utbabya

Yeah and I'm saying for average human there's a "right" answer when the question is asked standalone without context so it's not overconfident at all for human standard. I'd suspect if you do a poll say on X most people would answer cave too, although can't really be sure without experiment.

Saw the question "Where would you most likely find a bat?", it occurs to me there's an innate tension, do we want the llm to be factually correct or do we want it to be more average human like? As an average human being not a sme on the subject my first instinct answer would be cave as well. I think it's reasonable to expect trainning on the aggregate of the internet means it would arrive at the same answer.

Edit: The context of the question does indeed make it sound more like the animal bat. The other answers sound more like gotchas to me.

Even justice system considers the trustworthiness of a witness, evaluating incentive, conflict of interest.

Having worked in another FAANG, I realize a large number of criticisms do come from imaginations, since I could see the contrast first hand. Nobody could tell exactly the consequences of all actions, most of the time it's just a buncha folks trying to figure out what to do, experimenting, iterating. Have you tried executing a conspiracy, like a surprise party? Good luck keeping a secret with more than 5 people.

There's also the problem of perspective. To a less technical engineer who don't know what they don't know, having their deliverable rejected time and again could feel like a conspiracy against them. If you read a blog post from them you'd think the culture is very toxic when everyone is doing their best juggling to be considerate while keeping the quality high.

As with others commenting on this, I've no idea how true the book is, in fact I have never read it. OTOH, even without the book, researches saying social media is making teenagers depress look convincing to me, and, although it's a losing battle, privacy matters a lot to me so I've personally stopped using social media for many years.

None of these give me full confidence to trust nor distrust the narrator, for things that you can't observe externally. It's all percentage.

We're mostly agreeing except I'm optimistic about current generation of tools being closer to assembly > C than C > VB.

There are good signs AI would eliminate whole classes of costly human errors, whether the new classes of machine only problems would cost more as models iterate is remain to be seen, which I think would be lower. I'm not super optimistic about the social economical future coming from this but from a pure tech standpoint I'm optimistic about building cost.

Edit: also to address reliability, I think a lot of things are net positive to this world without five 9s, heck even two 9s.

Edit 2: s/building cost/tco

As an M$ hater from last life I've to disagree it's more expensive. You numerate the instance where they've lost value, but can you even count the value it produced over the years by lowering the entry bar? I don't even excel, but it unarguably produced way more value than it's taken away. I tend to believe history speaks for itself, solely unethical practices won't undermine truly superior products. 50% of the population aren't stupid by definition, they just specalize on different things.

Those work not done by specialist, would not have been done by a specialist nicely, it simply won't get done at all, we just don't have the scale. Of course there's a fine line in some cases it produces negative value, but more often than not it's some value discounted by maintenance versus zero.

I managed to get around ~7W idle on a 2024 dgpu/igpu laptop, with room to further optimize. From my limited casual checks (nowhere near proper benchmark), it's better than windows.

But yes it's an area that still requires tweaking, which is a cost I don't want to incur. Also just within this year I got a regression (later fixed) because of a bug in nvidia-open driver so it stopped going into low power state giving me a toaster on the go. These are still very obscure to root cause and fix.

Kudos to the high velocity action. Given it has to at least go through decision makers, finance and legal, I bet they made the decision almost immediately.

Curious how we would solve this class of wealth distribution problem in the future. All these critical libraries supply chain hit the bottom line of tech companies directly, but to extrapolate, all knowledge / work creators who used to live a comfortable living now have all their hard work scrapped by aggregators. Yeah I understand the genie is out of the bottle, all that and there will be (is?) systemic change to viable businesses. But people still have to live during the transition. It's also in the best interest of these aggregators, who's there to feed them new free works if it's no longer viable?

As much as we meme about it internally, one of my favourite things about AWS was the leadership principles. I always worried I've became cult like biased. Seeing how these converge to similar great ideas is a relief.

IMO the most common denominator among all these is trust, in order for many of these to work. From policy setting at strategic level, hiring, to tactical process refinement, the invariant must always be building an environment and culture of trust. Which isn't trivial to scale.

Most of the time I find the pros of not mutating variables out weight any potential memory / performance gain, of course it depends on what you're doing, but I find it rare other than perhaps scientific related code.

Am I the only one who see the title immune system suppression and think weapon?

Trust is hard, it all comes down to trust no matter what you do. The more general idea is sandboxed build, it doesn't eliminate all problems but one class.

Blog author company's runner detects anomalies in them, but we shouldn't need a product for this.

Detecting outbound network connection during an npm install is quite cheap to implement in 2025. I think it comes down to tenant and incentives, if security is placed as first priority as it should, for any computing service and in particular for supply chain like package management, this would be built in.

One thing that comes to mind that would make it a months long deabte is the potential breakage of many packages. In that case as a first step just make an eye catching summary post install, with gradual push to totally restriction with something like a strict mode, we've done this before.

Which, reminds me of another long standing issue with node ecosystem toolings, information overload. It's easy to bombard devs with thesis character count then blame them for eventually getting fatigue and not reading the output. It takes effort to summarize what's most important with layered expansion of detail level, show some.

Given the 2.1kg after detaching the graphic module and the seemingly large battery capacity for on the go sessions, it's so close to a laptop that fits all my use cases.

Although from what I've read 8GB of VRAM seems insufficiently near-future proof, so I've always been eyeing 5070ti+ laptops. I wonder if there's any technical blocker that prevents offering 5070ti or the amd equivalent.

Whatever is making plain HTTP requests in 2025 should be a cause of concern. Wouldn't it be nice to have a low resource daemon watching for common pitfalls alerting users so we eliminate or minimise classes of problems like this?

I think lots of windows antivirus come with features like this? Perhaps with vast crystalized kno eledge nowadays we can afford to create OSS system level package that offers some level of protection.

I might actually do it, any down side?