Anthropic's 'model is the product' philosophy — not hiring many people and expecting the model itself to do the marketing and convincing — is a 'father knows best' approach that is somewhat misaligned to humanity because it doesn't listen to or respect user feedback and opinions.
Pash argues Anthropic's strategy of letting the model itself be the product, rather than actively responding to user feedback the way OpenAI does with early adopters, amounts to a paternalistic 'father knows best' approach that is somewhat misaligned to humanity. ✦ AI generated
Pash · The Cognitive Revolution · 2026-07-08 · original ↗
starts at this moment · 28:56
“Do you have any other kind of uh synthetic takeaways or you know how would you summarize what you've seen in terms of who who prefers what at this point?”
I I I would say that I think the anthropic method is a little bit misaligned to humanity because it means that you're not listening to the people who uh actually have opinions and you're not respecting those opinions or uh resolving them as quickly as possible. It's more of a father knows best uh kind of framework where you know we think this is the way it should be
verbatim transcript · starts at 28:56
28:56right? So, um yeah, it's just it's just a different different way of like living in the world. Um I I I would say that I think the anthropic method is a little bit misaligned to humanity because it means that you're not listening to the people who uh actually have opinions and you're not respecting those opinions or uh resolving them as quickly as possible. It's more of a father knows best uh kind
29:19of framework where you know we think this is the way it should be and you guys are going to have to if you if you like it you can come up with this or not you know too bad. Um so it is what it is. Couple other things that jumped out to me that I I don't know that we can do justice to, but I would at least point people to
29:42check them out. Um, Li Lillian Wang, who's at um, Thinking Machines now, I believe, right? Former OpenAI, yeah, now co-founder of Thinking Machines, um, posted a very well-received blog post on harness engineering for self-improvement. And there's also one from Michaela Katasta who's the AI lead at Replet, continual learning for agents. And they're both really, you know, kind of sign of current times, I think, in the sense that we've been
30:17talking about, you know, recursive self-improvement. and it is now happening where it's not just anthropic people saying that it's starting but you're starting to see these um you know I guess you'd call thinking machines a neolab probably and in the case of replet I don't even know what to call replet at this point you know an agentic u coding platform but these different kinds of companies that have been aggressive early adopters
30:48are starting to raise their hands and say we've closed the loop too and that's how we're progressing so fast. That was the um Amjad from Replet basically just said like people people have been asking us how is Replet improving so quickly lately and the answer is we've closed the loop. The agent is now improving itself and here's how. Um, I definitely want to spend more time with those and I
- ·Anthropic lets the model itself be the product
- ·Fewer people actively engaging early adopters
- ·Model expected to do its own convincing
- ·Pash: approach is 'a little bit misaligned to humanity'
- ·Doesn't listen to or respect user opinions
- ·Decisions made as 'we think this is the way it should be'
- ·Contrasts with OpenAI's responsiveness to feedback