Janmaarten Batstra

@mcuban Yeah, great plan. Let’s all go local with Chinese OS models. 🤷‍♀️

We should federally tax Tokens at the Provider level. Not a lot. Less than 50c per million tokens. It will accomplish 4 things (at least ) 1. It will push the big AI players to optimize tokenization, caching , routing and localization Which will 2. Reduce energy usage. Saving them in energy costs more than what they paid in tax and reducing strain created by the growth in energy consumption Which will 3. Generate maybe 10 billion dollars a year to start, but over the next ten years could grow 30x to 100x Which will 4. Create a source of funding to pay down the federal debt or deploy, in response to the things AI brings that we don’t expect or don’t like At some point the models will pass it on to customers. Of course. That’s ok. Customers will have the ability to choose between providers. Or to do everything using open source models locally. Thoughts ?

@StephenKing That’s just not true. @elonmusk was placed on recall status, but no geographical restrictions were set. So as long as he can come back (reasonably) quickly (which he can, having a private jet) when called, no law or order is broken.

You went to China with Trump when the judge told you to stay put and attend the ongoing trial. So I guess the law doesn't apply to people once they pass a certain amount of wealth.

@elder_plinius @MatthewBerman Jensen is wrong on this one (And he knows it). Simple logic: if an attacker fails, he can try again and again. If a defender fails, it’s game over. With humans, attack capacity is constrained by Human Resources. With autonomous agents, it’s 24/7/365.

@MatthewBerman hmm

Jensen: if your AI is better than the attacker's AI, you're fine. A small malicious team is never going to have more compute than OpenAI/Anthropic and certainly not more than a country. Hope Jensen is right 🤞

@dvdeijk @huibmodderkolk En dat is (samen met regulatory, HR, budget en nog steeds een gierend gebrek aan bewustwording) de catch 22.

@huibmodderkolk Maar met die supply chain attacks wil je eigenlijk juist niet de hele tijd updates hoeven downloaden

Het Britse National Cyber Security Centre waarschuwt: door AI zullen organisaties binnenkort als een gek moeten gaan updaten om alle gevonden kwetsbaarheden in software te verhelpen. Alleen dan kunnen ze hun systemen veilig houden. ncsc.gov.uk/blogs/prepare-…

Mja. Denk dat ze gedacht hebben: geen schade want Lec al naar achteren gezet, maar wel ff puntjes op de i zetten om een signaal te geven voor de toekomst. Eigenlijk wel prima, anders is het signaal dat je wel een heel klein beetje over de lijn mag onder geel en dan komt daar weer gezeur over in een toekomstige case.

@Olav_Mol Nu al flink bezig. Kans op cancel… 50/50?

FIA/ FOM lijken de juiste beslissing genomen te hebben met 1300 lokaal starttijd van de GP van Miami… vanaf 1500 lokaal gaat het behoorlijk los daar. Alles nu dus in Ned vanaf 1900 op Grandprixradio.nl en in de app

Google is ook een exoskelet dan? En een calculator? En een telraam? 🤷‍♀️ Generieke onzin. Voor een % van de mensen zal dit vast deels kloppen. Voor een ander % vergroot het hun horizon, verbreedt het hun kennis en initieert diepgang. Alleen al de Grok knop bovenaan elke post toont dit aan. Ipv een post lezen en dat al dan niet voor waar aannemen, geeft het context en uitleg met een druk op een knop.

Hoe AI ons brein verweekt. Uit interview, door @LaurensVHG. volkskrant.nl/tech/deze-gedr…

@KRoelandschap HW3 of HW4?

Dutch authorities warned for extreme weather Thunderstorms, hail and lots of rain FSD has met its system boundaries and couldn’t handle the extreme weather conditions

@pbeyssac On the contrary. Disclosure was done very responsibly, late even. This is the age of AI, other systems will find the CVE as well. At least now people know and can blacklist Alg/A. In combination with the cPanel bug its way worse btw: x.com/interminded/st…

Je ne suis pas très exposé aux failles Linux (peu de serveurs), mais pas terrible la divulgation de #copyfail alors que les corrections n'ont même pas encore percolé jusqu'aux distributions majeures :(. En d'autres termes, même avec un système à jour, vous pouvez être vulnérable.

@UNGeneva @UN You are warning? Now? That’s about 2.5 years too late . Also, the AI you can press here on X (@Grok) on any post is a very good way to verify that post in terms of disinformation. Seems worth mentioning 🤷‍♀️.

