Haven't touched it in a few years but Bevy was a really nice breath of fresh air when I dipped my toes back into hobby game dev. I wasn't used to being able to create a game almost purely as code without having to touch a WYSIWYG editor.
For things of global relevance, I keep a git repo in ~/.skills. It has a script that syncs skills across agent harnesses when run. If I think of an interesting skill that seems like it'd be broadly useful, I put it in this repo. If I want to use someone else's skill, I read it with my eyeballs, decide whether I like it, and then copy and paste it into the repo.
This is a big reason why I feel like even though LLMs are _effectively_ AGI in some regard, they also are a hack around what most people figured AGI would look like before the advent of LLMs. Humans can do metacognition, output multimodally at the same time (verbal _and_ physical intelligence go together to produce an expressive face while one talks), have a good sense for what they do and don't know, continuously take in and respond to the world around them in a (mostly) uninterrupted fashion without "turns", learn knew knowledge and retain it for their whole lives, etc. When you reduce a human to a text generator, yes obviously SOTA LLMs perform way better, but rather than invent something that can operate as an always-running "being", we've grafted a harness around an intelligence that is bound purely to speak only when spoken to. Maybe organic intelligence is already that, playing out at a super high refresh rate, but I don't know.
Current AI is arguably much more capable of multimodal output than humans. It can produce an incredibly vast variety of audio, images, and video. Humans are limited to producing the sounds we can make with meatflaps in our throats, and contorting various parts of our bodies to produce crude symbols and shapes.
(Very capable!) Embodiment, persistent operation and continuous learning are indeed things that still set us apart from AI. None of those are fundamentally difficult to solve, though.
More importantly, none of those are particularly relevant for being "intelligent": If a criminal threatened to kill your family unless you solve some difficult problem that requires only intelligence and you could choose any single person, animal, or AI to help you with it, which would you choose? Be honest.
The thing I trust the most to solve tricky problems reliably is a specific very skilled programmer I've known for twenty years.
I wouldn't say he never makes mistakes, but his success rate is a damn sight better than any LLM I've ever interacted with (and I drive Opus daily, due to corporate demands to use LLMs).
Same partial answer as I gave your sibling commenter:
"OK, so the task ends up being to write a 40-page analysis of the result of a specific experiment in quantum chromodynamics, to be finished within 2 minutes. Does your choice for a human work, or would you have rather chosen any of the frontier AI assistants? Be very, very honest. Remember that the lives of your family are on the line and nobody will judge you for a lack of allegiance to humans."
This task is intentionally designed to ensure a human cannot do it.
The initial scenario is utterly, insanely absurd to begin with, but I tried to go along in good faith and gave you the true answer.
The result was a bad-faith rhetorical trap, so I'm done with this thread.
In another attempt at good faith, as part of bowing out I will add some actual response to your anti-useful cheap rhetorical trap:
I do not trust LLMs to get things right in high-stakes scenarios. I have seen the current models spit out falsehoods and errors regularly in the handful of fields I have expertise in, and have no reason to think they would do otherwise outside my expertise.
The scenario you describe is an absurd fiction, and no human making the absurd threat could evaluate the paper in less than hours (realistically even an expert would need days, and a nonexpert could not do it at all [short of becoming an expert]).
So, there's no point trusting a bullshit machine to save my family - it might very well get them killed, and whether it was right or not, what would actually matter would not be its correctness, but what the presumable bullshit machine evaluating my offered input spits out.
So, the best move I could realistically make would be to put a stab at prompt injection into the input.
For that job, I probably would actually prefer aforementioned programmer over any other option, come to think of it - I suspect he'd have better success than even another model (especially considering the safeguards the models no doubt have to try to keep users from using the models to inject other models).
Again - I'm disappointed in your worthless rhetorical cheap shot.
I suspect you'll have much better success convincing people LLMs are intelligent if you engage in good faith, listen to their perspective, and address their actual thoughts, instead of devising the sort of inanity that comes out of high school debate clubs, where people literally want to score points instead of find truth.
> This task is intentionally designed to ensure a human cannot do it.
It is one of a vast array of things the hypothetical kidnapper could come up with, some of which humans I agree will do better at (currently) and some of which AI will do better at. We clearly agree that in that array there is at least one task that a frontier AI would be better at than any human you could pick.
