zug_zug

Well now that it's documented in writing, that business will have to be shut down (-20% revenue), somebody will have to be blamed for it (fired, arrested?), and who knows how big the fines will be.

According to these allegations (May):

1. 20% of Nvidia's FY2026 compute revenue was ultimately driven by Chinese demand through intermediaries and alleged illegal diversion.

2. This was driven by Alice Huang (the spy)

3. Crucially this report has allegations that Jensen KNEW that these chips were being sold illegally, and that he even went on-site to these sites ALONGSIDE ALIBABA representatives.

Nvidia is down 5.7% this morning (probably on this news), but is trading at a P/E of 23x still -- my mental math if it's expecting to lose 20% of its revenue AND a CEO that it has more to fall.

Seems counter-intuitive and definitely warrants research to understand the mechanism. The most unintuitive part being local plants not returning.

> My question is what Americans think economic growth in the 21st century looks like?

Maybe it looks like a reduction in monopolies, increased competition, lowering of prices as automation pays dividends, and a general de-consolidation of power.

It doesn't look like 4-companies owning the whole stock market and incumbents using AI to cut every worker they can with no longterm plan, using AI to deny insurance claims, and rig the deck in the most unconscionable uncompetitive ways they can.

There are outcomes much worse than "flat" if we aren't careful, and even the AI CEOs say as much.

It's getting harder to believe this is accidental when it's 3 leftist situations in a row in such a short time.

> We didn't "lose" it, we never had it.

I think this is a silly claim of a very complicated topic. Our ability to trust various institutions (e.g. the newspaper, government, banks, markets, pollsters, academia, tv experts) in both their expertise and morality obviously has varied over time and by country. Some of it we obviously have lost (at least temporarily), e.g. trust in the US government, some other areas remain strong.

This situation of having entirely fabricated poll causing a candidate drop out of a race was not on our radar (obviously, or else McMorrow and her doners would have waited for more than 1 poll). Therefore it represents a reduction in public trust with negative externalities beyond just this one race.

Me too, I got one today, don't think I had anything in there though

Hard to say what the takeaway or lesson is here, other than having trustworthy professionals disseminating unbiased information is a public good and losing it creates paranoia and social distrust everywhere.

probably use tor and nothing tied to your identity

This doesn't seem like a full article without bothering to interview ANY of the arrested 130 people.

What I will say is that getting tiktoks in front of hundreds of thousands of people in a coordinated way isn't easy, and they won't come out if they weren't already passionate about something. If there are written records proving this was made to cause chaos specifically then they should subpoena tiktok for exact records.

Do you really consider Scuba as good as datadog? Or splunk?

The article's arguments "Look if you squint logs and traces kinda overlap" misses the point that if you want to use traces effectively you need to be able to graph them as flame-graphs and also query these relationships ("count average number of userLogin spans that failed that have a subsequent succeeding userLogin span with the same userId and same originatingIp")

To be fair even the more advanced products struggle with sophisticated queries, though splunk and sumo have usually met my needs just fine. Honestly even SQL has an order of magnitude more powerful querying than scuba.

Ludicrous take. I'm tired of people who aren't experts on observability having such strong opinions.

Having used scuba at meta, I can authoritatively say it's bottom-tier quality. Especially in the era of having competent LLMS able to write sophisticated queries.

Genuine expert level observability is being able to query things like "Of all the users who updated versions of the APP in the last week, what % of them had fewer errors after the update (per hour of usage) than before"

That will never happen in a "wide event."

I haven't looked into TCL at all, but everything I know on this topic I learned in one day, so if you have even a $20 a month AI subscription and you're passionate I'd say give it a go.

There was just a post yesterday about how to root your samsung. I rooted my samsung and removed ads, mics, and my agent is looking into this one as we speak [codex].

Happy to post to github if you'd like.

Yesterday's post (about rooting a samsung) was a $3,000 TV. I don't think it's about affordability.

There's another analysis there showing that they still send this even if you don't agree to it. I feel like this needs to go to court.

Just yesterday I hacked my Samsung Frame tv. The AI (codex) used an off-the-shelf root with Developer Console access (built in). Though root doesn't give you the signing keys to blow everything away and still boot (is my best understanding), but as a workaround you can still change anything you want after boot automatically.

I had it do:

- Stripping ads

- Disabling mics

- Turning off updates

- Leaving root SSH in place (protected) for future access

The key to getting codex to do this was that this is legal under DMCA, but after a certain point I had to get a new session to change all the "hacking" words into synonyms to not trigger false security alarms. Wouldn't be surprised or bothered if my chatgpt account gets banned though. I can write a blog post and publish the repo if you want

Could be. Another simple explanation is that there is a lot of bot activity messing with the algorithm.

I'm relatively sure I see bot activity on public facebook groups where I live in a blue county in a blue state and see hundreds of vitriolic angry comments that aren't aligned with voters current frustrations (i.e. economy)

The only one I can think of is Arcane season 1. What are you thinking of?

Fallout on par with Andor??? Wild take. The only thing Netflix has ever done that comes close to Andor imo is Bojack. I felt Fallout had one good season then quickly became rudderless and at this point is half a notch above all those shows that I can't even remember how many seasons I've watched because it all blends together.

Yes I mean the US, the DMCA is a US law. You can google it but "The specific legal source that allows you to hack or jailbreak your own smart TV is the U.S. Copyright Office's Triennial Section 1201 Exemptions."

Agreed. But in the meantime this type of hacking is entirely legal under DMCA, that's actually quite clear.

Though it would be nice if it didn't require such work.

But does this mean they don't turn this on by default resell that data? X house has dogs or whatever?

We clearly need a whole new privacy framework where everything is optional and opt-in, plus open software to monitor this compliance.

What are you talking about? I have a samsung frame that I'm in the middle of rooting because of this excellent information.

This is great. Thank you, I love this.

We can talk about TV all we want -- but the answer is youtube (and no not youtubeTV or whatever). We're seeing entirely free shows better than what you'd get on netflix (e.g. The Amazing digital circus) and a few orders of magnitude more content.

It's TV "2.0" (meaning decentralized creators no gatekeepers) and the younger generations are watching it far more than anything else (and producing impressive content)

Convenience is king -- unfortunately if I'm paying for something that has ads, doesn't have the movie I want for x months (or at all), rotates out something I had planned to watch, has a clunky UI, etc then actually jellyfin becomes the path of least resistance, aside from potentially being cheaper.

I recently did an experiment with LLMs where I asked them to rate the odds of a given outcome explaining an observation, and I found they are systematically biased much like humans [1].

Even if they can do great maths I think bayesian statistical fundamentals are not reliable at this point.

[1] https://think-twice.me/is-ai-biased-or-sexist-making-chatgpt...

> It wasn't about Theranos though, but the person behind it.

But the legally proven history of all the things she did (and still refuses to admit!) give a very clear picture of who Holmes is, a much more accurate picture than you get from hearing her lie and dodge questions.

1. I'm not sure who matters that much, the point is you need to make it more cost-effective to do the right thing. It could be CEO 6 days in jail, it could be that every action requires a signed affidavidt of safety and in the lack of such an affidavidt the manager is held accountable (cascading up to CEO).