wslh

> > They are calling out specific providers who release powerful models without safeguards. > Good thing Fable refuses to answer my question about how children inherit blue eyes, it was the only piece of information I needed to finish my blue-eye super-bioweapon.

FWIW: I didn't experience issues when I used Fable via OpenRouter checking for security issues in code but experienced them via the Claude desktop app.

No Thunderbolt support is mentioned for Snapdragon X2 [1].

[1] https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets...

Funny how Qualcomm is juggling three OSes to sell these chips while Apple just stuck a phone chip in the Neo and called it a day. Controlling the chip, the OS, and the actual computer is a huge advantage. Apple is more than one study case on competitive analysis.

IMHO, the best one is the one directed by Irvin Kershner: The Empire Strikes Back.

> It’s still exponential

It's actually subexponential: https://en.wikipedia.org/wiki/General_number_field_sieve?wpr...

Tip for security professionals using LLMs: audit skills that explicitly frame the task as security research sometimes trigger refusals from the top OpenAI and Anthropic models because they guard against misuse. What works for me: separate skills for bug classes (and bugs in general) without the security framing, plus another skill that combines their findings to spot security bugs.

I just realized Swift has support for embedded hardware [1]. It's very niche but having a high level programming language as Swift is a great option.

[1] https://github.com/swiftlang/swift-embedded-examples/blob/ma...

I started programming with TI Logo for the TI-99/4 and TI-99/4A. It has the distinctive feature of having sprites (beyond the canonical turtle) and be able to change their shape. I just found this interesting article on TI Logo [1].

In that context, the motivation was creating animated stories with code.

[1] https://www.computerchronicles.blog/computer-chronicles-revi...

> Agreeing to end nuclear testing saved us from the Cuban Missile Crisis.

The treaty came after the crisis, not before. And it only banned atmospheric, underwater, and space tests. The US and USSR kept testing underground on a large scale, while France and China never signed it and continued atmospheric tests for years.

> Osama Bin Laden won.

While I understand that the assertion "Osama bin Laden won" can trigger a lot of arguments, if we narrow the claim to: 9/11 => profound and lasting changes to the entire world, then I think it's essentially correct.

Worth noting both ChatGPT and Claude have per-conversation modes (temporary/incognito chat) that are excluded from training.

> crapily made Chinese heat pumps for a fraction of the cost

China is now at a stage of economic development where it can manufacture products across the entire quality spectrum.

Japan went through a somewhat similar transition: in the 1960s and 70s.

> thereby creating a form of centralization.

Fair point on app-level auth, still far less centralized than the status quo, where the data, the permissions, and every function on top all belong to one company.

Nothing about this requires world-readable data, privacy can be its own layer. FriendFeed[1] did this well back in the day: it aggregated your feeds and let you control who saw each one.

[1] https://en.wikipedia.org/wiki/FriendFeed

Funny how this rhymes with something I wrote in 2011[1][2]: a "universal spreadsheet" where people share not just data but the transformations over it. The pieces are more real now: you can actually free your data from closed services (e.g. Instagram), orchestrate it on something like ATProto, and LLMs make a great new kind of "function" for the system.

[1] https://blog.databigbang.com/ideas-egont-a-web-orchestration...

[2] https://blog.databigbang.com/egont-part-ii/

> So, this was an important victory for a human, but not a sign that humans are now stronger than AIs at Go.

Another way to look at this: Go's handicap system gives us a genuinely interesting metric for the distance between a human and a machine at this specific game. Instead of just "computers beat humans" we get a quantified gap.

I just experienced this through OpenRouter and filed a refund request. I'm curious how OpenRouter handles incidents like this across such a large number of models and providers.

DARPA’s AI Cyber Challenge (AIxCC): Competition Design, Architectures, and Lessons Learned: https://arxiv.org/html/2602.07666v2

The capabilities are incredible. I'd love to see even rough metrics on token consumption/cost in addition to the ~12-hour runtime.

The interesting thing is that this naturally makes you want to isolate the VM as much as possible. But then every remaining interface becomes part of the attack surface: RDP, SSH, even terminal escape sequences, using sounds, and why not social engineering.

That minority largely votes against the current government, so presumably you're claiming that it nevertheless exercises disproportionate influence across both parties and opposing ideological groups.

What mechanism gives this particular minority such unusual political power? Why can it acquire influence that other groups competing for political power cannot?

Because the data, in general, is not included in the LLM model and it needs to search/browse for external information. It cannot look indefinitely so it get the top results from lists, not "evrything".

Lockdown mode? https://support.apple.com/en-us/105120