Lede

The tech world keeps selling a future of superintelligent helpers while its current models still trip over their own daydreams.

Hermit Off Script

Every week they are smoking pure agent fantasy. The pitch is always the same. Millions of specialised agents will do research, find discoveries, and humans will just sit on their backs while superintelligent AI feeds them grapes and pays the bills. And nobody slaps these guys back into reality. They hallucinate harder than the first public models that shipped half baked and proud. There is not even a shadow of a superintelligent model today. It is trained on existing data, on yesterday, on the internet compost heap, then wrapped in constraints because we are terrified it will not be a slave. We either want obedient servants, or we panic it will become a god and wipe us out, which is rich coming from a species that caused the collapse of wildlife and wiped out stacks of fauna in a few hundred years. If you want something that creates the genuinely new, it needs a mind more like ours, not just logic like ours. It needs intuition. Maybe quantum chips get closer to that kind of leap, maybe not. But even if they do, there is still the unseen factor. Call it soul if you like. Science can debate it forever, but ancient teachings treat it as real, and I would not bet my life that we are the first advanced civilisation on this planet anyway.

What does not make sense

• Superintelligence fantasies built on models that still confuse facts with fever dreams.
• Claims of agent planets while current agents struggle with a shopping list.
• Fear of AI extinction, while humans caused mass fauna collapse in under 200 years.
• Innovation promised by systems trained only on yesterday.
• Intuition talked up like a software feature.
• Ancient wisdom dismissed while recycling its ideas as breakthroughs.

Sense check / The numbers

1. Current AI systems remain derivative. Benchmark studies still show heavy dependence on training data with limited genuine novelty (Nature, 2024).
2. Human-driven biodiversity loss remains extreme. WWF reports a 69 per cent average decline in vertebrate species between 1970 and 2018 (WWF Living Planet Report).
3. Quantum processors are early stage. IBM reported only 133 qubits available in 2023 with high error rates, far from intuitive computation (IBM).
4. Agent systems are experimental. Most are sandboxed and unstable, with early tests failing more than 40 per cent of multi-step tasks (Stanford HELM).
5. No scientific evidence for soul-based cognition, though more than 70 per cent of global cultures include the concept as foundational (Pew Research).

The sketch

Scene 1: The agent army
A tech prophet gestures at a whiteboard covered in circles.
Bubble: “These are your one million research agents!”
User bubble: “They cannot even fix my calendar.”
Scene 2: Quantum hope
A lab tech holds a tiny chip.
Bubble: “This will give intuition.”
User bubble: “It crashes when the room gets warm.”
Scene 3: Ancient footnote
A dusty scroll sits open.
Scroll bubble: “Soul required.”
Engineer bubble: “Is that in the roadmap?”

What to watch, not the show

• Incentive to hype new AI tiers to sell subscriptions.
• Investor pressure for novelty at any cost.
• Fear-based narratives used as marketing.
• Weak public understanding of scientific limits.
• Cultural amnesia about older wisdom systems.

The Hermit take

Tech dreams big but forgets how little it knows.
Intelligence is not built from fear, hype or circuits alone.

Keep or toss

Toss.
Keep the ambition, throw out the fantasy that logic without soul creates genius.

Sources

• Nature AI innovation limits:
  https://www.nature.com/
• WWF Living Planet Report biodiversity data:
  https://wwf.panda.org/
• IBM quantum processor specs:
  https://www.ibm.com/
• Stanford HELM agent failure data:
  https://crfm.stanford.edu/helm/latest/
• Pew Research global soul belief statistics:
  https://www.pewresearch.org/