It is a thought experiment, so there is nothing fundamentally wrong with it being extreme or unrealistic (thought experiments very often are), but for the sake of goodwill let's 'weaken' it a bit: the AI or human always has an hour to come up with the answer, the question and answer are in their preferred language, and the answer fits on four pages. You can't help them, though; The criminal 'prompts' them. They can use the internet as an informational resource, but they can't communicate/ask for help/post anything (with the spirit of this being: no loophole in letting somebody else do the task or parts of it for them). And of course all subject matter of all complexity is fair game (including but not limited to quantum chromodynamics experiments).
Given that situation, do you think the programmer you mentioned would be more successful than a frontier AI in more than 50% of the possible intelligence tasks?
Edit, addendum: Please, if you can, also let said programmer read this thread and give his opinion on it. It sounds like he would have interesting things to say on this.
Before "AI," humans have created a vast array of "multimodal output" (computer art, instruments, dance, architecture, etc.). Why are you giving the AI a harness and a plethora of tools and not the human in this comparison? Without these, the LLM too would be utterly useless.
Also, that situation is extremely contrived. If a criminal threatened to kill me if I misspelled a word, I would choose a dictionary. If a criminal challenged me to predict a next token, I'd choose the LLM. For all real precarious dangerous situations, I would obviously choose a human. Like immagine the hilarity (or tragedy) that would pursuit if ChatGPT tried to handle a hostage situation or a plane hijacking.
and those meatflaps are normally called vocal folds/cords btw
> Why are you giving the AI a harness and a plethora of tools and not the human in this comparison?
I am not. Multimodal models generate that output directly, without tools. Which 'tools' does AI use to generate all those images, songs, and videos do you think?
> Also, that situation is extremely contrived. If a criminal threatened to kill me if I misspelled a word, I would choose a dictionary.
Of course it is contrived, it is a thought experiment. Does not make it less valid. It is essential that you don't know what task it is going to be, just that it is a task requiring a lot of intelligence. This way question dodging loopholes like "I'd choose a dictionary" are impossible (and people will always try to find some cheesy exit rather than facing reality). You have to commit to something or somebody that has broad and general intelligence; you do not have the luxury of choosing the perfect tool for a very narrow task.
> For all real precarious dangerous situations, I would obviously choose a human.
OK, so the task ends up being to write a 40-page analysis of the result of a specific experiment in quantum chromodynamics, to be finished within 2 minutes. Does your choice for a human work, or would you have rather chosen any of the frontier AI assistants? Be very, very honest. Remember that the lives of your family are on the line and nobody will judge you for a lack of allegiance to humans.
Again, don't go for shitty loopholes. Engage with the thought experiment in good faith and thus as it is stated, not some conveniently distorted version of it.
> and those meatflaps are normally called vocal folds/cords btw
What? Next you're going to tell me that meatspinner is also not the name for the human male reproductive organ.. Maybe I need to get a refund on my Temu Gray's Anatomy.
Luna max is suitable for like 90% of the kinds of code changes I want to make. I only find myself actually reaching for a Sol or Fable tier model if the problem is very complex. If you're willing to build the guardrails and do some extra planning, Luna is very capable.
Everyone has their own distinct mix of jargon and uses some words more or less than others, but now it's as if you have 1000 coworkers who are all the same person, and talk the exact same way, and it just so happens that their writing is grating to read. It also doesn't stop when you log off for the day. The same guy follows you home in ad copy, message board comments, spam, and in some cases even the correspondence of your loved ones.
I have immense respect for the incredible work that the Asahi team does, but at the rate that Intel and AMD have been catching up on power efficiency I wonder whether anyone will really _have_ to run Linux on M series laptops. There was a clear case for it back when Apple silicon absolutely dominated anything else you could buy but with Panther Lake I'm not so sure that Apple's lead will continue to be big enough to justify needing to wait several years for full Linux hardware compatibility to be reverse engineered.
Find me another laptop that has everything, even if it's more expensive ! You can't, some have keyboard as good, some it's the display. In best case you got 2 or 3 things but then it goes downhill on the other, like the trackpad.
This is exactly why I run asahi. There really is not a laptop that comes close to the hardware quality of an apple laptop. I have even sacrificed x86 at the altar which is a much bigger step than battery life imo.
