You can’t save an organization from its own demise. It’s almost impossible to save a single individual from their self-destructive tendencies. But saving a whole group of people aligned around self-destructive tendencies is definitely impossible.
Some people in US like to compare powerfull export of powerful AI technologies with export of nuclear weapons. Following this comparison DeepSeek now export local nuclear bomb factories you can run in your garage: github.com/deepseek-ai/DeepI…
I asked my OpenClaw to translate article to share with the friends, and it rejected the translation due to copyright issues. I used Grok Bot for the same task and it gives me translation. I'm less and less sure that marriage of OpenAI and OpenClaw is a good thing.
"...the job that was my favorite one (being a typewriter) and it leaves me to do the part that I suck at (being a thinking person)."
List of my previous rejections: unable to solve captcha to register me for my ticket. Any human assistant can do that. There are online tools to solve the captchas I can pay for. There are handoff flow in some agents (which I ported to OpenClaw and use it, but unable to get it to the master)
As more people in IT worry about the future of their careers, I think every day about how I can contribute more to making more IT jobs obsolete.
Got this in my mailbox today. Can you tell me is it legal in the UK or should I report to the authorities?
People who are talking about the bubble in AI compute and waiting for hardware prices to go down after it bursts: do you understand how computationally constrained we are?
Teslas have 8 cameras and not 20, not because the cameras are expensive (they’re not), but because processing video in real time requires a tremendous amount of onboard compute, and yes, that’s expensive.
Your robot vacuum is dumb because you can’t have an edge AI server at home.
Your AI assistant is dumb because its compute is constrained to fit the subscription economics.
We can do so many clever things with AI right now, and the only thing stopping us, apart from talent, is a lack of compute.
We should be making 100 times more wafers to meet demand.
I'm not a biggest fan of Meta but their execution around AI products and their public communications are top notch. Such a big difference to any previous major launch from them I observe for years. @alexandr_wang is cooking, and he is cooking damn hard.
Even Meta haters not mentioning Facebook stealing all the data this time, which was a major PR pain point for Meta for many many years.
@alexandr_wang please do something about 5 cookie banners during the Meta AI subscription process, like one banner after each action. Looks like a joke.
Grok sounds like a comic book villain when you mention it in casual conversation. That’s why I think Cursor is a much better brand for SpaceXAI. Just try saying "I asked Grok" out loud without sounding like a comic book villain.
"Muse" is chef’s kiss. Great poets and painters had muses. It’s a creator’s thing! "I asked my Muse to do X" sounds great in conversation, and the name sticks.
Unfortunately, you will be unable to buy a personal inference server to run opensource AI models locally due to shortages of memory, GPUs, and regulatory barriers. What you can buy is a device with an always on camera for your bedroom, where you can order something online with your voice. And you will be happy.






