nothing accidental
there is nothing accidental about anthropic's connector "leak." there never is.
anthropic surfaced nine integrations at once - photoshop, blender, ableton - with a $280K/year blender fund and university partnerships attached. when rock stars in the 80s wanted to make the loudest statement, they kept silent and let the rumours leak. anthropic are modern rock stars and these are their rumours
generate for me \\ do it with me
openai and anthropic are staring at the same creative market and placing opposite bets.
openai: generate for me. build it all inside chatgpt. describe what you want and get the output. never touch a knob. feel the ✨vibes✨
anthropic: do it with me. don't replace the tool - be a tool extension. a photoshop user stays in photoshop. an ableton producer stays in ableton. claude becomes a copilot that happens to speak the application's language, not a reason to close it.
both bets can work, but they are aimed at entirely different users.
if you actually make music
music isn't a hobby for me - i'm a professional musician and i'm not even shy admitting it. that context matters, because the two approaches feel completely different when you're the person they're supposedly helping.
the "generate for me" approach is genuinely impressive as a demo. you describe a feeling, you get the result back immediately. you ✨vibe✨ and you pray that the result sounds like whatever was in your head. but for someone who knows exactly what they want and has the craft to get there, that extra step is excessive. it also adds uncertainty that ✨vibes✨ are inevitably packaged with. you end up correcting the output towards what you would have made anyway, except now you're fighting the tool's interpretation of your idea rather than just building the thing.
the "do it with me" approach - AI that lives inside ableton, that understands your session, that can handle the repetitive scaffolding whilst leaving the decisions and agency to you - that's an entirely different proposition. the routine tasks that eat studio time aren't the creative ones. that AI removes friction, not your judgement.
that's why the anthropic bet hits home for me specifically. it's designed for people who already have an established workflow and just want to move faster, not for people who need the workflow replaced by ✨vibes✨.
the caveat \\ right now
a fair warning: i am talking about the current situation. both approaches are heavily constrained by what the models can actually do. "generate for me" produces outputs that most professionals will recognise immediately as AI. "do it with me" is only useful if the embedded AI understands the domain well enough to be trusted.
a year or two from now, when the models are substantially better, these positions may look quite different. the bet you're making today is on distribution and workflow integration - and you're hoping model quality follows. i'm pretty sure that ultimately both bets will converge somewhere in the middle.
nine connectors is a statement from anthropic
and the statement is: we will be the intelligence layer that lives inside the familiar tools you already use daily, and we're moving before anyone notices that layer matters more than the model sitting on top of it.
for working professionals, that bet is won on arrival. we're not going to let ✨vibes✨ drive our careers. but we appreciate a smart tool that fits our existing workflow and helps us get sh*t done. whether the models get good enough to make it matter in practice is the only question left.
everything else was quite deliberate.
