Hacker Newsnew | past | comments | ask | show | jobs | submit | vitaflo's commentslogin

If you read the TL thread on this, the bot is technically running at 240 APM, but its commands get split off into separate commands for each unit, so the APM looks higher.

I've watched several replays on this. It using map hacks is lame but it also doesn't always take advantage of them. It sometimes does respect its own fog of war. It mostly wins with incredible micro (kinda has to, it's macro kinda sucks).

Most of its few losses come from drops (it never makes turrets and doesn't know how to handle them), bad macro (blocking its own ramp) or just incredibly unconventional play from its opponent (which is certainly not in its training set).


Specifically, it can micro across multiple screens of engagement in a way that we can't as easily.

StarCraft is just not a good benchmark exactly for these reasons: Endless complaining about the bots not being limited in the way humans are.

"Oh it's able to issue commands too fast." "Oh no, I give it full map access and it uses that."

Ego shooters are, naturally, also not good benchmarks.

OpenAI won DotA2 in 2019, a way better game.


It’s not “endless complaining” when the bot is functionally cheating. Yes if you let a computer use special mechanisms it will operate faster than a person who is using a mouse and a keyboard. A computer computers better than a person, what a revelation.

I don’t applaud people using aim assist in CS for the same reason. We literally don’t have access to the same tools. That’s why chess has been such a great target for years. One’s ability to manipulate the pieces has no bearing.


See, you're proving my point. To play StarCraft, you have to play like a computer, which humans suck at.

That's exactly why it's a bad game for human/AI competition.


dota wouldn't exist without Aeon of Strife (the UMS map from Starcraft that was later reworked into WC3 dota)

The other side of this equation is the people who now pump out socks for pennies. You can get what you’re advocating for but it means software development becoming something closer to sweat shops.

There were numerous wiki's that were flooded with this stuff, it's just the German one that got the most traffic.

It's the same basic training data, and if these agents were spamming the internet looking for a host wiki they probably found several and when finding other agents on one of them, they most likely just congregated there because it would have a higher value than one where they were the only agent on the wiki.


I still don't get how do you do this undercover. There are a billion websites. How do the agents stumble upon the same obscure unused wikis?


And these old manuals make it trivial to build a midi app for your old synths. The sysex specs are usually so detailed that it makes it very easy for Claude to follow and whatever it comes up with usually "just works".

I've built apps for all my old synths, not just to save sysex patches but also for midi control. It's actually amazing some of the hidden features these old synths have that you didn't even really know about because Claude finds all the connections in some cryptic sysex addendum in the manual.


A person eating one single hamburger uses more water than an entire year of LLM use for the same person. Yet we are very concerned about water use in data centers, but nobody cares about McDonalds.


Respectfully, I would like to see a source or at least the math for that.


Here it is:

[The water footprint of a single hamburger is *significantly higher than* that of using a large language model (LLM) for a year. For example, the water used for one hamburger is primarily due to the irrigation of cattle feed crops, while data centers like xAI's Colossus 2 consume much less water overall.](https://duckduckgo.com/?q=A+person+eating+one+single+hamburg...)

States that it's significantly higher but the breakdown shows a 660gal vs 1000gal (Single Burger vs 1yr LLM Usage) water usage...

Item Water Footprint (Gallons)

Single Hamburger Approximately 660 gallons

Year of LLM Usage Approximately 1,000 gallons

Results also contain an important disclaimer: "Auto-generated based on listed sources. May contain inaccuracies."

You be the judge ;)


I have to assume this means that they are about to release the final version of v4 Pro and it uses more resources than the preview did


I have to assume this means that they are about to release the final version of v4 Pro and it uses more resources than the preview did.


Preact is 3kb. The size argument doesn’t hold weight.


The author's library is only 2.47kB.


Where I work we call these people “very expensive keyboards”.


Finance is utter garbage now. 80% of the screen is AI bullshit and my portfolio is like 10%. And the design of it all is just a wireframe and not easily scannable at all. I’ve just stopped using it since I can’t get back to Classic view anymore.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: