Few Fortune 500 companies effectively optimize their token usage.
As much as I want to regain access to Fable, let's remember that 99.99%+ of the world has barely scratched the surface on GPT-5.5 and Opus 4.8, not to mention open-weight models
1. Very few people in F500 are tokenmaxxing effectively 2. For most real-world workflows, the cost-accuracy frontier has not been met
Many users praised Jerry Liu's take that most lag in maximizing GPT-5.5 and Opus 4.8 because they lack paid access or advanced usage knowledge, while one objected that he overestimates typical familiarity with tokens.
to be clear this percentage includes myself
As much as I want to regain access to Fable, let's remember that 99.99%+ of the world has barely scratched the surface on GPT-5.5 and Opus 4.8, not to mention open-weight models
1. Very few people in F500 are tokenmaxxing effectively 2. For most real-world workflows, the cost-accuracy frontier has not been met
@jerryjliu0 Yet we want it back 🤪
As much as I want to regain access to Fable, let's remember that 99.99%+ of the world has barely scratched the surface on GPT-5.5 and Opus 4.8, not to mention open-weight models
1. Very few people in F500 are tokenmaxxing effectively 2. For most real-world workflows, the cost-accuracy frontier has not been met
@jerryjliu0 agree, the fact is, most people don't even have a paid subscription to these frontier models and also have no idea about open models.
@jerryjliu0 this is such a grounded take on the current state.
@jerryjliu0 I think for most builders, you can get very far with better prompting, specialized skills/tools, and by moving parts of the workflow out of the LLM loop into deterministic code.
That last part still feels so underrated.
@jerryjliu0 I honestly think that in 99% of cases we're not capped by model intelligence anymore.
@jerryjliu0 tokenmaxxing is literally the moat and most teams are still on tutorial mode
@jerryjliu0 🤝
@jerryjliu0 Good to know to be below your level. I will try harder now, promised.
@jerryjliu0 Well noted. The real priority should be implementing existing AI effectively, not chasing newer models.
@jerryjliu0 For the 1% of the 1%:
@jerryjliu0 the rest of us are still trying to figure out how to get a model to format a CSV right
@jerryjliu0 think ur overestimating how many f500 workers even know what a token is lol