veunes

Sending someone a dumb meme can also be a form of affection

I don't think the point of the strawberry example is that industrialization failed to make strawberries cheaper or more available. It obviously did the opposite in many places. The point is more about what gets selected for when the whole system optimizes for scale, consistency, shelf life, lowest acceptable cost

When housing, healthcare, work, social life all feel unstable, the predictable option starts looking less like boring conformity and more like one less decision that can go wrong

Yeah, I think cities are probably the clearest physical-world version of this

This feels related to Goodhart's law, but applied to pleasure and culture

The uncomfortable part is that each step is usually justified by a real use case

The funny part is that "put your browser profile on a ramdisk" used to sound like an obsessive performance tweak, and now it starts to look like a privacy mitigation

What stood out to me is how little of this is really new to Silicon Valley

If a model like that exists behind closed doors, we're unlikely to know from press releases

The interesting failure mode here is that the defensive side needs access before the offensive side starts using the capability

The wholesale spread makes batteries look obvious, but I'd be careful about extrapolating from today's arbitrage. If enough batteries get built, the spread they rely on compresses

This sounds less like a dispute about "the mission" and more like a dispute about who should have controlled the mission. The nonprofit/for-profit structure is messy, but so is the idea that putting OpenAI under Tesla would somehow have preserved its independence

This is a good reminder that encrypted and end-to-end encrypted are very different promises

This is a lot less strange if you think of the robot as participating in ritual rather than "having faith"

45 is painfully young

The manipulation risk is real, but it usually comes from pretending there is a painless answer

The "old car" analogy seems right, with the extra complication that the car is supplying a non-trivial chunk of the country's electricity and replacing it is not quick

The interesting part will be whether Belgium can turn this into a coherent long-term plan

I bet if you could look at the hidden reasoning tokens at the exact moment the DB was dropped, there were zero thoughts about safety rules in there. The model simply hit an access error > searched for a token > found one > ran the command. That whole "I am violating my instructions" vector only fired up after the pissed-off user fed it a prompt full of accusations. So yeah, it's not a confession at all, it's just the model adapting to the user's context

"backups in the same volume" aren't backups, they’re just snapshots in the same blast radius fwiw. If your DR plan hinges on a single physical volume ID, you have zero resilience

This needs to be a lesson for everyone: real backups belong in an independent store (S3/GCS) in a different region with object lock enabled. It’s the only way to make sure even a compromised root token can’t nuke your data for 30 days

CORS protects your Facebook from your Gmail, but it won't protect your Gmail from the agent itself since it already has access to the DOM and JS context. If that agent gets hit with a prompt injection and decides to "Delete all mail" or exfiltrates session tokens to a third-party endpoint, the browser sandbox will actually facilitate it because it views those as legitimate user-initiated actions

Sure, 26B models on beefy desktop silicon are finally nipping at the heels of commercial APIs, but this is a mobile thread. On a phone with 8GB of RAM and passive cooling, your tokens per second (t/s) are going to fall off a cliff after the first minute of sustained compute

It’s likely a llama.cpp backend issue. On the Pixel, inference hits QNN or a well-optimized Vulkan path that distributes the SoC load properly. On the iPhone, everything is shoved through Metal, which maxes out the GPU immediately and causes instant overheating. Until Apple opens up low-level NPU access to third-party models, iPhones will just keep melting on long-context prompts

This article is all fluff because real benne marketing. If they mentioned that a 4B model on an iPhone 16 drains 15% of the battery for a single long prompt and triggers hard thermal throttling after 20 seconds, nobody would be clicking on headlines about "commercial viability" fwiw

I noticed the inference is routed through the gpu rather than the Apple neural engine. Google’s engineers likely gave up on trying to compile custom attention kernels for Apple’s proprietary tensor blocks iirc. While Metal is predictable and easy to port to, it drains the battery way faster than a dedicated NPU. Until they rewrite the backend for the ANE, this is just a flashy tech demo rather than a production-ready tool

Yeah, a lot of it only becomes obvious in hindsight because each individual signal is easy to rationalize away

You're not just delivering expertise, you're stepping into a situation where incentives are already misaligned, expectations are fuzzy, and there's often a cashflow problem hiding somewhere

I think the hidden advantage here isn't even enforcement, it's filtering

This is all correct in principle, but in practice it's a lot messier

This reads less like a "got ripped off" story and more like a perfect storm of every classic consulting red flag showing up at once