Open-source parity narrows the moat while scaling returns diminish — the hype cycle wobbles
“Multiple senior researchers departed over safety culture concerns”
“Grok benchmark position improving but still not frontier”
“Gemini image generation controversy revealed rushed deployment culture”
“Amazon and Google backing creates dependency risk, not security”
“Llama adoption across enterprise and research represents genuine success”
Intel Drop — February 24, 2026
The industry sits at 61/100 — Extra Crispy territory — which feels about right. Five companies racing to build artificial general intelligence while their moats erode daily, their talent gets poached, and open-source communities replicate their products for free. The hype cycle is wobbling.
OpenAI leads the cookedness leaderboard at 72, a distinction that combines genuine business risk (talent drain, open-source competition) with maximum drama (Sam Altman's governance chaos is the gift that keeps giving). xAI follows close behind at 68, where the Elon Musk chaos multiplier inflates the drama score to near-maximum levels regardless of what Grok actually ships.
Meta is the plot twist: the least cooked at 38, because choosing to give away Llama means they can't be disrupted by the very strategy they invented. Anthropic and Google sit in the middle, both heating up but not yet crispy — they have real resources and real products but face the same structural question everyone else does: what's your moat when Llama runs on a laptop?
- 1.Open-source model quality now matches proprietary models on most benchmarks
- 2.AI company talent drain continues as researchers move to better-funded startups
- 3.Meta's Llama strategy appears to be working — commoditizing the competition
- 4.SEC filings show Meta and Google R&D spend growing but returns unclear
- 5.Elon Musk's political entanglements creating enterprise hesitation around xAI