Open-source models already eat 30% of token spend — the real AI app layer is barely being built
bankhosa_01 · reddit · 2026-11-22
A Reddit essay pushes back on the narrative that three frontier labs will own the entire software economy: token prices are collapsing, open-source models already capture 30% of token spend, and top labs run nearly identical transformer architectures underneath — a utility market like AWS/Azure/GCP, not a sacred monopoly (ex-OpenAI researchers building Anthropic and taking major enterprise share in two years is the proof).
Labs layering bloated vertical features dilute their focus, and 95% of real business work never needs frontier AGI reasoning: a solid mid-tier model wrapped in a real execution harness, clean cross-tool memory, and deterministic guardrails beats a raw frontier model operating blind.
Conclusion: labs capture the compute layer, and that's it — AI's real value sits in a new application layer above commodity models that most builders haven't started building yet.
More from AGI Musings
- NeurIPS 2026 position paper: agentic AI scientists aren't built for autonomous discovery — CatAstro_Piyush · 2026-11-22
- Grady Booch: LLMs are architecturally incapable of abductive reasoning; symbolic layers essential — Grady_Booch · 2026-11-22
- teortaxesTex mocks BAP's doomer AI forecast: "BAP is in his doggy era" — teortaxesTex · 2026-11-22
- Writing works subconsciously: metaphors, ambiguity, and how LLMs read like pond reflections — erikphoel · 2026-11-22
- You're allowed to hate your AI, just not love it: the gendered double standard — serialchilla91 · 2026-11-22
- Debate erupts over Alpöge/Claude Riemann proof: so broad no one may grasp it all — aran_nayebi · 2026-11-22