I was wondering what other services could be glued onto the deal.
Wondering about this I had several horrifying thoughts. Somewhat palletable seems to add a hosting account with a subdomain so that one can share conversations and other ai creations, upload other things and use them for future reference. That way visitors can adjust to the content being machine generated.
If it's costly to run, maybe it's reasonable to expect that persisting settings, which is also costly, has some level of gatekeeping? I'm as ambivalent about LLMs as the next guy, but these are frankly nonsense concerns.
The ‘costly’ bit is referring to when asking for a SELECT statement instead of just receiving the SELECT statement I am given pages of overly verbose flowery text, reminders, and custom generated flow chart diagrams.
My understanding is most of these things charge per token so I am assuming generating more tokens incurs more costs, and I’m saying it’s generating needless tokens that frustrate me, so cui bono these additional costs?
Also more token generation means more energy resources consumed, so it’s burning the planet to frustrate someone it is trying to convince to become a paying customer.
How are these nonsense concerns?
It is unnecessary to persist any settings if I can just click a “just answer the question” checkbox before clicking send on the prompt.
Hey, I am free of any delusions in regards how I interact with the tool in question, and any expectations on returns therein.
The ceo asked for feedback and I provided it.
They can ignore me, I’m fine with that.
The reality of the situation is that they allow you to use the service without an account, and it is the only way I will ever use any of these llm services.
If they want to clarify that they only want feedback from paid or signed up users then I would gladly withhold my feedback.
As for the feature in question, I suggested a check box, so I’m unsure where you are getting mind reading. There are a number of other such checkboxes when prompting.
As for the feature itself, it would seem to save them server costs to implement it, and just in general, I think it’s bad form to hide QOL features to encourage sign up, but that is their choice as the service provider.
(A similar annoying QOL feature ransom is YouTube refusing to do PiP unless you’re logged in. Twitch on the other hand allows PiP without a login.)
I think their “here’s our worse model for free, pay for better ones” is a fine monetization strategy, and is a fine compromise for getting people like me to use the service at all. I mean they are still clearly collecting and using my queries for training and feature expansion.
(Even better is their idea to allow a few uses for a better model for free that resets in some time frame to let me see what I’m missing without using those better models, unfortunately, as I mentioned earlier, that is buggy)
Hiding QOL features seems like a foot gun.
They hardly tilt the scales on cost, like a better model would, and make people who you have yet to convince to pay think even your paid service is worse than it actually is, making them less likely to want to pay in.
For instance, apparently the feature I requested is already implemented, and as an accountless user I was unaware that it was even an option had I paid or signed up.
Does YouTube consider PiP an account generating feature? For me, it pushed me away from their service.
The only reason I used invidious was so I could listen to YouTube audio with my phone screen off while I went for a run.
Provide for free the features that would make people want to use your thing, and charge them for the features that cost more to serve.
I am uninterested in creating an account to use an llm service, and doubly uninterested in giving one my email address.
These are separate concerns.
For instance, HN allows account creation without providing an email.