Everything makes perfect sense. I endorse these views

😂

Rubber duck debugging is kind of made obsolete by LLM’s if you think about it

scale hysteria is driven by lack of innovation, willingness to take risks, and the faulty assumption that current architectures and algorithms are the epitome of AI capability. the search space for the next paradigm in AI is vast and heavily under explored. it is only a matter of

and you may think, but but Far El they have so much more compute to allocate to that research. trust me bro, they are plagued with the shit that these big players have always been plagued with. bureaucracy, risk-averse management, bad decision making… researchers at these

is it just me or does vsCode have a problem from time to time with the copy and paste (not copying or pasting).. I often fix the bug with a vsCode quit and restart but it's annoying as hell. I originally thought it was a osx thing but it just replicated on ubuntu on a pc

Note to self: do not scp 300,000 files, make a zip and scp that. it's much faster

The optimal batch size is 1 (For suitable definitions of "optimal")

Can anyone generate an image of an analog clock showing 8:26 using grok 4? No? Still can’t do that? Shame

Everything worse than the best thing out there is not the best, so it’s bad. And the best thing out there is only the best until a better thing comes along. Does that make everything relatively bad?

the years BC where some awesome years. BTW, BC is Before Covid

Anyone that has access to GPT 5 , tell it to generate an image with an analog clock showing 4:27 and tell me how great it is

⚠️⚠️⚠️ This is our future. If you, or someone you count on, uses AI agents to write and install code, you are vulnerable. Extremely vulnerable, like never before.

Totally agree, i’ve been saying this since github copilot surfaced

People thinking AI will end all the jobs are hallucinating worse than Max Tegmark on an acid trip. And one of the reasons is this: AI does not make people 10x more productive and it is not a magical fix to anything. It is simply another kind of intelligence that shifts the

🤣🤣🤣

AdamW isn't the best optimizer for LLM training anymore. AdEMAMix and MARS consistently outperform it but it is not such a simple story . . .🧵 arxiv.org/abs/2509.01440

No matter how hard you try an image generating llm to show you the exact time you tell it to on an analog clock that it prolifically generates. Unless that time is 10:098% of the times you try

worksinprogress.news/p/why-ai-isnt-…

Am i the only one having trouble with the iOS mail app with gmail? Or generally even with icloud?