June 29, 2026
Hall Pass for Hallucinations?
The Permission Slip
AI’s trillion-dollar excuse note? Commenters are absolutely not buying it
TLDR: Cringely argues the AI world may have used “bigger models will fix it” as an excuse to ignore chatbots making things up, while claiming a small team found a cheaper fix. Commenters were deeply split: some said the warning is overdue, others called it hype, self-promotion, or a fake solution to the real problem.
The real fireworks around "The Permission Slip" aren’t just about one essay or one AI boss — they’re about whether the whole industry got handed a very expensive excuse to keep spending mountains of money and hope the mistakes magically disappear later. The article argues that big AI companies have basically treated “just make it bigger” as a free pass, even for the glaring problem of chatbots confidently making stuff up. Cringely’s bombshell claim: his small company says it tackled that problem with a cheaper design instead of endless spending.
And wow, the comments did not quietly nod along. One camp said, hold on, maybe giant spending still makes sense if only huge systems can do things smaller ones can’t — like advanced bug hunting and security work. Another camp flat-out rejected the article’s central premise, saying these systems are already known to be unpredictable, so acting shocked about mistakes feels outdated. Then came the sharpest knives: one commenter mocked the whole thing as a “plane that never leaves the ground,” arguing that avoiding made-up answers by tightly limiting what the system can say is not the same as solving the underlying problem.
The funniest drive-by? A brutal commenter dismissed the piece as “Claude-generated slop about how Claude is wrong” with “self-aggrandizement” on the side. Ouch. So while the article tried to expose a trillion-dollar gamble, the crowd turned it into a brawl over hype, honesty, and whether this was a warning siren — or just another tech guy victory lap.
Key Points
- •The article says Dario Amodei’s *Machines of Loving Grace* argued that scaling compute would eventually solve hard AI problems, including hallucinations.
- •It argues that this view encouraged AI companies and investors to prioritize spending on larger models and infrastructure rather than directly addressing hallucinations.
- •The article cites major spending examples including Stargate, hyperscaler investments, and an Anthropic-Akamai arrangement as evidence of the industry’s scaling bet.
- •The author says 2Brains Inc., founded in 2022, addressed hallucinations through an architecture that separates language generation from fact retrieval and verification before user output.
- •The article concludes that either scaling will not remove hallucinations or it will do so only at very high cost compared with architectural alternatives.