I’ve been a big Lenovo fan, settled on the X1 carbon. I’ve had three of them, G5, G6 and G9. The USB-C port broke on both G6 and G9 models. Apparently it’s a common problem. Unfortunately the ports are soldered directly on the motherboard making it difficult to fix(should be a small separate board for easy replacement). I was going to desolder and replace the ports, but haven’t gotten around to it.
The only thing that’s egregious is the batteries.
They tape the batteries to the case, and to remove them you have to do this complex twirling motion with tweezers to pull the tape out from under the battery. If the tape snaps during this then the documentation basically says get a new case.
Apple saves itself a lot of the trouble with superior build quality. With Samsung devices (so no cheap crap), USB ports last me about three years of heavy use.
With my old iPhone 12, 2019 16-inch MBP or my 2022 MBA? Not a single issue. The housing is simply constructed so tight and, being made of solid metal, rigid that the housing bears the torsion load and not the tiny PCB solder pads.
The only problem I have with my new USB-C iPhone 15 is that it is hard to clean pocket lint out of the socket, compared to the old Lightning connector.
> With Samsung devices (so no cheap crap), USB ports last me about three years of heavy use.
What are you doing to these machines?! I work in the safety and security biz, including in the field, and my trusty HP EliteBook 2730p is still rocking along on its first USB ports. That machine is over 15 bloody years old!
That is a convertible laptop that meets MIL-STD-810F standards on vibration resistance, it’s not really a fair comparison. Also, the change from dedicated charge ports to USB-C charging increased USB port usage considerably.
> Also, the change from dedicated charge ports to USB-C charging increased USB port usage considerably.
I never needed one, but a good USB-C adapter (or maybe even small dock) with power passthrough should solve this. Burning through whole machines because a single USB-C port cakes out after just three years is an awful lot of waste.
Mostly, keeping them in my pocket with a USB cable attached to a power bank. Places an awful lot of torsion force on the connector. I tend to go through 2-3 charges a day.
Another issue is that I have cats that like to play with everything being too close to a table corner.
I try to buy good hardware for practical reasons. I insist on repairability for emotional reasons around ownership and independence. You do not have to value the same things as me, of course.
Exactly. As a lifelong linux user, I bought a Neo on release for my SO and was blown away by the quality at that price point.
I've had high-end laptops from all the big vendors, and even at the highest price tier, other vendors don't deliver such build quality. I hate how all non-Apple laptops feel. The case, the hinge, the materials, everything. If I coud reliably use linux on Apple hardware, I'd never buy another laptop brand again.
1. Trackpads. I hate 'em. Keyboard-integrated trackpoints and/or trackballs is what I need. Detachables with a trackpad are acceptable as one can bring a Bluetooth substitute (that's how I do it on my HP x2 G4); on a non-detachable only a trackpoint, in addition to the trackpad, is acceptable (on my HP 2730p or my workplace ThinkPads I only use the trackpoint, the trackpad is turned off).
2. Form factor. Not counting their "castrated" tablets, Apple doesn't build 2-in-1 or 3-in-1 detachables (the best general-purpose mobile computing form factor) or convertibles (T-hinge models represent the second best form factor). The last orthodox-chassis laptop purchase I made took place in 1999. Outside of a vintage or retro machine I might entertain (which itself loses out to wedge-style keyboard computer cases), that form factor is dead to me.
3. Apple is not the only outfit that ships with good displays. Et cetera.
> wait several years for full Linux hardware compatibility
I'm not convinced it's only "several years".
Asahi was launched in January 2021. So 5 and a half years ago. Asahi, as far as I can tell, supports most of M1 and M2 features, so generations launched in November 2020 (6 years ago) and June 2022 (4 years ago).
I would be very surprised if a decade from now the story won't still be the same thing: 90% of features for systems half a decade old with software package gaps in important areas and everything slightly less optimized than MacOS.
Asahi is doing incredible work but it's still 5 people in a garage vs the huge corporation.
This can be a good tradeoff. An M1 Ultra still runs circles around 90% of consumer hardware sold today, and with Linux it has easily another 10+ years. Now support for M3 is coming.
> Asahi is doing incredible work but it's still 5 people in a garage vs the huge corporation.
