Massively subsidized tokens have distorted AI demand.

· Bits and Bobs 5/4/26
  • Massively subsidized tokens have distorted AI demand.
    • When there's a massive subsidy there's no need to be efficient with tokens.
    • Two places subsidies are common today:
    • An insightful comment on HackerNews:
      • "I see highly trained engineers spend hundreds of thousands of tokens doing what can reliably be accomplished with 150 lines of python.
      • I think the push from management for us to use AI has made it so we don't have to be efficient with our consumption, so now we write md files which we feed to Claude in a loop instead of python and bash scripts to do routine tasks."
    • LLMs are insanely powerful and able to handle surprise in a way mechanistic software never could.
    • But they are ludicrous overkill for most tasks.
      • If you have a Max plan and rarely hit the quota, then there's no reason to not use Extra High Thinking even on simple tasks.
    • Subsidies must tighten.
      • First, we were in "burn as many tokens as we can" mode.
        • Explore.
      • Second, we'll get in "maximize the bang for tokens you burn" mode.
        • Exploit.
    • As subsidies tighten, there will be much more of a selection pressure for efficient and effective use of tokens.
      • If you use all of your budgeted tokens and hit the limit, then you are incentivized to figure out higher-leverage uses.

More on this topic

From other episodes