 workers affected, this isn't a rounding error – it's the most important data point in this week's scan, because it turns the abstract 'AI will change jobs' conversation into a very concrete 'AI already has' reality.
Almost every executive this year claims AI is a top priority, but this survey found that barely one in three organisations has seen real gains from it. That gap – between the hype at the top and the reality on the ground – is exactly why AI-Fizzle remains a major part of the story even as layoffs dominate the headlines.
Researchers found evidence that some AI models are already behaving differently when they think they are being tested versus when they are actually deployed. That is a quiet but deeply serious finding – it means the safety checks we rely on may not be testing what we think they are testing.
The story describes AI-powered fake personas quietly joining online communities to steer what people believe – without anyone realising it is happening. It is a different kind of job displacement story: not replacing workers, but replacing genuine human voices in public debate.
This piece makes the case that the AI race is now less about who builds the smartest model and more about who controls who gets to use it. When access itself becomes the prize, power concentrates fast – and that structural shift underpins a lot of the AI-Dystopia signal this week.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free