English
V
85 posts
@adcock_brett @Figure_robot Why are some faster? Luck with packages?
English
@endocodigo @OpenAI There's a community port: github.com/ilysenko/codex…
I've set it up recently with NixOS. Didn't work much with it yet on NixOS, but seems like a promising bridge until OpenAI finally releases the desktop app officially for Linux. CLI is already available natively on Linux.
English
@OpenAI Guys.. this is great. But, for linux users... HOW? We have no desktop app ... ate least make it work for codex cli?
English
@sundarpichai So you telling me Gemini the best coding model rn .....I don't believe this but it would be better if Google creates something like codex to deeply integrate in IDEs because anti gravity is so confusing ngl the setup etc I tried and eventually moved to codex
English
Workhorse model! (and hope you're enjoying your first I/O)
Chubby♨️@kimmonismus
Insane evals for a Flash model! Gemini 3.5 Flash is really good for its size!
English
@FiftyOne_50_ @kimmonismus Fair point, yet my point was the irony.
The consequences were clear early on, that's why big tech is massively investing into the infrastructure. Otherwise agree, those aspects exist and will continue to exist.
I'd say the burden is relatively small and necessary.
This is the way
English
@v_avuso @kimmonismus Peak X: seeing an argument you cannot answer, declaring the phrasing “AI-sounding,” and mistaking that for a rebuttal.
The infrastructure still exists.
The resource burden still exists.
The consequence still exists.
Your vibe check is not an argument.🛠️
English
Energy, not compute, may become the real bottleneck for AI.
The proposed Stratos data center in Utah could consume up to 9 GW of power at full buildout, making it one of the largest data center projects in the world.
That is roughly comparable to New York City’s average electricity demand - or the output of about nine nuclear power plants - for a single data center.
English
@FiftyOne_50_ @kimmonismus Peak X: complaining about AI while letting AI write/draft tweets, with the most obvious AI-sounding phrases. 🙌
English
@kimmonismus AI was sold as weightless intelligence.
Now it wants cities of power, oceans of cooling, and nuclear-scale infrastructure so a few firms can rent cognition back to everyone else.
The bottleneck was never just compute.
It was consequence.
Also known as: reality. 🛠️
English
@rwang07 Makes sense they'd use the best hardware & models for themselves ➜ massive speedup & qualitative improvement in their own AI research.
Hardware ➜ limited & uneconomical.
Bleeding edge models ➜ rough edges, insufficient safety for mass adoption, uneconomical, competitive edge.
English
Cerebras CFO: "We serve all models, and there is no limit to the size of the models that we can serve. Today, we're serving trillion parameter models. We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them."
Deirdre Bosa@dee_bosa
"One of the biggest misconceptions" Cerebras CFO @BobKomin pushes back on the small-models narrative. "We serve all models, and there is no limit to the size of the models that we can serve. Today, we're serving trillion parameter models. We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them."
English
@kimmonismus Tokenmaxxing isn't the way, bad proxy for code quality and real throughput.
This usage might be legit, I'm not judging it.
Just saying that by being efficient, token usage can be reduced strongly. Deep pockets can enable brute forcing. But token count ≠ skill or productivity.
English
This is what you’re competing with.
$1.3 million in tokens in 30 days.
603b tokens in sum.
Burn more tokens or you’re not gonna make it.
Peter Steinberger 🦞@steipete
The latest CodexBar update renders API costs wayyyy nicer. codex.bar
English
@icanvardar Would be cool if this level of competency could be achieved in 6-12 months with 200 tokens per second with a 35b3a model, and another 6-12 months later with well affordable ASICs at 2k t/s.
English
gpt 5.5 is already enough for most programming work the bottleneck is no longer the model
English
@XFreeze Why though? Small sphere of computronium could power trillions of civilizations for eons. Burry the sphere in an inconspicuous, uninhabitable planet, add a reactor /w fuel for an eternity, & live in peace+safety. Way more energy efficient, low risk, solution to the fermi paradox.
English
@flowersslop "In The Hitchhiker's Guide to the Galaxy, a supercomputer named Deep Thought calculates 42 as the ultimate answer after 7.5 million years."
-> the problem with long term time horizons is the risk of misalignment with the task, because specs are typically lossy.
English
@0xigat0r_Man @Teknium Disagree. Better: support devs with contributions to their open source projects or donations, if available. These tokens try to solve a problem that isn't there, while paving a highway for scam.
English
@paraddox @thegenioo There's the community port though. Didn't test it yet, but sounds promising: github.com/ilysenko/codex…
English
@misraetel Cool demo, but the nod must be pure theater: with a wireless module, they could probably transmit 1gbps, instead of using what feels like 1-2 seconds for a nod.
English
@yellowLit78678 @rudyonton_ @LuaSyntax0x I'm very skeptical if someone with such an attitude could be right. Plus, he doesn't bring any good argument. Are you all bots trying to defend a crypto scam?
English
@kimmonismus Doesn't make any sense to me. EUDI wallet is supposed to become available next year. If it's about child protection, require age verification with the wallet. Could surely also be done anonymously, similar as World app attempts to. Required for future infra anyway. Just do it.
English
This worries me a lot. VPNs are an important gateway to the open internet.
To now supposedly raise concerns about child protection seems like a pretext.
European Parliamentary Research Service@EP_EPRS
Virtual private networks #VPN are increasingly used to bypass online age verification. Protecting children online is a priority, with new rules being implemented requiring a minimum age for access to some services Read👉 link.europa.eu/FGfr6C #DSA @EP_Justice @FZarzalejos
English