Unchecked AI use in advertising risks fuelling information crisis, the UN warns. A new @UN brief says #AI-driven systems are accelerating the spread of disinformation, hate speech and polarizing content. Learn more: buff.ly/sALFZbk #AIForGood

@ProtonPrivacy @ConlustroR The word shocking (with the …) begs to differ. Also, your post is factually incorrect. EU published preliminary findings, that suggest a breach. That’s very much not the same as ‘found guilty’. But do as you please, ofcourse. Freedom of speech ftw. x.com/digitaleu/stat…

@ProtonPrivacy @ConlustroR I didn’t cite anything. I explained to you what people read, not what you wrote. Look at the numbers. After 1 hour, your first post in the thread gets 12k views. The second one 2.2k. And you confuse creating legislation with marketing it.

I am fully aware as to your position and I fully support it. But assuming the EU commission will not use any 100k+ followers account that states anything they can use to their advantage is presumptuous and does not align with what is currently going on in Brussels. And next to that, >360k followers just read ‘Meta endangers kids!’ From you, and (intended or not), this does influence people’s minds. People don’t read complete stories and consider all provided facts. They take 5 second bites of a story and a headline. That’s reality, fair or not.

@interminded @ConlustroR The EU is not looking to our X account for inspiration, and people should be informed. If you read down the thread we've given a take on issues adjacent to this: proton.me/blog/keep-age-…

@ProtonPrivacy @ConlustroR Yeah, that’s true. But, @ProtonPrivacy; the net effect of it, intended or not, is that you give the EU commission extra fuel to push for their highly unsafe and surveillance hungry ‘age verification app’.

@ConlustroR No, we've reported on something that is happening by spelling it out in a thread.

@DKThomp @Miles_Brundage That’s a sign the govt has no idea what’s actually going on, what it all means and how to handle it. Pretty much just as the rest of the world when it comes to AI 🤷‍♀️

I can understand a plan to treat AI as a "normal" technology and let Nvidia export powerful chips to China. And I can understand a plan to treat AI as an "abnormal" technology that compels the govt to stop private companies from selling their products and services because they're too dangerous. But current AI policy treats AI as both normal and abnormal. We're selling China the compute necessary to build technology that we're also declaring too dangerous to release. And I think that's weird.

@ChrisRMcGuire The fact that it is illicit does not mean it’s not industry standard .🤷‍♀️

Sorry but that just isn’t true—distillation attacks are illicit activity, not an industry standard. They are against the terms of service of all frontier AI labs. There is a reason OpenAI, Anthropic, and Google all put out reports warning about it: none of them do it.

@BernieSanders Given that those ‘oligarchs’ all went to the govt and politicians to plead for regulation BEFORE IT WAS TOO LATE and those politicians (including you, @BernieSanders) did nothing at all, why are you blaming them? You know very well that if they quit now, China an NK will not.

The moment in time where a 14 yo script-kiddo with a laptop and an investment of 0 (zero) euro can steal all data of > 10k websites (all email, login data, internal stored data, user data) within 24 hours is…. Today. What could possibly go wrong… 🤷‍♀️. #CopyFail #cPanel

It gets worse. CopyFail (CVE-2026-31431) + the fresh cPanel auth bypass (CVE-2026-41940) = trivial remote-to-root chain on hundreds of thousands of shared hosting servers. Fully automatable by agents: • Scan exposed cPanel ports (2082–2087) • Auth bypass → impersonate any tenant (no creds) • Drop your 732-byte Python PoC → instant root on the entire box Attacker owns every website, database, email & backup on the machine. Worm potential is insane Bigger than Heartbleed? 🤷‍♀️ Autonomous agents changed the game.

Time to talk about this one. CopyFail (CVE-2026-31431) — a 732-byte Python script that roots every Linux distro shipped since 2017. 🧵

⚠️ URGENT WARNING – Remote-to-Root Chain on Shared cPanel Servers A trivial combo is now public (disclosed last 48h): •cPanel auth bypass (CVE-2026-41940) •copy.fail Linux kernel LPE (CVE-2026-31431) Attack flow (fully automatable by agents) gives Instant host root on the entire shared server Consequences: •Attacker owns all websites, databases, emails & backups on that machine •Can turn the server into a new attack node to scan & infect others (low barrier worm potential) Hundreds of thousands of shared hosts remain vulnerable right now. Remedy (do this TODAY): •Apply cPanel + kernel patches immediately •Blacklist algif_aead module ASAP (stops copy.fail instantly, no reboot):