Not sure why the "vs" scenario, this is a project that makes no promises. And I think they do have some support from Fedora / Red Hat fwiw.
That's a good tradeoff for durable hardware with available spare parts. That's not a good tradeoff for hardware that was designed so it cannot be repaired, such as newer Macs.
I'm still hoping one of these mega AI LLMs will figure out how to install Linux on the many old iPads that become nearly useless after Apple stops updating them and you can't install new apps.
Oh definitely, there's great value in it for those who already own Macbooks for one reason or another. I just personally have found myself shying away from buying a Macbook multiple times over the years because the Asahi Linux experience was really really rough at the time & probably still is for the generations I would be interested in buying.
Yea, this project doesn't just benefit someone who goes out and buys a new Apple laptop. It's going to be most beneficial years from now, after Apple deliberately obsoletes today's hardware, stops updating them with software, and tries to turn them into e-waste.
It depends on how long Apple want to support the devices, if the support matches the longevity of the devices maybe nobody really needs Asahi. I believe all Intel Macs have received their final macOS version now so it is almost time for Apple to make that decision for the M1 generation. I think an M1 Pro, Max or Ultra (or even an M1 with 16GB of RAM) still has many good years left.
OS LTS support barely matters when planned obsolescence is baked in at the hardware level. Wearing parts (eg. SSD) are soldered on newer macs and there's intensive efforts to make sure even if you have the tools you can't repair it.
Thanks, that's a fair data point. Although i'm sure it depends on actual user workloads.
I'm not saying that I support closed-source firmware, locked bootloaders, or even that replacing SSDs is the one true path to sustainability.
I believe it's a whole field and we shouldn't tolerate compromises on any level. It should be 100% illegal in any civilized country to sell the kind of lame hardware we have today. It should be 100% repairable, 100% open bootloader & free firmware, 100% public tech docs/datasheets. Original manufacturer OS support is not insignificant, but i believe it's only significant because we have so little control/understanding of the hardware/software stack.
You have trade offs though. Both with added effort during the development but also just physical limits with the amount of additional losses you have over clicked vs soldered connections
That's true. I can't say i'm 100% against soldering parts, that would be silly. What concerns me is that undocumented parts that are unavailable to buy as spare parts are soldered and have firmware that enforces pairing with a specific other part.
It would be reasonable if all components on a motherboard had to be sold as spare parts by law (not just the entire board that costs the same as the entire device, that's malicious compliance) and all spec sheets were made available. That might just make the economy of a repair shop work again.
Sure, and in a zero marginal cost world that would be fine and dandy. In practice, it would mean significantly increased development and documentation costs for a low percentage increase in sold units.
Selling spare parts was worth when things broke often and labor was cheap. We don’t live in that world anymore. That means a broken device will no linger get fixed. But on the plus side there are far fewer broken devices. I’m pretty sure that has the same problem shape as “there is no glory in prevention”
The thing about SSD failures is that in normal use, you can't wear one out. But in perfectly normal use, some of the units just die after N years (for very small number N in some cases). They don't warn. They don't show increased fault or wear counters. They just die and completely disappear from the bus and stop accepting any commands. There is no recovery and, I repeat, no warning.
That's why you have backups and drives are easy to replace, though reinstall/restore can be tedious.
Every piece of Apple hardware I own has outlasted its software support by at least 2X. Or, as I'd rather put it: The software support has been prematurely cut off at best at 50% of devices' lifetimes.
Yeah, but there was a time when running Linux on a laptop was one of the most complex ways of running Linux. Now I just expect everything to work, especially if I buy from Lenovo, Dell, or HP.
There was a time when Wi-Fi, or Wayland, or XComposite, or Linux gaming, or just accelerated graphics at all or, or, or... People get excited, they work on it, other people get excited and mess about, and eventually it just works.
I would love to one day pick up a decommissioned M1 Ultra and use my OS of choice on it, and I'm grateful for the people who are doing the hard thing now.
The main power advantage of the apple chips is when the system is mostly idle - under mid to high load power to watt has been comparable for a few years already (at least AMD and Apple, intel was way behind - but I guess they caught up recently). The main reason behind my M3 MBP back then wasn't so much power efficiency - but that I can get it with 96GB of RAM.
