HNHacker News
TopNewBestAskShowJobs

wtallis

22,096 karma · joined September 20, 2009

submissionscomments
wtallis··on StarFighter 16-Inch
Case temperature is very much at the discretion of the laptop OEM. Some OEMs take regulatory limits on skin temperature seriously and ship a well-tuned thermal control system that keeps the case at a comfortable temperature. Others push close to the legal limits to keep fan noise in check. Others ship plastic enclosures so they can get away with even higher temperatures (since plastic has lower thermal conductivity than metal, and thus a harder time cooking your thighs) at the expense of more noise.

The StarFighter has a metal case, so when running at high power levels (45W sustained according to the spec sheet) it will either get uncomfortably hot somewhere on the case or at least a bit noisy from the fans, but since it's a bit thicker than the 2019 MacBook Pro it should be able to cool itself more effectively. But when running at the performance level you're used to the power draw should be plenty low enough to make temperature and fan noise not a problem: roughly double the peak CPU performance means you can turn down the power limits a lot and still have a better-performing machine.

wtallis··on StarFighter 16-Inch
A 2019 MacBook Pro would have an Intel Skylake processor (N-th re-release), made on Intel's stagnant 14nm process. The older Intel option for the StarFighter has its CPU cores made on an Intel process two generations newer, and the rest of the chiplets made by TSMC. The newer Intel option moves the CPU chiplet to TSMC as well. They're in a very different league for power efficiency than your current machine, both from the fab improvements and from having a microarchitecture that's not from 2015.
wtallis··on StarFighter 16-Inch
I think Framework is a good example of how smaller laptop OEMs end up shipping late, often on the order of three quarters. This is something else entirely, if any of these configurations are recent arrivals (I don't think they are).
wtallis··on StarFighter 16-Inch
I only use a number pad for playing a few games, and for bulk data entry. Neither of those use cases are something I prefer using my laptop for, and even on my desktops they're rare enough that I'd much rather have the number pad separate and largely out of the way.

What do you use a number pad for often enough to not only see it as mandatory for you, but to leave you unable to imagine how anyone could live without it?

wtallis··on StarFighter 16-Inch
Is there something new here? The processor options seem to be two generation old Intel, one generation old Intel, and one generation old AMD.
wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
Major feature updates require consent, yes. Automatic updates are for bug fixes and security patches. It is disrespectful to the user to automatically deploy any update that changes the UI without warning or consent. I like my toolbar buttons to stay where I left them.
wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
When looking at the power consumption across the whole network path and not just a single link, most of the power draw is probably baseline static power costs of keeping all the routers and switches running. Which means that judging the impact of a download in terms of Watts per MB/s is a pretty bad way of analyzing this.
wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
It should only need to load one copy of the weights, but each tab/site will need a separate context and KV cache.
wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
> Most of this is answered under the tag "Intent to Experiment" and the associated link. It's not a mandate that they're forcing on the web today, it's a public experiment intended in part to solicit feedback for a potential spec: https://groups.google.com/a/chromium.org/g/blink-dev/c/6uBwi...

That does pretty much settle the question: if it's still just an experiment, Google should be asking for consent and web developers should not be assuming that this is the future of browsers, and nobody should be acting like this is a foregone conclusion.

wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
My biggest beef with the browser's unchecked scope creep is that its role as an application platform is fulfilled in a manner that's completely at odds with the original purpose of a browser as a user agent. They're running untrusted third party code by default, letting sites embed spyware, and constantly adding more anti-features. Treating this as normal is insane from an unindoctrinated viewpoint. Acting like people who question this status quo are being unreasonable is an insult to everyone who cares about privacy or security.
wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
The question of whether web browsers will one day be expected to normally include an LLM is not at all relevant to the situation at hand, where that absolutely is not the expected or typical behavior of a browser. Google should be asking for consent now even if they expect in several years time to be able to presume consent for this.
wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
Google's marketing for their latest new browser feature nobody asked for shouldn't be taken at face value. Somebody outside Google needs to provide a well-reasoned assessment of the feature proposal.

And having skimmed that page, it really doesn't answer most of the important questions. Are other browsers expected to ship Google's model, or put a different model behind the API that Google has documented as being specifically for Gemini Nano?

wtallis··on Google Chrome silently installs a 4 GB AI model on your device without consent
You haven't even tried to provide a hypothetical example of what a web app should try to do using a local LLM, nor addressed the obvious questions about how that kind of thing should be standardized, what level of local LLM capability is reasonable for a web app to expect, or how permissions for that should be managed given that a local LLM is not just a tax on local storage capacity.

So why should anyone take it as a foregone conclusion that this is an instance where web devs should get what they want? In general, the browser should be acting in the best interests of the user and not automatically granting the wishes of every web site that wants to drain your battery.

wtallis··on BYOMesh – New LoRa mesh radio offers 100x the bandwidth
Meshtastic supports store and forward for ESP32 nodes that have a few MB of RAM, but not for the nRF52 devices that can't practically buffer much. I've only used the latter class of devices, so I don't have any experience with how well Meshtastic's store and forward works in practice.
wtallis··on BYOMesh – New LoRa mesh radio offers 100x the bandwidth
I suspect protocols like B.A.T.M.A.N. would consume all available LoRa airtime just with neighbor discovery. On Meshtastic, nodes announcing their presence more than once per hour is seen as unnecessarily talkative.

I also don't think it's possible to make Meshtastic into the kind of reliable and high-performance mesh network some users want without effectively dropping support for most battery-powered devices by requiring much more frequent transmissions to maintain connectivity and participation in the mesh.

wtallis··on BYOMesh – New LoRa mesh radio offers 100x the bandwidth
The issue you linked to is about MeshCore using channels that are too narrow. A mesh system claiming to offer 100x bandwidth is probably not violating regulations in that particular way.
wtallis··on Ti-84 Evo
I use emu48 on my phone emulating the HP-50g, which was almost exactly the same size as the phone so my muscle memory somewhat carries over (minus the tactile feedback of a real keyboard). I still have the physical calculator on my desk at home, with no batteries in it so it's only usable within reach of its USB cable.

