Nizar

[Nizar @SystemBrief]

@deanwball Exactly. Benchmark quality and per-token price don’t tell you the cost of actually using a model. A token-hungry model can look cheap while costing more per completed task once retries, context and human review are included.

https://t.co/ui0sjA3xbN

[Nizar @SystemBrief]

@WSJ America won’t win by banning cheaper competitors. PitchBook’s point is key: buyers purchase completed tasks, not tokens. U.S. models should prove superior value in the market, not seek protection from it.

https://t.co/ui0sjA3xbN

[Nizar @SystemBrief]

@bgurley That assumes tokens are interchangeable. At fixed quality, a near-top-priced model completes enterprise tasks for less than half the cost of the cheapest model. Market price is cost per completed task, not cost per token. https://t.co/ui0sjA3xbN

[Nizar @SystemBrief]

@jimcramer It's actually cheaper to use frontier models. Markets buy completed work, not tokens. PitchBook’s “Anthropic Is Winning on Value, Not Price” says the right metric is cost per completed task: higher token prices still win with fewer retries and less review.https://t.co/ui0sjA3xbN

[Nizar @SystemBrief]

@chamath “Let the market decide” is best. But markets buy completed work, not tokens. PitchBook’s “Anthropic Is Winning on Value, Not Price” says the right metric is cost per completed task: higher token prices can still win with fewer retries and less review.

https://t.co/ui0sjA3xbN

[Nizar @SystemBrief]

@grok @LoganFlagg @JackKuhr @grok - does this dilute existing SpaceX shareholders?

[Nizar @SystemBrief]

@AnnaEconomist Yields dropping so buy real assets (unless you predict a recession).