It's nice that I can go a full office day without charging - the previous x86 notebook couldn't, but a decent one from the M3 generation probably could as well. But also if I put proper load an that thing the Macbook battery is done in less than 2 hours.
I don't see how? it is a 50Wh battery. Even at 6W that is about 8h of usage? On what OS + workload on x86 are you seeing just 6W usage? Or is it that each day your usage is only 2 hours?
> And I bet you can soon get a ThinkPad with the same intel chip, that should have similar battery life.
Yes you can, speaking from experience. I have the T14 gen 7 with the 75 wh battery, and the 356H CPU, it only uses 4-6w/hour when coding with a lot of browser tabs open, a headless VM running, docker containers and so on, online meetings is about 6-8w/hour, this is under Linux, with default power setting from Opensuse.
It can easily last more than 10 working hours, I limit mine to 80% charge as it's more than enough for when I'm on the go.
I have a M1 Pro from work and it surely won't last that while working. Throw some video calls at it and working outside (so, screen at the brightest level) and at the end of the work day I have to plug it in
With load on it and a bright room (= display reasonably bright as well) my M3 also doesn't always last 8h. But it gets very close. Intel machines are more like 3h in that case.
It’s more about not having to drag a cable + brick around everywhere and not needing to care about the location of the nearest power outlet, potentially for multiple days depending on usage patterns. You can instead leave the charging bits at home or the hotel and charge overnight.
This also means that battery cycles accumulate more slowly and the battery’s lifetime is stretched further.
Also, generally more efficient chips don’t get hot as easily which makes the laptop more comfortable to use and prevents fans from becoming audible unless you’re really pushing the system.
Probably nobody and I remember when a phone battery lasted one week: nevertheless sometimes we forgot to charge and we run out of battery. So a long lasting battery is good, too long lasting is a waste, which maybe it's your point.
However nowadays it could mean running some agents all day long (remote models) and plug in to power only at the end of the day.
i snatched a lenovo yoga slim aura with a core 5 226 for 650 euro. Its build quality is on par if not better than the colleague's macbook, it's pretty good for working, runs fedora natively, battery last at least 12 hours unless i'm playing games.
On the battery life front, both Intel and AMD have caught up. Apple greatly inflates their battery numbers; to get 18 hours, you need the lowest spec model and you need to basically not use it. Chromium or Firefox open? Bluetooth active? Yeah, you’re looking at 12 hours, not 20.
12 is still good, but intel as of lunar lake (2 generations ago!) matched that. Now, the performance is still behind, particularly single core. But the efficiency is there, and these are standard x86 chips.
The efficiency doesn’t come from ARM, it comes from the process node and chip design. Apple uses the latest process nodes with an SOC design. When intel does the same, as they have done, they match the efficiency easily on x86.
The real gain is the SOC. When you shove the ram and chipset into the CPU you save massive battery.
> On the battery life front, both Intel and AMD have caught up.
Intel and AMD's cpus don't have batteries ;)
Their chips are approximately as efficient as Apple's chips from ~2 years ago, but that's not their fault, it's Apple getting the entire allocation of new chips for about 1.5 years.
> to get 18 hours, you need the lowest spec model and you need to basically not use it. Chromium or Firefox open? Bluetooth active? Yeah, you’re looking at 12 hours, not 20.
I work on compilers, often maxing out all the CPU cores, and have stopped carrying the power adapter altogether. I plug it in maybe once in three days.
> The efficiency doesn’t come from ARM, it comes from the process node and chip design. Apple uses the latest process nodes with an SOC design. When intel does the same, as they have done, they match the efficiency easily on x86. The real gain is the SOC. When you shove the ram and chipset into the CPU you save massive battery.
Yes, I agree. IDGAF about whether the battery lasts longer because of ARM or magic pixie dust.
Yes, I meant that Intel and AMD laptops have caught up to battery life, due to the chips.
I have both an M1 Pro MacBook Pro and an Asus lunar lake laptop. The M1 battery life is… good. But again not what Apple claims, if you actually use the computer. The lunar lake laptop has better battery life.
