Currently downloading a 130gb game* wondering what the fuck could possibly be in that data
*Not the one you're thinking of. Or the other one you're thinking of
@luked.bsky.social
Head of AI for Take Two Interactive. Doctor of Game AI (from before the techbros moved in I swear). Dog dad, pitmaster, D&D-holic. ๐ด๓ ง๓ ข๓ ณ๓ ฃ๓ ด๓ ฟ in ๐บ๐ธ
Currently downloading a 130gb game* wondering what the fuck could possibly be in that data
*Not the one you're thinking of. Or the other one you're thinking of
Love "research". You can justify a lot when it's a write-off
11.08.2025 21:22 โ ๐ 3 ๐ 0 ๐ฌ 1 ๐ 0Looking forward to seeing them!
11.08.2025 19:13 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0This also applies to image generators
11.08.2025 14:22 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0This right here
10.08.2025 05:20 โ ๐ 5 ๐ 1 ๐ฌ 0 ๐ 0Hah came to make a similar crack. "authentic experience" ๐
10.08.2025 00:08 โ ๐ 2 ๐ 0 ๐ฌ 0 ๐ 0part of my resentment towards LLMs is to do with them crowding out anything else we could be talking about. there are so many things I could be reading to get better at my craft but I'm presented with "you should actually get worse at it on purpose because nothing matters any more"
08.08.2025 19:27 โ ๐ 638 ๐ 126 ๐ฌ 7 ๐ 21Which notionally this should be good at: Read the info I'm feeding in. Rewrite it a specific way. LLM table stakes you'd think
I got it to work once after extensive correcting and retrying. The n8n workflow agentization of it remains elusive
The use case was to take a photo of one of my many recipe books' pages and have it parse out all the info, formulate a json payload and fire it against my recipe server's API so it became part of the database
09.08.2025 03:39 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0Today I came up with an actual genuine authentic AI workflow idea that I tried to build in n8n.
It did not go well.
Starting to suspect this is all bullshit...
How very dare anyone try to hold these companies to account for pillaging the entirety of human civilization in order to make machines bad at math.
Whatever will the bad at math industry do???
Oh no! Anyway...
Source: Ars Technica share.google/phLXJFvqeEDE...
Happy birthday bud!
09.08.2025 00:08 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0One side says that LLMs lack knowledge of the world because theyโre extensively trained on text and have no model of the real world, the other side says LLMs have PhD-level intelligence. Guys, guysโyouโre both right
LOL Musgrave
08.08.2025 17:06 โ ๐ 767 ๐ 121 ๐ฌ 8 ๐ 3This one hits close. Romero hurt but John and Brenda are ethereal elders, it's... Distant? Removed? Bithell is my era's boy-made-good story.
If none of these titans can make it work, we are all so so so cooked.
In the wake of the announcement regarding redundancies at Bithell Games please go and follow these wonderful, talented people go.bsky.app/U1R3UmV
08.08.2025 16:18 โ ๐ 110 ๐ 63 ๐ฌ 1 ๐ 1bsky.app/profile/mike...
08.08.2025 16:11 โ ๐ 133 ๐ 11 ๐ฌ 2 ๐ 0awww mate I'm so sorry. Love to you and the team โค๏ธ
08.08.2025 16:05 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0Very possible but I feel like that's being very generous to the capabilities of ChatGPT. Being a sniveling weasel of a scumbag lying with data visualizations is a very human thing
08.08.2025 04:15 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0Had a major deadline for today. Shipped at 11.57 EST
Grad student in me shouted "STILL GOT IT!" ๐
Well I guess we get to steal this without attribution then cos it's too good of a thought to let disappear
07.08.2025 23:49 โ ๐ 22 ๐ 0 ๐ฌ 1 ๐ 0That's a 9 year old going places. Smarter than these "10x engineer" assholes apparently
07.08.2025 23:45 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0I have half an idea it was you or @jfg.land but someone recently posted that they're just gonna make disagreeing with genAI output discrimination and that'll be that
07.08.2025 22:46 โ ๐ 2 ๐ 0 ๐ฌ 0 ๐ 0Forever is good
07.08.2025 21:26 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0Mike doing the lord's work so we all don't have to.
07.08.2025 18:48 โ ๐ 21 ๐ 6 ๐ฌ 1 ๐ 069.1 and 30.8 are the same
52.8 is bigger 69.1
74.9 is almost 3x the size of 69.1
They're working real hard to obfuscate what a tragically small incremental step this new thing is
But hey don't worry. They raised another $9bn on $500bn valuation
Surely no economic fallout when this bubble pops!
A graph taken from OpenAI's GPT5 launch that fully misrepresents data
It's absolutely unreal that we live in a timeline where an alleged major company can put up a graph like this as part of their new product launch.
What are we even doing anymore???
Spent the morning trying to design a new hiring process that can overcome all the various ways folks are starting to weaponize AI to game interviews - particularly for a remote-first team.
But tell me again how great GenAI is.
Fuckin hell dude
06.08.2025 22:14 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0I'd probably go halfsies on that
06.08.2025 02:06 โ ๐ 2 ๐ 0 ๐ฌ 0 ๐ 0