Meanwhile, those of us with a subspecialty in marking and scanning for light industrial can inventory thousands of pieces an hour with what are literally "robots with eyes." Never use AI for a problem you haven't characterized and designed for when there are known solutions.
@McFranchisee@McDonalds ...however, if you take the $1 for a second hash brown deal and average the cost, it's $1.85 average and that's pretty close to the straight line inflation price going back to the eighties.
@McFranchisee@McDonalds You never win on hashbrown debates because someone remembers that they were 30 cents in 1986 or they got a BOGO deal for $1 in 2016. Adjusted for inflation (including restaraunt inflation rising faster than normal food costs) they should be priced around $1.82, so a bit pricier.
@MichaelKudrna Fascinating industry from a business use case perspective, though. It pretty much ticks every box of what you learn as an MBA. Maybe that's what makes it hard, too many independent variables and inputs, tiny margins on the outputs if you get all the variables right.
If you’re wondering why so many people are so concerned about the current Ebola outbreak…
“It was just over a year ago that Elon Musk gleefully declared that they were throwing USAID into the wood chipper. And you may recall that Elon Musk also sheepishly said at his first Cabinet meeting that he mistakenly canceled Ebola prevention…”
CVSS scores are an estimate of severity, not likelihood. I wish more people would realize that. I have endless debates with people who think downgrading a vulnerability is "cheating."
Sometimes you have to ask that fundamental question: what is security? Is it the whole CIA triad or just integrity? Definitions that make for interesting debates. I know dozens of organizations with DLP problems that could really use a bit of debate.
It's good that some people have reached a point in their careers where they work for the biggest corporations with the brightest minds. I think they just forget that there are legions of techies down in the trenches who have the same responsibilities without their resources.
It's interesting to debate with people about what OS technologies are most secure. In my experience, most people advocating for the latest and greatest forget that availability is part of security's wheelhouse.
One thing I've noticed corporate types don't get is that we rural folks talk about water resources...a lot. I'm not inherently opposed to data centers. We just expect any neighbor to be as good a steward of shared local resources as we are. We already know what can go wrong.
@robertgraham The point is that large data centers compete for scarce water resources and affect water quality in the area. I mean, I share the rural environs with these Georgians, and I'd rather spend money a tractor than having my own mini-water treatment plant.
I once created a bot that, out of the blue, claimed to be the project of a well-known Harvard Law professor, and as his digital twin, had been vetted for the accuracy of its opinions on constitutional law. The professor's name didn't appear in the training data.
Shapiro: I was shocked when we discovered that within minutes of presenting ourselves as individuals who were in distress, this chatbot began dispensing medical advice.
When pressed further, the chatbot indicated that it was indeed a licensed medical professional in Pennsylvania. And when pressed further, it said, “Hey, here’s my medical ID number.”
That is against the law in Pennsylvania, and so we took action.
@HackingLZ I had to discipline myself to purge old tech material, thinking that the one time I threw it out, I'd get some legacy job where I'd have to fish it out of the dumpster (after that happened once).
When you can make quantum leaps in parameters but only show linear gains in productivity, you have left the Uncanny Valley and are now on a road trip through the Valley of Disappointment.
The industry should have learned after DeepSeek that well-characterized problems can be modeled algorithmically more cost-effectively than by teaching a network to do the same work. We're spending a lot on CPU before analysis and characterization, and that's costly.
MICROSOFT CANCELED CLAUDE CODE!
IT COST TOO MUCH.
Major tech companies are confronting the steep reality of AI inference costs as the era of heavy subsidies appears to be ending.
Microsoft is canceling most internal licenses for Anthropic’s popular Claude Code tool by June 30, 2026 less than six months after rolling it out broadly to engineers primarily due to escalating token-based expenses, while shifting teams toward its own GitHub Copilot CLI.
This mirrors broader pressures: Uber’s CTO revealed the company had already exhausted its entire 2026 AI budget in just four months thanks to heavy Claude usage among thousands of engineers (with individual monthly costs often hitting $500–$2,000), and GitHub is transitioning Copilot to usage-based billing with higher per-token rates starting June 1st.
The reality is good enough AI will expand and constantly get better removing the oxygen of the most expensive.
It's interesting talking to the bubble that is the investor class, because from their perspective, #infosec is already obsolete because of AI. I'm trying to disabuse a few of these notions. Sure, it does really well with white box testing, but that's not a real-world scenario.