the movie oversimplifies a lot from the books. similar to how the movie depicts a heighliner as a tunnel between two worlds (which is also wrong). FTL travel is not impossible without spice, just unreliable.
specifically, "Without melange [aka spice] or some other means of projecting a heighliner's course, every translight Guild voyage risks disaster."
additionally, before the ban on AI (butlerian jihad), they would have had AI perform this operation for them and likely discovered arrakis before the jihad
Interesting take. I don’t think Apple is using foundation models or private cloud for coding (could be wrong here though). But I was not aware that Apple was moving to Claude as a means of data protection.
The point of my post is to illistrate how high api prices are. The idea that a consumer can pay $200/mo and spend more in a day than an apple engineer is funny to me (Apple probably gets discounts and whatnot, but as I pointed out elsewhere these likely are nowhere near the discounts from subs). I do want to say that 300 dollars a day is still a lot of money even in the US. Cost of living is higher and people make a lot more, but not so much more that $300/day is a trivial amount. it’s just not a lot when you compare it to the amount of api spent by someone with multiple subs.
@Philfreeze96@CarlosAurelius@zenoware A lot of small companies/startups do, but yea as someone else pointed out you don’t get ZDR
If you turn on all the privacy settings though it’s not that much more data shared but don’t take my word for this
@dantosinfernos@zenoware This is a good point and I would agree with you broadly
Apple seems to favor Claude code, though from what I can tell from job postings and other info, but hard to say tbh. And as someone else pointed out, companies get discounts
Yea no it’s good feedback, I just assumed most of these were just made through the normal chat interface without any special rules against python
I think it’s a spectrum though, and you obviously have one side where the agent doesn’t get to iterate against its output to the other where you just use a coding harness and let it find the best approach/web search/use whatever tools
@bedirhandd0@AdamHoltererer Ok, fair. I take your point. Retried without python (and removed the subagent prompt since it wasn't doing anything anyways)
@basedpolymer@AdamHoltererer Fair, but I think also the takeaway from this post shouldn't be "agents can't recreate a good approximation their own logo" which is the conclusion most ppl will probably come away with
@bedirhandd0@AdamHoltererer I still consider this one shot, no /goal or /loop or subagents involved. the agent didn't end up using subagents (thay aren't availible in chat mode)
@AdamHoltererer this is my second attempt tbf, but you need to set models up for success and give them a good prompt. this was in chatgpt (chat mode) on pro.
more info:
@AdamHoltererer this is my second attempt tbf, but you need to set models up for success and give them a good prompt. this was in chatgpt (chat mode) on pro.
more info:
Leto knew that Arrakis was a trap, and he assumed that the Harkonnens had a spy within his midst. However, he did not know who the spy was.
Yueh in particular underwent imperial conditioning, supposedly making it impossible for him to defect. He also attempts to hide his betrayal (this isn't really explained in the movies though).
Leto did not (and would not) have let it happen ofc, but he did not and could not predict the Harkonnens converting Yueh