Anything that goes beyond what that calculator's UI can reasonably handle is going to end up in a Jupyter notebook or something like that.

wtallis··on You can beat the binary search
And you always start off knowing the total length of the array, and the width of the datatype.

Actually deciding what to do with that information without incurring a bunch more cache misses in the process may be tricky.

wtallis··on You can beat the binary search
Yes, this can be seen as unrolling the loop a bit. It improves performance not by significantly reducing the number of instructions or memory reads, but by relaxing the dependencies between operations so that it doesn't have to be executed purely serially. You could also look at it as akin to speculatively executing both sides of the branch.
wtallis··on Online age verification is the hill to die on
There are so many caveats and limitations that bringing it up in this context is downright dishonest. The most you could fairly say is that some of the philosophy driving some of the meshtastic developers is what you want to see applied to the development of an internet-scale network (which in reality would have less technology in common with meshtastic than with the current internet).
wtallis··on Online age verification is the hill to die on
Did you intend to link to Meshtastic as an example of how not to achieve your goals? Because it definitely isn't capable of scaling up to anything like the whole internet, and the project struggles to agree on any goals they want to reliably achieve.
wtallis··on Networking changes coming in macOS 27
That's when Samba gained official easy to use support for being used with Time Machine. I'm pretty sure it was possible long before then, IIRC by changing a setting on the Mac to allow selecting unsupported network volumes.

I don't recall when I stopped running netatalk on my NAS and switched to pure Samba, but I think it was before 2018.

wtallis··on Networking changes coming in macOS 27
Where "new" in this case could be a NAS running Samba from 2011? Samba added official support for Time Machine much later, but I think it was possible on earlier versions with some extra steps.
wtallis··on Asahi Linux Progress Linux 7.0
That's one of the several forms of support under discussion, under the specious claim that it would become the expected level of support as soon as Apple declared any level of support for Linux. But as the comments you're refusing to understand have explained, Apple could meaningfully "support" Linux in the form of providing hardware documentation, without making any promises to help any customers troubleshoot Linux running on that hardware.
wtallis··on SWE-bench Verified no longer measures frontier coding capabilities
Intel's done it again recently, this time targeting Geekbench: https://www.intel.com/content/www/us/en/support/articles/000...

Both that and the SPEC compiler shenanigans are cheating by changing the test, not just over-specializing the product being benchmarked.

wtallis··on SWE-bench Verified no longer measures frontier coding capabilities
That's a slightly different problem. There's no thing as saturation for a performance benchmark like SPEC; we can always conceive of a faster processor (even if we don't know how to build one). Saturation is the problem that once you are at (or near) 100% pass rate on a test of pass/fail questions, there's no room for the score to keep going up and the test has lost any power to discriminate between competing options.

However, both kinds of tests are susceptible to over-fitting: an LLM can be trained on the exact test questions, and a CPU can be designed with eg. branch predictors and cache sizes tuned specifically to handle a particular benchmark or workload.

wtallis··on SWE-bench Verified no longer measures frontier coding capabilities
I'm not sure people are really trying to interpret this kind of benchmark as being accurate in gauging the magnitude of improvement. It seems pretty obvious that doubling your score on some benchmark where 100% means "correctly answered all of these specific problems" doesn't translate directly to performing twice as well on all problems. I think what people want from these benchmarks—and what they do get to some extent—is answering the question of "is model A better than model B", especially the subset of "is this local model better than last year's frontier online model".

The marketing departments touting each model do want to claim superiority on the basis of slivers of percentage points, and that's probably always a stronger claim than the test results can reasonably support. And the benchmarks are obviously susceptible to cheating and overfitting. But when the scores aren't saturated and do show a big discrepancy, that kind of result usually seems to align with what people report from actually trying to use the models in the relevant problem space.

wtallis··on Asahi Linux Progress Linux 7.0
As used by commercial hardware and software vendors, "support" can mean anything from "we'll come fix it for you when it breaks, or your money back" to merely "theoretically, it should work, and we won't get in the way of you trying". Likewise, "unsupported" can mean anything from "don't complain to us if it doesn't work" to "we're going to spend significant engineering effort to prevent it from working".

A stance of "here's some hardware documentation, implement the drivers yourself" definitely falls within that spectrum of "support", and is the kind of "support" for Linux that some hardware vendors have in the past been lauded for, eg. when AMD started documenting their GPUs.

That level of "support" from Apple for running Linux bare-metal on Apple Silicon would be an improvement from the status quo, and in practice would probably be sufficient to get good drivers written and upstreamed in short order, given how much interest there is in running Linux on these devices.

wtallis··on Asahi Linux Progress Linux 7.0
There's nothing rude about it; the Nuvia CPU core is pretty much the entire selling point of the Snapdragon X Elite product family. Everything else on those chips is underwhelming. But the provenance of the CPU core is really irrelevant to the question of Linux support, which is gated by driver support for the rest of the SoC, which didn't come from Nuvia. So focusing on the Nuvia aspect is a bit of a red herring.
wtallis··on What the FCC router ban means for FOSS
See https://openwrt.org/releases/25.12/notes-25.12.0 and https://openwrt.org/docs/guide-user/installation/attended.sy...

They have the tools and infrastructure for assembling custom firmware images on-demand, and have recently added it to the default images, so they must feel like their infrastructure is ready for significantly increased demand.

← PreviousPage 7 of 34Next →