My takeaway: charge for usage but sell predictability. In selling, the edge is data only you have plus showing up in person. And on Apple's new EU terms, RevenueCat's math says lower fees alone rarely justify web checkout for smaller apps. https://t.co/JxEe3ztiSQ
Today's business read: pricing is moving to usage, but buyers still want to know the bill in advance. Plus HubSpot's restructuring, why AI SDRs flop in small markets, and what actually gets you named in AI answers. Here's what stood out:
Christian Strunk checked 1,032 AI answers. The best-linked third of companies were named in 72% of answers, the least-linked third in 35%. Rerun 3 hours later, 18% of those results flipped. Links still matter, and AI visibility is noisy. https://t.co/JUwse4YudB
My read: the cheap tier is now good enough for many agent calls. One new paper found the best local small model matched hosted models at 4 of 5 call sites. The hard problems are moving to routing, containment, and how we train juniors.
Frontier AI got cheaper and moved closer to the device on the same day. OpenAI put GPT-6 in ChatGPT's free tier, Anthropic cut small-model prices by roughly 75-90%, and Microsoft and NVIDIA made Windows a governed home for local agents. What mattered:
Best study today: an RCT with 133 patent attorneys. AI lifted drafting quality for all (+0.38 SD at 90 days), but the unassisted skill gain came entirely from seniors. Juniors saw no average gain. AI raises the work, not automatically the worker. https://t.co/SYh95Ed9Tz
I recently came back to X after a long time after LinkedIn became boringโฆ I realized people have so many questions here.. polls, open ended questions and stuffโฆ if you want survey people use survey apps manโฆ please donโt spoil X just like everything else.