At least we know it wasn't written by a bot

For what it's worth, I enjoyed reading your article. I think it will resonate with many.

I've never used any of the stuff you've released for free, but in the name of those who did use it -thank you

just a clarification, this was not written by me (maybe i should clarify that in the title, i just copied the article title). but it did resonate with me, especially the bit about free users because we tried using the same model with my startup and failed. so yeah lessons for next time

Time for only-AI hacker news

Does it use AI to filter AI?

Is it a move to sell more of the paid Google Search API calls? If so, that's a sign of distress.

Surely, the cost of serving search results to bots isn't that high

Putting together all I know about tokenminning. Point your agent at it.

1. Prompt hygiene (schema over prose, make no mistake) 2. Cache whenever you can 3. Context hygiene (seriously, start a new chat, it doesn't take much) 4. Routing 5. Control max_tokens, be smart about RAG 6. Semantic caching

What works for you?

Author here. Continued Apollo's analysis from August. Data: US Census Bureau's Business Trends and Outlook Survey. It tracks AI usage across 1.2m US businesses every 2 weeks.

- Large enterprises (250+ employees) peaked at 13.4% in July 2025, dropped to 11.7% - Every firm size category declined since July except smallest (1-4 employees) - Smallest firms continued rising: 9.65% → 10.3% - First sustained multi-month decline since tracking began Nov 2023

OG research: https://www.apolloacademy.com/ai-adoption-rate-trending-down...

Why do you think this is?

Which one should we implement first?