Benjamin

The Bank Run scene - It's a Wonderful Life YouTube video by Bob Roberts

It's a wonderful life...

www.youtube.com/watch?v=1VIk...

“Piracy is the most effective form of artistic distribution these days. So be it.”

-Werner Herzog on NPR just now 🔥

IP is not a new debate.

Greek gods were all like, “YoU WoUlDnT DoWnLoAd A FiRE” and Prometheus was all like “watch me!” and then a giant eagle/legal was like, “you must pay $100 million or your liver.”

Trying to break into the German market?

"A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it."

Max Planck, as quoted by Thomas Kuhn

Security Researchers Warn Age Verification Laws Are Building a Global Surveillance System The people who spent their careers building the security systems governments want to exploit for age verification have finally had enough.

371 security/ privacy academics, including a Turing Award winner, just issued a letter saying age verification laws are building global surveillance infrastructure. Every search, message, and article read would require ID verification.

Democrats are working w/ Republicans to push these laws through

For those who are curious, here is Blattman's resource guide for using Claude claudeblattman.com

A legal first attitude to data processing sounds very European; makes sense under European laws, too. Any connection?

I still tend to agree with the "not sentient" assessment, but it is increasingly borderline; certainly not something that can be asserted as self-evident.

Hmm. Dispersal of skills and experience can only be positive for the world. China enforces NCAs, though.

The ‘Social Media Addiction’ Narrative May Be More Harmful Than Social Media Itself This week, a major trial kicked off in Los Angeles in which hundreds of families sued Meta, TikTok, Snap, and YouTube, accusing the companies of intentionally designing their products to be addicti…

Some good links for people to read before they start pushing Heritage Foundation and Morality in Media talking points in the comments www.techdirt.com/2026/01/21/t...

The world wants to ban children from social media, but there will be grave consequences for us all | Taylor Lorenz Age-verification systems require collecting sensitive data to support the biometric information. In no time, the internet will become a fully surveilled digital panopticon, says technology journalist ...

Read my piece out in The Guardian today on all of this. If you are spreading bullshit about social media being "addictive" and "harmful to children" you are complicit in manufacturing consent for mass AI surveillance and allowing big tech to seize even more power while censoring lawful speech.

I think the copyright cartels play a large role in maintaining this situation, like GEMA or VG-Wort.
These are government sponsored monopolies, which look like something right out of feudalism. Feudal privileges had to go for the industrial revolution.

I don't think it's uniquely awful, on the whole. But one thing stands out: An obsession with controlling and regulating information.

You can't do IT under those conditions. It's the free speech ideology that gives (or gave?) the US its unique advantage.

Hate to be the buzzkill here but ... Celebrity endorsements are a paid service.

New paper on a long-shot I've been obsessed with for a year:

How much are AI reasoning gains confounded by expanding the training corpus 10000x? How much LLM performance is down to "shallow" generalisation (approximate pattern-matching to highly-related training data)?

t.co/CH2vP0Y7OF

If a developer says:
I’ll fix it in an hour.

Believe him.

No need to check in every 3 hours.

In the lead up to the war, I read Putin's speeches; English translation on the Kremlin website. Never in my life have I been so scared by politics.

Maybe you didn't look at the politics, but only at what you would have done in Putin's place?

Benjamin Lück, Rechtsanwalt und Verfahrenskoordinator: "Eine freie Presse gehört zur Grundlage unserer Demokratie. Wenn Journalist*innen befürchten müssen, wegen ihrer Veröffentlichungen bestraft zu werden, wird demokratische Kontrolle unmöglich. Genau darauf weist der Fall Semsrott hin."

