I get what you're saying about embedding data, but see no valid reason to support it as a feature. Prior to v30, 80b of op return data was supported and other methods were tolerated abuse. I'm concerned that now with Core's full support other chains will pile on top of Bitcoin, rather than building their own network. Ethereum chain is 15TB in 11 years. Does BIP-444 not address all of these methods?
Programmable money I'm good with, as long as it isn't enabling arbitrary data on chain.
They're consensus valid now. If BIP-444 or similar is successful, they may not be in the future. The whole "if you don't like it, leave" idea could be equally applicable to developers who want to change things like mempool policy. It seems to me like some want to turn it into BSV or Ethereum. I would prefer they just use those chains if that's what they're into.
What makes Bitcoin unique is not JUST that it was the first successful cryptocurrency, nor that it has finite scarcity, but ALSO that it tries to do one thing rather than everything.
Developers will always want it to do more things. At some point this will become unsustainable, without a clearly defined purpose. It also creates an ever larger attack surface on one of the most disruptive technologies created.