The base M1 is probably much better, but I still think these lofty 20 hour battery life claims from Apple are, well, lofty. It might be true if you playback local media with hardware acceleration and run in airplane mode. But as soon as you’re using all the antennas, the battery life goes down a lot. Bluetooth especially drains battery life.
correct! it's better on lenovos. All metal, trackpad has a proper click, keyboard doesn't shit itself. Screen is not retina, but then again i don't need that ridiculous resolution as the OS i'm using doesn't employ a toy UI with bubbles and rounded everything. Before this i had macbooks for 15 years, colleagues have this generation's macbooks at work, i like this one a lot more.
I'm a bit opponent of Apple's walled garden, but the build quality of this laptop is quite far away from a MacBook... Except the display, this OLED WXHQwhatever is stunning.
Source: I bought the same (with core i7, so it's performant, etc.).
> What do I need any higher specs for, other than battery life?
To run poorly made software. Despite Apple having the best hardware around, they still manage to slow down M1 devices on Sequoia. Somehow.
I think it’s a double-edged sword. Powerful hardware means powerful hardware. But it also means developers get lazy, and start writing poor quality software because it still runs OK on the latest and greatest.
Luckily Linux doesn’t have that problem, which is why you can easily add 10 years to a laptops lifespan just by using Linux.
As someone who does a lot of work with local LLM's, today's systems feel woefully under-powered. I'm looking forward to a future where my laptop has 10x the memory, 100x the memory bandwidth, and optimized cores to make inference workflows that currently take minutes or hours go down to seconds or milliseconds.
While we're at the point where traditional software is pretty much fast enough for all but extreme use-cases, with LLM's it feels like we're back to the days where you press compile and go have a coffee or chat to your colleague.
That's what I mean though. The comment I was responding to was talking about "forever laptops" - my point is there's plenty of room for new capabilities which will make current hardware obsolete. Just like how GPU's didn't exist at all, and became a standard part of computing.
And given how fast the hardware and software is evolving, I can easily imagine a future where we all have very capable models running on our own devices for an embedded intelligence layer that's doing most of the day-to-day tasks, and only have to outsource to a super-smart cloud model for specific things.
I do family tech support. Hardware refresh is driven by facebook adware js knobbery. Proof-of-work bot countermeasures, too, although i don't really resent that.
If you run noscript then yeah, why not keep rocking your pentium.
You can't run an egpu (amd or nvidia) on macos so that's one thing I could see switching for. I did just discover tinygrad though due to a hackernews comment, but not sure its performance.
Also Tahoes absolutely horrendous performance for the last year..
I'm not having that experience. So far each major model update has been at least slightly better than the last, in ways I've found useful. Can't say it's perfect, or able to do exactly what I want without a decent amount of instruction/implementation/docs, but it's been useful enough to keep paying for it.
Oh no the models are absolutely getting better, I'm just amazed that only 6 months ago I was using gpt-5.3-codex, and now I can use gpt-5.6-luna for similar results at like 1/15th the cost. Now 5.6-sol is being slashed by 50%? Amazing.
I like Go but a couple of these "advantages" wash out when you add the scale and typical usage patterns of agents.
| Go is Readable / Go is Maintainable
It's true that Go, as a low-magic language, tends to be very same-y looking across projects, which is incredible for being able to reliably understand your dependencies' source code. And its tooling is world-class. I love this about Go.
But in practice I've found that, working in a monorepo with multiple teams, contributors that don't have cross-team legibility as a priority will just write SO much more code. And with business logic, often the fact that I can read the code on a line-by-line level doesn't matter if I don't understand the wider context to know how something might effect spooky action at a distance.
Pre-agents, I witnessed a fast transition from a codebase that I could mostly hold in my head to one where large swathes of it had been written and rewritten until they were unrecognizable to me. Now we have agents and, since they are still mostly not good at software engineering in-the-large, the process of knowledge debt accumulation (and ofc tech debt accumulation) in a codebase accelerates tenfold without concerted effort in the other direction. Go being easy to read does not intrinsically help with that.
My problems reading code are understanding what the new vocabulary actually means, what it does in context, why it exists, etc. Doubly so when someone is pinging me to review a new 13000 line AI MR every 24h. Understanding an individual line because of some complex C++ feature, I don't remember it ever being an issue.
- An agent messed with my git config for some reason so I had to go back and fix the commit authorship
- I created a PR mostly based on work that someone else did, so I reauthored most of the commits to be in their name
reply