[theptip]

It’s pretty reasonable to say “demand is way up, quality is up, supply is constrained, and so price needs to rise”.

It seems weird to segment this way though. Surely it’s better to just give Sonnet to your bottom tier, rather than cut out the entire Claide Code product entirely?

Give folks a taste rather than lock the whole product behind a $100/mo plan.

[mewpmewp2]

But if Sonnet is bad it would give bad impression of the product, no? And it also takes compute, so you give a bad hallucinating impression of your product while still losing compute.

[theptip]

It’s not bad though, it’s crazy good in comparison to any model older than 1y old. If you don’t have access to any vibe coding at all it’s gonna be life changing.

But I think you are right, as long as Codex and Gemini are cheap alternatives then vs. 1yo models isn’t the correct comp.

Then it’s probably better to just resegment the whole Claude Code product as an enterprise only tier. (That also has the advantage of kicking out all the Claw subscribers that screw over the token limit economics for normal $20/mo users.)

[wobfan]

I mean, this is why they do A/B testing. This way of testing stuff is not new at all, people who act genuinely surprised need to do a reality check. Companies want to maximize profit. They do this by testing what creates the biggest profit. A/B Testing is one of the ways to do this, and it has been used for decades in precisely this way.