Muse can submit a request to be whitelisted for beta features like calling.

Jev outperforms clef and OpenAI decisions on most of my tasks, outside of when I need multi modal image going into it. Jev is also cheaper but costs are minimal overall.

I just dont understand why anyone needs a general purpose classifier. Just build a classifier for your specific use cases.

Fortunately Jev is cheap, so I dont think it matters too much, but I think its robbing people of the opportunity to learn and implement this themselves.

Also, I dont really want 3 companies responsible for censorship/classification.

That's why it's useful, it's for when you don't know what needs to be classified or if you want to offer this classification models to your users.

Because training your own model requires substantial effort around data annotations, and then dealing with data drift etc in production.

It’s not training that’s the hard part, it’s when you don’t know your use case or it keeps changing all the time

Hm I had built out a system that tested a few hundred decisions and clef beat Jev pretty handily, I think there was actually 0 instances where clef regressed and it improved about 20% of the false decisions.

I had the opposite experience. Clef was worse in accuracy, and an order of magnitude slower.

That's why it's a nonstarter for me.

The flip side is that it helps people find items they want, leading to less friction and buying more. Not that I trust it to be to the benefit of the consumer by default, but we've definitely seen better conversion and AOV increases after moving from straight lexical search to hybrid vector.

Does this not cause you to lose cache because the effort is in the system prompt, so you will burn your usage faster dynamically changing effort in the same conversation?

I use my Vision Pro like 5x a week, it's rough around the edges but I figured it's niche enough that it wouldn't get a meaningful upgrade and be obsolete for 3+ years - seems to have played off well. If you travel a lot it's also amazing.

I used the Rewind app on Mac and it was very nice to have the ability to search almost anything you did / saw on the computer. If it's local, opt in, and secure then it's potentially very worth exploring.

thats the point - local, opt-in, nothing leaves the machine. rewind was great but the cloud part scared people off. this runs the whole analysis on-device so the search/chat works without anything being uploaded and it works more as a background service.

If you're vibing such complex things you should probably be in the habit of also generating detailed documentation and commits so the ai can follow breadcrumbs, add some playbooks for how to debug and it's actually pretty good. Too complex for local models context though - so you're probably still correct albeit there are ways to mitigate or delay this.

>there are ways to mitigate or delay this.

Yeah, like writing the code yourself!

Conductor has a feature that lets you define a script to copy .env and other things as needed from main repo into new worktrees on creation.

I don't use it much anymore, but last I did I would run into port conflicts with docker projects.

Most of the time it's when I want to go back and have a skill made for future reuse, but with remote control I've had some sessions open for remote diagnostics and it just works better than starting from scratch - even having lessons learned to create memories and update Claude.md.

I know it's wasteful but often I've got a surplus of tokens and not enough of my time - so it's a trade off I've been fine with.