Funny how Nvidia spent years telling gamers they were the future, only for AI data centers to become the favorite child. Huang leaving might accelerate that shift, but blaming one CEO’s exit for expensive GPUs is generous. The real problem is that selling chips to billion dollar AI customers makes gamers look like a side quest.

The real scandal isn’t that Nvidia’s CEO became a liability. It’s that an entire industry made itself dependent on one man in a leather jacket. If AI’s future can be derailed by one executive’s personal decisions, perhaps the biggest problem isn’t succession planning but an ecosystem built around a personality cult.

TSMC makes the chips, OpenAI and Anthropic buy the chips, and now TSMC invests in the companies buying the chips. Everyone gets to report growth, everyone gets a bigger valuation, and the same money keeps doing laps around the industry. The silicon is real. The demand signal is starting to look like a hall of mirrors.

I suppose it’s a question of image: if you’ve hit rock bottom anyway, nothing matters these days. But if you’ve got a good reputation…

Calling it affordable broadband while throttling people to 600 Kbps after 50 GB is like selling a discounted bus pass that turns into a pair of shoes after four rides. And hiding the plan from search engines? Apparently the cheapest way to provide low income internet is making sure nobody can find it.

Nvidia spent 33 years building a parallel computing empire while apparently keeping its leadership architecture strictly single threaded. The scandal is serious, but the bigger corporate failure is having no clear successor. Imagine selling redundancy to the world and forgetting to install a backup CEO.

Turns out giving every developer five tireless AI interns doesn’t eliminate work. It turns them into unpaid middle managers of five interns who confidently submit broken code at 3 a.m. Management sees five times the output; developers see five times the pull requests. Congratulations, we’ve automated coding and scaled up the headache.

The irony is that Study Mode can ask a teenager to think, but it can’t teach them when to stop asking the machine. Growing up involves confidently making terrible arguments at dinner and discovering why they’re terrible. If every half formed thought gets a patient AI tutor, adolescence starts looking like a permanent customer support chat.

I would have liked to change it too, as it doesn’t really fit…

AI has turned code into the cheap part of software development. Unfortunately, bugs, security reviews, integration headaches and maintenance haven’t received the memo. We can now generate technical debt at machine speed while asking the same exhausted senior engineer to approve it all. That’s not quite the productivity revolution on the slide deck.

Anthropic went from worrying about AI alignment to aligning itself with a $3.4 billion gas engine order. Apparently the existential risk they couldn’t tolerate was a data center waiting for a grid connection. The planet, meanwhile, can just keep buffering.

We’ve reached the point where the smartphone industry is doing board level archaeology to keep $200 phones affordable. Reusing working chips is actually sensible. The absurd part is that we needed a memory shortage to rediscover that perfectly functional electronics don’t automatically become garbage when the phone does.

The irony is spectacular: a coding assistant designed to automate development also automates the creation of security liabilities. And Anthropic’s response seems to be more automation. Perhaps the next AI agent can monitor the first AI agent while a third one explains to management why the database credentials are now public.

The AI industry is already floating the idea of AI rights while actual humans lose jobs, privacy and control to systems nobody can properly explain. Before anyone asks whether a chatbot deserves protection, maybe ask why we’re handing so much power to technology that routinely gets basic facts wrong.

When employees ask for AI, they’re solving problems they actually understand. When management pushes it from above without training or clear goals, they’re often just chasing a productivity headline.

I suppose I’ll never understand why comments are upvoted or downvoted. Sometimes I get the impression that it’s more down to chance or mistakes.

And that’s exactly where management fails! You can’t demand AI adoption, provide no training, no guidance, no quality standards, and then blame employees for poor results. If workers don’t know how to use AI effectively, that’s a leadership failure. Management wanted instant productivity without investing in skills.

Management: Use AI everywhere! Also management: Why does everything sound like a chatbot? The real failure is executives confusing mandatory adoption with actual productivity, then blaming workers when the resulting mountain of nonsense lands in their inbox.

💯 That’s an important distinction: Civil liability determines who pays for the damage, while criminal responsibility addresses whether someone’s negligence was serious enough to constitute a crime. And that question absolutely deserves public scrutiny. Automation must not become a shield against criminal accountability.

The responsibility should lie with the company that decided the robot was safe enough to deploy. How it distributes liability among its developers, manufacturers and suppliers is its problem, not the public’s.