John Mell 🇺🇦🇦🇺
4.2K posts
John Mell 🇺🇦🇦🇺
@braincell__
founder with unlimited runway
Melbourne, VIC Katılım Haziran 2009
488 Takip Edilen139 Takipçiler
>codex running while you cook with your partner
>codex running while you push your kid on the swing
>codex running while you call your mom
>codex working on the thing you dreamed about for years but never had time for
>codex doing this while you hang with your loved ones
English
@aidan_mclau @viemccoy hey twitter algo, can you show me more of aidans tweets from now on? maybe like a 1.3 multiplier to your metric? sorta? ballpark?
English
@viemccoy we need to put you in charge of openai naming
English
@Hamidk_12 @tszzl you have a budget? i just use nvidia nemo free
English
@GaryMarcus yeah man, I agree with you, but they're taking the "derivative" and running with it! it seems to be working OK ... for the time being. honestly, they can just keep patching it and it will start to resemble a faux-agi. to coin a phrase. it just works.
English
almost all LLM “thought” is derivative.
below just sounds like hype to me
Jon Hernandez@JonhernandezIA
📁 Jensen Huang, CEO of NVIDIA, says GPT was never just about generating images or text but about generating thought itself. The real shift is that AI is no longer becoming a tool you query occasionally but a system that reasons continuously. Once computers stop waiting for commands and start thinking persistently, every layer of computing and human work gets rebuilt.
English
@destiny82728 its true i live in australia its terrifying dont come here
English
@DanielleFong I think this is kind of what @GaryMarcus talks about. They are "rendering" an increasingly plausible high resolution picture of AGI, with all the passes, but there will always be holes, where those "intelligence pixels" don't quite gel together.
English
what is wrong with opus 4.7 really
trillion dollar company. doesn't know how to search itself without prompting.
something is wayy over tuned in training sometime, and undertuned on stuff you rely on tool calls to load in awareness off
English
John Mell 🇺🇦🇦🇺 retweetledi
wading through bots and LLM-written replies here is getting more tedious by the day.
retweet if you agree.
English
@tekbog everyone was just having fun, prancing along all happy, and then
English
if after the bun rust rewrite and 1M code merge in a few days you still think that software engineering hasn’t changed massively idk what to tell you
English
Maybe 10 years from now we'll be saying it wasn't an AI bubble, it was just nature taking its course?? The sand talks to us... listen to the sand! ....listen for the wind that carries it
English
@Bizzle77C @JoyfullCat Mine did it when there was an annoying fly on the window
English
@max_spero_ @solipsnitsyn You do understand that you can have your own pro model by using 5.5 thinking with OpenCode and instruct it to browse the web? Pro is obsolete. Try opencode. It can generate research and reports for you with even cheaper models.
English
@solipsnitsyn Is 5.5 pro available on the 20x sub? I can’t even figure out how to buy the 20x sub for my team
English
John Mell 🇺🇦🇦🇺 retweetledi
@ultranarrative @sama I am a c++ guy. I did similar things like you 20 years ago and I feel silly and dumb for it. I hope you had fun. :)
English
i would like to talk to people who have built amazing things with 5.5 that weren't possible with earlier models. i am especially interested in examples that took ludicrous token budgets. thanks.
English
@giles That's why in C++ if you see a reinterpret_cast, you call a team meeting right away. A straight C cast, you lose your job.
English
John Mell 🇺🇦🇦🇺 retweetledi
@aidan_mclau @pigeon__s My theory is a lot of Claude character comes from allowing specific creative writing books to exert more weight during training. You have to have taste for that.
English
