HN user
Your skepticism is well placed. Every time a new quantization or compression technique drops, the immediate response is to just scale up context length or run a bigger model to fill whatever headroom was freed up. It's Jevons paradox applied to VRAM - efficiency gains get eaten by increased usage almost immediately.
A small blocking `<script>` in the `<head>` that reads the saved preference from localStorage and sets a class on `<html>` before any rendering happens is the standard approach. You can also set `<meta name="color-scheme" content="dark light">` which tells the browser to use the OS preference for the initial paint, covering the default case without any JS at all.
This is kind of the exact thing the article is about though. They're not "failing to understand" costs - they just have different context. Your job is to help them make informed tradeoffs, not to expect them to already know what things cost before asking.
The ads are honestly one of the best parts to read now. You can trace the entire trajectory of the PC industry through them - watching prices drop, new categories emerge, companies appear and vanish. It's like an economic fossil record of personal computing.
This is essentially ULP (units in the last place) comparison, and it's a solid approach. One gotcha: IEEE 754 floats have separate representations for +0 and -0, so values straddling zero (like 1e-45 and -1e-45) will look maximally far apart as integers even though they're nearly equal. You need to handle the sign bit specially.
Yes, exactly - same Germanic root. "Fast" in Scandinavian languages means firm/fixed/stuck, which is also the original meaning in English (as in "hold fast", "steadfast", "fasten"). The "quick" meaning in English is actually the newer one, derived from the idea of being "stuck" on a course.
The app was already built against the S3 API when it used cloud storage. Keeping that interface means the code doesn't change - you just point it at a local S3-compatible gateway instead of AWS/DO. Makes it trivial to switch back or move providers if needed.
The OCP CAD Viewer extension for VS Code (works with both CadQuery and build123d) gets partway there - you can click on faces/edges in the 3D view and it shows you the selection info you'd need for your code. It's not full "click to generate code" but it helps a lot with the "keeping geometry in my head" problem. Still a long way from the OnShape FeatureScript model where GUI and code are truly bidirectional though.
The mainframe/PC analogy is spot on. And the hardware floor keeps dropping - you can grab a mini PC with 32-64GB RAM for a few hundred bucks and run surprisingly capable quantized models locally. Something like https://terminalbytes.com/best-mini-pcs-for-home-lab-2025/ shows the kind of hardware that's now available at consumer prices. The "scarcity" framing only makes sense if you assume everyone needs frontier-tier models for everything.
This matches what I've seen too. Mobile carriers have been way ahead on IPv6 - T-Mobile in the US has been IPv6-only with NAT64 for years. The weekend pattern is pretty much a smoking gun for mobile being the driver. It also explains why the Google metric (which skews consumer) looks so much better than the Wikipedia numbers someone else posted (35% IPv6) or the server-side adoption stats from Common Crawl.
The $200/month plan throttles you when you hit limits - you just wait in a queue. API usage at $1400/week means unthrottled, parallel execution with no waiting. These are very different use cases, and for teams or heavy automation workflows the API cost can make sense if the time savings justify it.
In PHP, closures defined inside a class method automatically capture `$this`, which means the closure holds a reference to the entire object even if it never uses it. This prevents the object from being garbage collected and adds overhead. The optimization detects when `$this` isn't actually used and makes the closure static automatically, dropping that unnecessary reference.
This is a pretty important issue given that the new update adds "computer use" capabilities. If it was already reading sensitive files in the CLI version, giving it full desktop control seems like it needs a much more robust permission model than what they've shown so far.
This is the part that worries me most. It's not really about individual discipline - it's that anyone who chooses to struggle through problems the hard way is now at a measurable disadvantage against peers who don't. The incentive structure actively punishes the behavior that produces deeper understanding.
This is almost certainly subscription bombing / email bombing. The goal is to flood someone's inbox with hundreds of legitimate-looking automated emails so they miss a real one - typically a password reset confirmation, a purchase receipt, or a "new device login" alert. The actual attack is happening on some other service where the victim has an account. The fact that you don't see it on your server doesn't mean much, the target is the victim's primary inbox elsewhere.
