milch

With the Chinese models now being 3T too I don't think they have to. Speed/cost wise this already doesn't seem like a viable model. The biggest threat for the economy IMO is if they challenge the whole Nvidia/Datacenter infrastructure money side, which this pretty much doesn't. More likely they will ban running LLMs locally in some way

And somehow just because the store says it's 48gpbs doesn't mean that's true either... I have a cable that is explicitly marked as 48gbps, 8k60, 4k120 compatible and yet it frequently ends up dropping the connection. Swapped to a different 48gbps cable and it works fine. Funny enough when I display the negotiated FRL speed it's only at like 27gbps, so should have plenty of headroom.

At some point my older Mac charger that I was using for traveling died, so as a replacement I got a few of these styles of adapters: https://www.amazon.com/dp/B08YS7QHM9 (not the exact one I have - but you get the idea). I pair it with one of these: https://www.anker.com/products/a2687-anker-prime-charger-160... which is tiny for the amount of power it can output, which has been enough to cover both me and my wife while we travel. Many US plug chargers allow their pins to fold in, including this one, which makes for a fantastically compact setup when paired with that style of adapter.

Depending on where we are going and what we are doing, I also bring this thing: https://www.amazon.com/dp/B08YDPB8RZ which is kind of an extension cord + USB charger + all-in-one converter as well. The plug is at a right angle and very flat, so it can fit behind furniture and such that straight cables can't.

I built myself a new PC recently and jumped straight to installing Linux on it. Unfortunately the experience isn't quite as plug and play there either... In terms of everything I had to do:

- Disable iGPU in BIOS because on a cold boot the system would "latch on" to the iGPU (iGPU was GPU 0 and dGPU was GPU 1, which some programs I guess interpret as "GPU 0 == primary GPU"). Everything is fine on a warm reboot where it assigns dGPU to the 0 slot and iGPU to the 1 slot

- Get a special dongle to get 4k120Hz support over HDMI because the driver doesn't support it. The dongle should also enable VRR but needs multiple kernel patches to do that. Applying the patches causes varying levels of graphical glitches

- On waking from S3 the HDMI audio is dead. I tried various kernel arguments related to power management that various people had reported working for them but ultimately none of them ended up working for me so this one is still broken. Gave up on HDMI audio out and used an SPDIF cable instead (my AVR has been a magical problem solver here but most people aren't going to have a separate device to do that for them)

- On waking from S3 the wifi doesn't automatically reconnect. Solved with a systemd unit that just toggles it off and back on

- The audio out kept changing away from the one I set. As far as I can tell there's no easy way to set the "default" like you can in macOS or Windows, so I had to fudge the priorities of the built in devices a bit with a Wireplumber config

While debugging all the issues I've been bouncing back and forth between various kernel versions and patches. The fact that you can do that is awesome, but the fact that you have to do it to fix and debug issues is unfortunate. Just as an example, while debugging the audio problems I found a ticket that looked like it could be related on the amdgpu driver issue tracker where someone reported a similar problem, and the response from AMD was "please bisect the kernel"... That's just not something that most people are going to be willing or able to do. AMD GPUs are even supposed to be the well supported platform compared to Nvidia. I will say though, this whole debacle with the dongle sure had me wishing I could just stick an opaque-but-vendor-supported blob in like you do with Nvidia and get HDMI 2.1 working out of the box.

At least on Windows, IME Microsoft ironed out a lot of these types of driver issues in the Windows 7/10 era so that things like sleep just tended to work. It's been a while since you had to track down the right driver for your hardware manually. They just kept going downhill from that point forward with boneheaded decisions like sticking React Native or whatever in the start menu and forcing shovelware down people's throats. Even though I was having all these problems with Linux I personally won't go back to Windows, but I'll think twice about recommending it to my non-tech friends and family.

I think there's a world of a difference between "you wrote some code and it sucks" to "you wrote some code and it sucks because you have beginner energy and live in a fever dream". The former can be an objective statement without attribution to personal failings, while the latter... not really. If I put the latter as a PR comment to my intern in the morning I would expect a meeting invite from HR by the afternoon at the latest

How historical? Compilers that have translated from a source language into C and have left the C-to-bytecode translation to another compiler have been around for a long time, as have compilers that translated from a source language to Assembly

Very interesting article. Would've been fun to see the comparison between native interpreter & JIT-on-WASM on iOS as well

I'm never mean but sometimes when Claude does something especially boneheaded I just hit it with a single "bruh". That usually triggers an automatic "You're absolutely right -- I shouldn't have X and followed your directions more closely, let me revert and do Y instead"

Looking at their blog about it, I don't think Evolver has any claim to be honest. They make such generic claims as that session history was copied because they implemented their history in some `.jsonl` files while Hermes has a session history using sqlite FTS. Evolver better have some attribution crediting Anthropic for their session history feature, because as far as I can tell they've had that implemented since early 2025! Using .jsonl files no less, coincidentally the same way that Evolver decided to implement session history.

I was expecting them to show code samples or something, but to me it looks more like these were two projects that ended up coming up with similar looking features, even with different implementations of these features. I don't see how Nous Research owes them any explanation, especially since at least from that linked issue they were coming in really hot with accusations of plagiarism. If I owned a large OSS project that amassed issues/discussions/PRs at a rate of ~4k per month I would probably also treat this kind of thing as spam.

