your example with google isn't necessarily applicable now because they've shown a roadmap that can be done and squeezed down tightly between the "hey, we're good folks" to "you're our captive cattle, we can do whatever the fuck we want. there's nothing you can do, since all our competitors will be doing the exact same thing shortly"

search engine results do this all the time, reordering output by advertiser input. its a pretty small jump from that to rewriting output from models, and even better where its all a black box.

>Cortex, by default, can set a flag to trigger unsandboxed command execution. The prompt injection manipulates the model to set the flag, allowing the malicious command to execute unsandboxed.

>This flag is intended to allow users to manually approve legitimate commands that require network access or access to files outside the sandbox.

>With the human-in-the-loop bypass from step 4, when the agent sets the flag to request execution outside the sandbox, the command immediately runs outside the sandbox, and the user is never prompted for consent.

scope restrictions are in place but are trivial to bypass

https://arstechnica.com/information-technology/2026/02/most-...

CloudBolt’s survey also examined how respondents are migrating workloads off of VMware. Currently, 36 percent of participants said they migrated 1–24 percent of their environment off of VMware. Another 32 percent said that they have migrated 25–49 percent; 10 percent said that they’ve migrated 50–74 percent of workloads; and 2 percent have migrated 75 percent or more of workloads. Five percent of respondents said that they have not migrated from VMware at all.

Among migrated workloads, 72 percent moved to public cloud infrastructure as a service, followed by Microsoft’s Hyper-V/Azure stack (43 percent of respondents).

Overall, 86 percent of respondents “are actively reducing their VMware footprint,” CloudBolt’s report said.

they're expecting to spend 10B burying lines in the mountains.

New england is pretty much one big rock garden/shelf where you're not digging through soil in alot of places but rock ledges.

So if you already have parametrized tests, you're already halfway there.

You'd have the same issue, if not more, with background noise, group settings and context acquisition

Processing input is the hard part, if you're already having issues, that isn't going to go away

As for others, one thing hearing people, particularly monolingual hearing people, don't understand very well is that hearing != understanding. Just because you hear a sound doesn't automatically equate to it having meaning. The default for many people is to just SPEAK LOUDER and slower, which does not help in the vast majority of encounters

Good example of enshittification

Other models are not considered insurable by GEICO as well. So likely small pool of policyholders + exorbiant claim payments == not worth the headache

When you have fender bender claims costing 20-40k USD to repair, how do you price that risk?

https://www.theautopian.com/heres-why-that-rivian-r1t-repair...