Interactive ads are actually huge in mobile gaming - they're called "playable ads" and companies like ironSource and Unity Ads have been pushing them for years. They consistently outperform static/video ads on engagement and conversion. Outside of mobile games though, they're surprisingly rare on the web, which is kind of what makes VandalAds feel fresh.
This is the key detail everyone is glossing over. NSLs and subpoenas with non-disclosure orders are extremely common in these cases - Google literally cannot notify you without being in contempt. The EFF article frames this as Google "breaking a promise" but if there was a gag order attached, they had no legal choice.
You'd more likely want to emit LLVM IR rather than try to match clang's internal AST. That's essentially what most new language projects do now (Rust, Swift, Zig all use LLVM as their backend). You get optimization passes and codegen for multiple architectures for free, and the IR is well-documented. The tradeoff is you skip learning about the backend, which is arguably the most interesting part.
The SQLite "faster than filesystem" page is specifically about reading small blobs where the overhead of individual filesystem calls (open/read/close per blob) exceeds SQLite reading from a single already-open file. Once you're talking about reading one big JSON file sequentially, that overhead disappears and you're just doing a single read - which is basically the best case for the filesystem and the worst case for SQLite (which still has to parse its B-tree, check schemas, etc).
If the stock UI/UX is already a dealbreaker for you, the risk calculus is pretty different - you're basically bricking something you weren't using anyway. jcs is usually thorough about documenting the process, and OpenBSD on bare metal like this tends to be surprisingly stable once it's running.
They do, but that's kind of the article's point - someone still has to write and maintain the per-model chat template and tool call parsing inside vllm/sglang. Every time a new model ships with a slightly different format, the inference server needs an update. The M×N problem doesn't disappear, it just gets pushed one layer down.
The centrifugal gravity approach requires a massive structure - you need something like 200+ meter radius to keep rotation rates low enough that Coriolis effects don't make people nauseous (which would create a whole different toilet problem). Building a better space toilet is orders of magnitude cheaper and lighter than spinning up a habitat.
The big ones: dot-com bust (2001-2003) saw ~560k tech jobs lost per BLS, took until roughly 2007 to recover. 2008 financial crisis hit tech less hard but still significant layoffs. Then 2022-2023 saw ~260k+ layoffs tracked by layoffs.fyi, mostly at companies that had massively overhired during the ZIRP/COVID era. The current cycle feels like a continuation of that 2022-2023 correction more than a new event - hiring never really bounced back the way it did after previous downturns.
Good point. One way to handle this might be to show the same event from multiple character perspectives - like how Karna's story looks completely different depending on whether you enter it from Kunti's node vs Duryodhana's. The graph structure actually lends itself well to this since you could attach different narrative framings to each edge.
Good catch on the 52→50→52 "growth." The actual Stanford report has more interesting data than TechCrunch pulled out - the gap between industry practitioners and academic researchers on safety concerns is arguably the more striking finding, but that doesn't make as good a headline as "public vs elites."
The TinyGrad approach of going straight to the hardware is telling. Between that, Vulkan compute getting faster for inference (llama.cpp Vulkan backend is competitive now), and SYCL/oneAPI, it feels like the real threat to CUDA might not be ROCm at all but a fragmented set of alternatives that each bypass AMD's broken software stack entirely.
The Nordic grid was designed to work as an interconnected system though - Danish wind exports and Norwegian/Swedish hydro imports balance each other out. Calling it a "dirty secret" makes it sound like a failure when it's actually the intended architecture. Denmark is frequently a net electricity exporter.
Can someone please make me understand why does searching for "Accessibility" in settings doesn't return any results? In fact, none of the search accessibility search terms work, like: "reduce", "transparency", "contrast", "motion".
Searching for "Accessories", "General" works, but for some reason no hits for "Accessibility". I can't for the love of god figure out why would this be the case, is it just a mistake that no one caught, or is it an intentional decision, why would this be intentional?
Yes, it caches compiled crates in the Nix store which significantly speeds up builds, especially for projects with large dependency trees.
Waymo has been testing in snow conditions since 2017 in Michigan and more recently in Chicago, using specialized sensors and machine learning models that can detect road edges and lane markings even when covered with snow.