Benedict's Newsletter: No. 635

OpenAI's competitive position

OpenAI has no unique tech, limited user stickiness, no network effect, and incumbents have matched the tech while leveraging superior product distribution — the plan remains unclear.

AI automation PE roll-ups

The thesis of buying legacy companies and making them efficient with AI seems 'a bit too deterministic'.

LLM hallucinations

Anyone claiming hallucinations are solved is an idiot — the author's own tests show ChatGPT still returns wrong or estimated figures even when correct data is explicitly available.

VR / Metaverse

VR headsets are good enough that if there was a 'there there' we'd have found it by now — no one cares, and the word 'metaverse' became functionally meaningless.

Kalshi prediction markets

The author sees no way that Kalshi's prediction markets are NOT gambling, regardless of how they're framed.

Frontier LLM models as commodities

Frontier models are commodities, but something can be both a commodity and really hard — the number of leading companies is shrinking just as it did with chips.

AI Automation PE Roll-Up

Jeff Bezos is raising a $100bn fund to buy legacy companies and make them more efficient with AI — a quasi-PE strategy betting that AI can unlock value in existing businesses rather than building new ones.

Instead of backing AI-native startups, this bets on applying AI to legacy businesses post-acquisition — a deterministic transformation thesis most VCs avoid.