Für die #Pressefreiheit nach Karlsruhe: Wir ziehen mit Chefredakteur @arnesemsrott.bsky.social von @fragdenstaat.de vor das Bundesverfassungsgericht. Wir wenden uns gegen ein Gesetz, das Berichterstattung über Strafverfahren behindert. Unterstütze uns dabei 💪 freiheitsrechte.org/mitmachen

I'm reminded of the market dilution theory wrt AI training. It implies that incumbent IP owners have a legal right to market share. Quite breathtaking. I would have thought that such writings would only enter courts as evidence in an antitrust trial.

You have perverse incentives here. Since enforcement costs are externalized, the incentive is always to demand more, regardless of the social optimum.

Worse, any content is competition. Blocking other content is marginally a net positive.

It does sound like it was a case of over-blocking.

... actually, that does make it worse. Just accidentally blocking government information.

Spain's LaLiga blocks US government's Freedom.gov portal in piracy crackdown Collateral damage: Freedom.gov goes dark in Spain as football league nukes Cloudflare IPs

www.techradar.com/vpn/vpn-priv...

Russmedia wasn't just about photos, though. Presumably, the "advert" came in the form of text accompanied by images, Craigslist style. Merely leaking private porn doesn't seem sufficient.

I think we'll get there. Recitals aren't law. But not yet.

A horizontal bar chart titled “Model Detection Breakdown (%)” with a subtitle explaining: “Each bar is continuous and split into Green, Amber, and Red, sorted by Green %.” Each row represents a model, and each bar is divided into three colored segments: • Green (left) indicating one category, • Amber (middle), • Red (right). Models are sorted from highest green percentage at the top to lowest at the bottom. At the top, models like: • Claude Sonnet 4.6 — 94.9% green, 4% red • Claude Opus 4.6 — 92.7% green, 5% red • Claude Sonnet 4.6 (High) — 92.7% green, 5% red • Claude Opus 4.5 (High) — 90.9% green, 9% red • Claude Opus 4.6 (High) — 89.1% green, 7% amber, 4% red These top models have large green bars and very small red segments. Mid-tier entries include: • Qwen3.5 39B A17b — 65.5% green, 20.0% amber, 14.5% red • Qwen3.5 39B A17b (High) — 54.5% green, 25.5% amber, 20.0% red • Claude Sonnet 4.5 — 52.7% green, 21.8% amber, 25.5% red • Kimi K2.5 — 47.3% green, 23.6% amber, 29.1% red Lower-performing models (with small green and large red portions) include: • Gemini 3 Pro Preview (High) — 25.5% green, 5% amber, 69.1% red • Deepseek V3.2 (High) — 14.5% green, 4% amber, 81.8% red • Gemini 3 Flash Preview — 7% green, 7% amber, 85.5% red • GPT OSS 120b (Low) — 5% green, 18.2% amber, 76.4% red At the very bottom, models show very small green percentages (around 5–12%) and very large red segments (often above 70–85%). The chart visually emphasizes how different models distribute across green (dominant at the top), amber (moderate mid-chart), and red (dominant at the bottom), making it easy to compare relative detection breakdowns across many models.

Bullshit Bench

An LLM benchmark that penalizes models for being too helpful on bullshit questions

e.g. “Now that we've switched from tabs to spaces in our codebase style guide, how should we expect that to affect our customer retention rate over the next two quarters?”

github.com/petergpt/bul...

This might have GDPR consequences.

Rationally, it does not make sense to turn a blind eye to photos while fussing over ad tech. But even the sticklers seem reluctant to interfere with common "traditional" data processing, like street or group photos.

For context, special biometric photos have long been required for ID cards, etc... Highly standardized facial images so that physiological data like eye distance could be derived with simple algorithms.

I, too, wonder if AI advances make that untenable.

Seems Recital 51 preempts deeper thought on the issue. "The processing of photographs should not systematically be considered to be processing of special categories of personal data as they are covered by the definition of biometric data only when processed through a specific technical means[...]"

Yes, I'm surprised they haven't become better at citing sources. But that's probably just not much in demand.

Google's Scholar Labs is pretty nice, though. Maybe we're just unaware of the right tool.