This week the gap between who AI helps and who it hurts got harder to ignore. The numbers have a story to tell, and it's not the one the press releases are selling.
New to Fizzle, Spread, Domination, Autonomy and Chaos? Here’s why I renamed the five worlds – the Aaronson & Barak originals appear in brackets.
Big companies are reporting their best-ever profits while cutting hundreds of thousands of jobs – and they're openly blaming AI for the cuts. That's the headline underneath the headlines this week. We track five possible futures for AI, from the hopeful Spread future where AI is a useful tool that makes everyone's life better, to the scary Chaos future where AI becomes so powerful nobody can control it. Right now the data says we're living somewhere uncomfortable in the middle: AI is definitely working, just mostly for the people who already have the most. The question nobody is answering yet is whether that's a bug or a feature.
The headline metric in this series is the Distribution Gap – the distance between Domination (AI-Dystopia) and Spread (Futurama). A positive number means the darker, more concentrated-power narrative is winning the discourse. A negative number means the optimistic, broadly-shared-benefit story is pulling ahead.
This week the gap sits at –17. That sounds like good news for the optimists, and in isolation it is – Spread holds 43% of all sources, Domination sits at 26%. But read the full arc and a more complicated picture emerges.
The series started ugly. Scan 1 (13 April): gap of +5. Scan 2 (14 April): gap surges to +24 – the widest Domination advantage in the entire series, the moment the concentration-of-power narrative was most dominant. Then a long, uneven retreat. By Scan 5 (18 May) the gap had flipped to –3, the first time Spread led. By Scan 8 (14 June) it hit –20, the deepest Spread advantage on record. Since then it has been narrowing – slowly, unevenly, but consistently. Scan 9: –16. Scan 10: –22 (a brief reversal). Scan 11: –20. Scan 12: –18. This week: –17.
That is three consecutive weeks of the gap tightening toward zero. Not a dramatic swing, but a direction. Spread is not collapsing; it's still the dominant single world in the chart. But Domination is recovering ground, and this week's layoff coverage is the engine.
Twelve sources landed in Domination this week – matching its Scan 12 count and continuing what is now a two-week stabilisation after months of decline from that +24 peak. The layoff cluster is the story: Forbes documenting 21,000 Oracle job cuts, TechCrunch's running log of companies citing AI while posting record revenues, CNBC reporting AI is now the leading stated reason for layoffs (40% of May 2026 cuts). These aren't soft trend pieces. They're receipts.
The Alphabet Workers Union petition – 4,500 signatures delivered to Sundar Pichai – is the one source that cuts against the corporate framing. Workers are organising. That's a data point the other layoff trackers don't carry.
The geopolitical layer adds texture. Japan and Nvidia's 27,500-GPU sovereign AI factory, Huawei's Atlas 950 SuperPoD, and Xi's WAIC keynote offering AI training slots and weather systems to thirty nations are all classified Domination – not because AI is inherently dangerous in these stories, but because the infrastructure is concentrating inside a small number of state-backed power centres. The White House framework preempting state AI regulation fits the same pattern from a different direction.
Spread (Futurama) holds twenty sources – its second-highest raw count in the series, behind only its Scan 8 peak of 45%. The enterprise data is genuinely contradictory this week, which is what makes it interesting rather than reassuring. Digital Applied reports 6.4 hours saved per worker per week. Futurum Group sees agentic AI shifting enterprise ROI from productivity metrics to P&L outcomes. PwC and Deloitte both frame AI as a manageable, human-governed productivity multiplier.
But those sources exist alongside the Fizzle evidence, and the gap between the two is uncomfortable. Writer.com finds only 29% of enterprises see significant ROI. Terminal X cites S&P Global data showing 42% of companies have abandoned most AI projects. The enterprise story isn't Spread vs Fizzle – it's both, simultaneously, depending which part of the market you're looking at.
Fizzle (AI-Fizzle) drops to 11% – its lowest reading in the entire thirteen-scan series. That's worth flagging. The ScienceDaily attention-span study (top AI models deteriorate sharply as task length grows) and the anthropomorphism-in-language study are doing real work here, but five sources is a thin presence. The fizzle narrative – that AI is fundamentally overhyped and underperforming – is losing airtime precisely when the ROI data suggests it should be gaining it. That's a structural absence worth watching.
Autonomy (Singularia) holds at 13%, consistent with its post-Scan 9 range. Anthropic President Daniela Amodei's suggestion that Claude may already represent a form of AGI is the single most consequential source in this category – not because it settles the debate but because it moves the Overton window on what "AGI" means in a commercial context.
Chaos (Paperclipalypse) is at 7% – up from 5% last week and its highest since Scan 4. The Cloud Security Alliance's alignment gap paper and the Future of Life Institute's safety index (nine leading AI companies evaluated, experts warning of unprepared races toward recursive self-improvement) are the drivers. It's still a small signal. But Chaos has now ticked up three scans in a row. That pattern hasn't appeared before in the series.
Thirteen weeks of scanning and the discourse has produced a clear macro story: AI productivity is real, AI concentration of power is real, and the question of who captures which outcome is almost entirely absent from the mainstream coverage. The layoff trackers document the displacement. The enterprise reports celebrate the productivity. Almost nothing connects the two into a distributional question: who exactly is 7% richer because someone else lost their job?
That's what's missing. Not fear, not optimism – accountability for the arithmetic.
The articles most responsible for bending the five worlds this week.
Oracle cut 21,000 people while posting record profits – and named AI as the reason. This story is the clearest single example this week of AI working exactly as advertised for shareholders while doing the opposite for workers. It's the Distribution Gap made concrete.
When AI becomes the number-one reason companies publicly give for laying people off, that's a turning point in how the story gets told. This report also asks whether AI is being used as cover for cuts that would have happened anyway – a question the other trackers don't raise.
Only 29% of companies see meaningful returns despite huge spending and near-universal executive enthusiasm. Sitting right next to the glowing productivity reports, this source creates the central contradiction of the week: AI is working brilliantly and failing badly, at the same time.
Saying out loud that today's AI might already be AGI is a significant move – not because it's definitely true, but because it shifts what we think we're building toward. When a major AI lab president makes this claim, it changes the political and regulatory conversation around what 'control' even means.
This report argues that the safety methods we have now won't work on the systems we're building next – and that the gap between them is already open. Combined with Chaos hitting its highest reading since Scan 4, it's a warning sign that deserves more attention than it's currently getting.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free