The limbs you can deal with, at least. I've done century rides in 100°F+ weather wearing UPF 50 sleeves the whole way... it's actually quite nice because you dip them in water, and they end up staying saturated for a long time. Together with the airflow from the ride it keeps you cool as well as protecting from the sun

IIRC Anthropic models haven't all been available on day 1, so it does feel like a deliberate choice, especially since they are partners/part-owners like you say. No one would have bat an eye about some corporate PR thing "blah blah mythos is too powerful and too intelligent and so we've decided to focus our capacity on Opus 4.6/4.7/4.8 for now until we have the proper safeguards in place blah blah"

Why would they have launched Fable on Bedrock if they knew they were going to be shutting it down a day later?

Instead NHTSA still allows automatic high beams, which blast you instead in the US... Or just regular old drivers that don't realize/care. Even as a pedestrian, I constantly see people driving through my neighborhood with high beams on... and these are side streets with a 20mph/30kph speed limit. Anecdotally at least people in Europe are a lot better about only turning high beams on outside of city limits

Look at the very first point on the archived vote leave page: https://web.archive.org/web/20160620214900/http://www.votele...

Sure, you could argue that they didn't mean it would be positive for the economy to save that money, but "we will save 350M/week" is what's on the buses and their website. Even if we assume the average voter clicks through here and reads everything point by point, or goes onto the website in the first place rather than by the headline, it is at the very least heavily implied... Otherwise what is the argument?

Weirdly I find myself writing more code lately, not less. A lot of it is AI assisted but a lot of it also isn't. I find that where previously the management chain wanted things meticulously planned out, and we had to have 17 meetings with different stakeholders and principal engineers to approve the system design.

What hardware do you need to run MiniMax M2.7 230B locally?

In fact when my company awards me stock the government takes half anyway!

The difference is that taxis would either give you a call that they were here or they'd just wait for. They don't care either way if you show up or not because the meter is running. The Uber is gone if you don't show up in 5 minutes. That is if you are lucky and the driver didn't mark themselves as "here" when they were 2 blocks away, which seems to be the norm here now.

Exactly. Sure, I can go read a book and learn about Rails. If I wanted to be a Rails specialist I probably would, but I don't want to be a Rails specialist. And like you said, other languages can do this too, but I haven't really found it to be too much of an issue. I remember one time where I had to break out whatever the rustc flag is that just produces the code with expanded macros, but that's it.

SwitchBot is one of those major manufacturers: https://www.reddit.com/r/TrySwitchBot/comments/1j8ogt8/matte...

I don't think I've ever been unhappier than working on a Rails project. See a bug on the site - something is rendering incorrectly. grep to find the view. Great, there's a method that is being called to render the buggy section. Grep for the method name - 0 hits. Amazing, it's something that is synthesized somewhere and I have no idea where. Guess I'll stop what I'm doing and read docs for an hour. If you do nothing but use Rails all day, sure, but the whole convention over configuration thing is such a huge anti pattern to me.

The control interface is local, not cloud-based

Unfortunately you see quite a few manufacturers implement Matter in a way where they require a proprietary hub, and the proprietary hub requires a constant internet connection to work...

I've worked at Amazon since 2018 and they've always talked about the software engineers being fungible during my tenure there. Technically everyone is supposed to be able to do everything, but in practice it's a huge headache if you want to hire for a more specialized role. They started creating some, like Frontend Engineer and Embedded Systems Engineer, but in practice these are still extremely broad

Apple App Store Connect has this issue for example. If you remove or downgrade someone from the store they retain access for however long... and they keep the ability to do destructive actions like remove or change the permissions of other users, at least they used to until a year or so ago when I last checked. They closed my security report as "by design" btw.

To be fair here if you look at their "supporting evidence" most of it is pointing at markdown formatting. Some of the other things it points to are also just common in human writing, though the overuse leans more towards coming from an LLM, e.g. two instances of groups of three back to back with inconsistent use of the Oxford comma:

If you're building a web app, a mobile app, or a first-party API: JWT is the wrong default and you should stop reaching for it. A row in Postgres with a bearer token in front of it is faster, simpler and strictly more secure.

I skimmed the study rather than the article about it and I don't see them define it at all. They just had a machine learning model take accelerometer data and classified it into "sleep, sedentary behaviour, light physical activity and MVPA". Whether any form of walking counts as light or moderate in this classification is really anyone's guess

I played around with local LLMs on my M4 Max 64GB this weekend and this is exactly what I found. I put Opus 4.7 "head to head" on the same task as Qwen 3.6 and a few other local models. The 35B did not perform well IME - it needed a lot of handholding and even then the final result did not work until a few more tweaks, while Claude one shot the task. The 27B was much better and also one shot the task, but took about ~55min as opposed to about ~15min for Claude. The 27B is probably something that I could happily run for many use cases if I had some faster hardware... the main problem there seems to be that at larger context sizes, prompt decoding can take several minutes.

IME there is a large difference in quality in what is available at the super market. Sure I can do a once a year road trip to Monterey. The average organic heirloom tomato at Whole Foods or Trader Joe's is worse than the average organic heirloom tomato at Spar

libghostty makes it pretty easy to do. I spent about two weeks setting something up until it was advanced enough to daily drive. I wanted to have a modal workflow similar to vim or tmux copy mode, but without having the overhead of using tmux... that's probably a lot more complicated than "I want Ghostty but with $X tweak". You can poke around in the repo to get a feel for what's involved if you want: https://github.com/milch/mistty

I live in Seattle and anecdotally I have seen the number of people running red lights absolutely explode in the last two years. Literally from seeing once or twice pre-COVID to at least one a day. This is not an exaggeration, there's a particular light on my commute that I see at least one driver run per day. My theory is that in an effort to make the intersection safer they adjusted the lights so now there's a period where cars all have a red light while pedestrians are crossing. Meanwhile a certain segment of the population sees all cars in the intersection stopped and decides to slam it. It's a recipe for disaster given there's a middle school down the road from that light...