@hamptonism A fair comparison is between Nvidia GPUs and ARM GPU or Google TPUs.
I’ve compared image model inferencing on Nvidia GPU with Apple M series GPUs and found it to be 100x faster, hence the comparison here.
@hamptonism Any other semiconductor company CUDA beaten Nvidia’s moat, but they didn’t, not even Apple Silicon.
Why will a new semiconductor company succeed?
@vip92226041 Any other semiconductor company CUDA beaten Nvidia’s moat, but they didn’t, not even Apple Silicon.
Why will a new semiconductor company succeed?
@hlopez_ Stop doing work.. Try delegating to Codex or Claude first and if they can’t do it, think about how to solve it better and try delegating again.
I'm signal-boosting this because Luke clearly knows what he's talking about. I think his diagnosis that Microsoft's running cloud services are only metastable and cannot handle the restart storm is almost certainly correct.
After a 11 hour Outlook Outage, I’m wondering if Microsoft tried asking any of the 60000 employees who have been given VEO about how the system works or they’re relying on AI’s sycophantic “Yes, you’re right” answers for system flow understanding and debugging
We're investigating reports of users experiencing issues accessing Exchange Online mailboxes through multiple connection methods in the Americas. More details can be found in the Microsoft 365 admin center under EX1550543.
@thomasfbloom We evolved from Gen AI papers to Gen AI proofs in less than 3-4 years, and did you magically expect the proofs to be coherent like Einstein or Maxwell wrote them?
If you’re still worried about token costs in Nov ’26, you’re probably not using GPT-6-Luna-XHIGH enough. It can be your default workhorse for almost anything you do on a computer.
Save the big guns Sol and Opus for greenfield projects and truly complex work.
@CAULDR0N_ Are Agents villagers chopping wood when they’re busy working? 🙂
Do buildings created by villagers represent the features or projects that the Agents are working on?
@theo It is insane to me that there are still devs not using ACP (https://t.co/H9GwRoRXb9) to connect to any agent from Zed/VSCode/IntelliJ, just because Dev Twitter tells them to use GUI vs CLI
@kmeanskaran +1 from my observations. Beyond a point, you can’t do better than a Coding agent like harness with m skills + n tools (either MCP or function tools)
The lesser instructions in SKILL.md the better as future models emerge.
Most of your time has to be spent on Evals
@ThePrimeagen@trq212 I have to agree with this.
Any of the below setups are 100x more fun than a vanilla Mac with basic window managers:
XPS + Omarchy + Hyprland
XPS + Ubuntu + Niri
Mac + https://t.co/b26x2XHSNH + Hyprland/Niri layouts
@trq212 OmniWM window manager lets you do this with App rules to categorize spaces manually and route apps to specific spaces (spaces are not same as Mac workspaces).
But using Jev to have a distraction free workspace experience on Mac is clever!
https://t.co/b26x2XHSNH
I don’t think I’ll be able to shut up about how good OmniWM window manager on Mac is.
It instantly lets you switch between Niri and Hyprland style layouts if you’re on the latest Mac OS Tahoe or Golden Gate version.
https://t.co/fQSI0tr7Ad
Distro hopping is fun. Add window manager hopping to that. Your options are vanilla Gnome or KDE, Hyprland, Sway, Nori.
I’ve made a small utility here that helps you go from 0 to 100 on any Distro by installing your favourite apps, tools etc instantly as you hop Distros. Try it here -
https://t.co/FZbUMEMvAv
@mattyp This is a welcome change but I’ll never combine my X account with AI coding or agent accounts. X has weird rules for blocking accounts - Unless you fix that nobody is going to want to merge their Cursor or GrokBot or Grok API accounts with X.
Because My software is an “Intelligence cache” of how to solve your problem and that of thousands/millions of other users who have similar problems.
Came across this concept of “Intelligence cache” on X and it kind of explains why everyone will not vibecode their own SaaS like Notion or Jira.
As an architect, I don’t think it’s just functional aspects that you’re paying $15/month for, you’re also paying for non-functional aspects like scalability, accessibility, security, auditability and a hundred other -ilities that all users will need.
https://t.co/lw4G5TJW3V