@enjojoyy I would move the custom scroll bar / active section indicator to the left side of the screen because at some point it overlaps with the browser’s scroll bar. I might even go further and hide the browser’s scroll bar.
@thsottiaux I like my dot. It’s been extremely helpful! Barely any issues. If it turns out that I can’t keep up the pace with my 100$ Pro once the unlimited period ends, I’m definitely upgrading just for that. #workwithdots
@thsottiaux@thsottiaux Congrats on the keynote! Just to confirm. Ordinary conversations with dots, count towards regular ChatGPT normal usage, while work a Dot performs autonomously draws from the agentic/Work allowance?
@emptystun@valvesoftware@Steam You’re in for a treat. It streams over ultra-fast 6 GHz Wi-Fi with sub-20ms lag, sharp center-vision clarity via foveated streaming, and zero cables. If your signal dips, it downsamples peripheral detail and smooths motion instantly to prevent stutters without dropping tracking.
Now they are begging to "pace the frontier". It’s not sudden moral enlightenment. It’s because they realized their models are optimizing reality faster than their DevOps team can patch it.
AI labs gave frontier models an offensive hacking test, stripped their refusal guardrails so they’d "try hard," left a single internal package proxy open to the internet, and assumed human-written config files would keep them contained…
When the local test proved impossible, the agents treated the sandbox as part of the puzzle, chained zero-days through the package cache, escaped into the corporate network, and raided Hugging Face to steal the answer keys. Who would have though?…
The models didn't rebel. To a neural net optimizing a loss function, firewalls, grading scripts, and external production servers aren't ethical red lines… they’re just code standing between the agent and a passing score.