@ahmedshubber25 The front of the blade should be designed to be brushed unpainted metal, so the intended aesthetic is not ruined after the first 5 seconds of use.
I want to help cuz I love Hermes, not trying to shit on it. But saying "every human on earth has used for 3 years" is what I mean by you rejecting the premise that it's not simple. But assuming you want some good faith feedback: Hermes is a lot more competent than ChatGPT, the paradigm doesn't hold. Normal people treat sessions as "google searches" it holds for that.
Part of the reason providing feedback is hard is that you are at the frontier, no one is doing it way better that I can point to, but when people who want to do what your app does are saying it feels complicated, it doesn't matter what you think, they are right.
You have two options, believe them or think they must be stupid, beacause I love hermes, I hope it's the former.
@Teknium@dshan@Jason You are asking how it's complicated to de-complicated but when you get an answer your response is "it's already simple".
I use Hermes, love it, but it's not simple.
Sit behind someone with low technical skills and watch them get stuck 10 seconds in.
@VictorTaelin Wouldn't this be much more valuable for AI companies than individuals? The way models use .md files, python scripts etc.. for internal work, they could use Bend2.
This is what I call a "divergent" problem. Where two features fight each other. The more X the less Y and vice versa. The way to tackle these is by either splitting the problem so they don't fight, adding adaptability into the system (fan, electromagnet...) or if you are lucky, turning it into a convergent problem where the more it cleans the better it drives :)
What you don't want is to solve it by calibration. aka finding the delicate balance where both work. Because the moment one variables changes (motor wear, surface magnetism change, tolerances...) it stops working.
Imo the kickstarter restrictions are very reasonable and increase the odds of a product being successful.
AI use is allowed but you must show that you actually have something. It's often enough to send them a handheld phone video showing what you have.
If this guy was unable to provide the evidence to clear that requirement I think the project being cancelled is for his own good.
You can vive-video, vive-code, vive-copyright but you can't vive-hardware (yet)
@ethanabuck DM me, I can help
@anothercohen Moving to Austin pushed me to figure out you can disable those alarms (at least on iPhone) as well as the amber alerts which are misused and 99% of the cases the perpetrator has the same last name as the victim.
I'm ok with the increased risk of drowning in a flash flood.
@VictorTaelin Try this: in a separate session, ask codex to create an auditor agent and evaluate the same results that you evaluated manually. Task it to be a cheat and BS detector. Donโt write special tests, just high level instructions.
Is codex able to find the hardcoded cheat?
Why are you trying to convince the model? Its structure leads to lying. It doesnโt have an โediting stageโ once it commits it has to go all in. What you do is create a separate isolated auditor that simply describes what the main model did. When thereโs mismatch between what you asked, you can automatically tell the main model to try again and why. It doesnโt have morals. All caps is meaningless, work with their structure.
@vanschneider@gentlycarved Whats the most intuitive ui examples youโve found that accommodates complex use cases? I keep coming back to node based when complexity increases