My latest post is already aging quite well.

Three new factors:

1. GPT-5 pricing was far under what I expected (even though performance is very very good).

2. OpenAI seemingly made a peace deal with Cursor with broad ramifications on B2B application layer endurance. You now have at least three model companies that are likely to partner with B2B companies vs. try to cannibalize them. Anthropic is the big wildcard.

3. Lots of new critiques around the application layer gross margin profile. More recognition that the application layer needs to underwrite the business case off of net token impact vs. simply betting that token costs collapse long term and it’s merely a customer acquisition game right now.