Software is 10x more valuable than inference tokens because tokens do nothing for the user, just like a database request by itself does nothing.
Software is what makes inference valuable because it builds a workflow that transforms tokens and data into practical benefits.
Look at the payment plans for Lovable, Figma Make, Claude Code. None of them charge by token. They charge by obfuscated 'credits'. We don't know the current credit economics, but it is certain that the credit markup will increase and probably eventually reach 10x of the token cost. Users will gladly pay for it because again, tokens do nothing for them. It is the Claude Code, Figma Make products that make them productive.
The claim is too absolute. Software amplifies value, but inference cost and capability still shape what’s possible. Users aren’t demanding obfuscation; they just want predictable pricing and clear ROI. Does anyone want hidden math in their pricing?
In many markets, transparency wins. Think of Carfax or banking fees or airbnb pricing for example, when regulators or competitors force clarity, buyers benefit and trust grows.In a functioning government that serves the people (regardless of party) we would see this
People believe they “need” these AI products partly because they’re saturated in both earned and paid media. In '23 there were nearly 400k articles covering AI. I think we can all safely assume its more now, and when we include financial reporting, quite inescapable.
Software is what makes inference valuable because it builds a workflow that transforms tokens and data into practical benefits.
Look at the payment plans for Lovable, Figma Make, Claude Code. None of them charge by token. They charge by obfuscated 'credits'. We don't know the current credit economics, but it is certain that the credit markup will increase and probably eventually reach 10x of the token cost. Users will gladly pay for it because again, tokens do nothing for them. It is the Claude Code, Figma Make products that make them productive.