Forget legal power ... HOW? What technical mechanism does the FCC have that can block firmware updates over the public internet? Are we proposing a China-style "Great Firewall" for U.S. consumers?

I've been a rider for about 15 years. The absence of shifting is one of the things which makes EVs significantly less fun (in both cars and bikes/scooters). Even video games and movies recognize this in how they implement futuristic EVs.

The clutch on a bike is also more important than the clutch on a car, and it's a big part of the feel of a bike. Motorcycle clutches are 'wet', you can be half-on and half-off clutch. This is useful for helping control against engine torque to the wheels, 'engine braking' as well as controlling launch. For anyone accustomed to riding, it's a necessary feature, because it's literally how motorcycles work. Remove it and it doesn't feel like the same thing; it removes a lot of the enjoyment and tactility of the activity, and subsequently the enjoyment, of controlling a machine. It feels like you're doing something (and you are).

EVs feel more like a railcar, it's the exact opposite of the freedom of movement that motorcycles give you.

On the contrary, evidence seems to suggest that AI has long been very good at generating zero-day vulnerabilities. It has a little more trouble with identifying and avoiding them.

There's not an agency in the US government at this point which doesn't drastically manipulate its data to fit either internal politics, or political party agendas.

The problem at AWS is that they largely don't have 'core competencies' anymore, and haven't realized it yet.

They used to be a company which embraced new ways of doing things and doing small, agile things quickly. That hasn't been the case for half a decade now - in part due to cultural changes pushed from the top, but largely hasn't been the case for a while.

You'd think a cloud company with a fully distributed global infrastructure would have been one of the forefront proponents of remote work, and they did lean in on that a little bit at first, but quickly reversed course - in part due to the kinds of people they'd started hiring in excess not working. Those people are predominantly NOT the traditional hard charging, results-oriented people they used to hire, and are instead people who seem to prefer meeting over doing.

His own sister has made rape and sexual assault accusations against Altman, which supposedly went on for decades. He's "questionable" at best with regard to the murder of one of his prior coworkers/employees who was going to blow the whistle. I'm not sure what Musk has done comparable.

Do you think that might be politically advantageous of him, perhaps as payback for unthroning his wife, and that he realizes that the chance of them getting released is zero despite anything he might say?

Neither is a particularly endearing character, but of the 2, Altman gives me more of the "serial killer who'd poison your entire family to get what he wants, turn your back and he'll take your wallet" type of evil, and Musk more the "we're going to leave everyone on earth to die, want to have my child?" kind of evil. I'm not sure which is worse ... but in this case, Musk clearly has a better argument against OpenAI than OpenAI has in its defense - merely from the 'nonprofit to profit' basis, ignoring the intent and other aspects about safety.... which is certainly concerning in its own right.

OT, but... the way these big shops seem to be doing 'safety' is an idiotic bolt-on approach, from what I can see: it's all after-the-fact. If you give your agent/model a foundational prompt like "You're a helpful bot", that colors everything it does after that. They need to have a moral/foundations layer which does the same thing, perhaps even trained on its own very insular dataset that's been curated to meet objectives that can help it rank the value of different data. That way Reddit or some postgrad's humanities paperwork doesn't get the same criterial evaluation as, say, Socrates, Einstein, or the Bible. Maybe they do that - but it certainly doesn't seem like it, based on how easy it is to get them to disobey "safety" guidelines.

That is literally ages in 2026 internet time. The world itself moves much faster. Entire nations have been toppled in less than a day, in the past year. A week is an eternity to wait to mention something like this, particularly when it's literally just a repost of something that was posted a week ago.

This is literally the third /. mention of this in a very short period of time, nevermind the fact that it's been broadcast literally everywhere and is the biggest security vuln found since sliced bread (or heartbleed). It's been fixed and available for "ages" now on every major distro.

Precisely why I jumped to local inference, and spent about $20/mo checking the work and planning on cloud models. You don't need much hardware to effectively replace claude.

Sir, are you aware of how thrust is generated? Compression.

I'm getting about 35 tokens/s on a 3060ti 12GB with Qwen 3.6 MoE Q4, KV 8/8, and would get about 50-60 with KV 4/4. That's with 128K context.

I looked it up. Asus's fiscal year is January through December (same as the calendar year).

They got bought by EMC, which at the time was a Dell company. Then they got acquired by Dell directly. Then they got spun off as their own company, which lasted a year or two before Broadcom snapped them up. Through the whole ordeal, they were sustained mainly by a combination of legit vendor lock-in and people just drinking the Kool-Aid.