Not only that, but asking it to do a security vulnerability assessment of your own project is a very valid and important thing, and there is no way for it to know what is yours vs someone else's, so we just lose this capability?
The amount of FUD and notion that hardware depreciates in this manner is widely held. I blame Michael Burry of the Big Short who is perpetuating these lies to the investor community today.
There's a bunch of retro hardware which should make people pause and realize they're stupid to assume hardware slows down on average even 5% 20 years later (it's probably closer to 2% and I'm being generous).
HVAC/power delivery and generation are the major factors, and if you didn't skimp/get defective parts for this and replace failed moving parts (usually fans), your hardware is basically the same 20 years down the line as it was today.
Also using LLMs locally doesn't even induce sustained 100% GPU usage over significant periods of time for most real (agentic coding in OpenCode) use-cases.
There are tons of things that can start failing on hardware. I don't realistically expect some LLM usage to materially reduce the lifespan of the laptop, but running it 24/7 for AI usage makes me think that I'm more likely to get 3 years out of the device instead of 10.
Still trying to understand if a Macbook Pro M5 Max with 128GB is likely going to be able to run coding models well enough that I can cancel my Codex, or even go down to the $20/month plan.
A 128GiB MacBook Pro in Canada is what, north of CAD $11k after tax? That’s around USD $7k. At $20/month for a cloud AI subscription, you’re looking at almost 30 years of service for the same money.
How long do people realistically expect a laptop to stay competitive with SOTA local models? Especially in a space where model sizes, context windows, and inference requirements keep moving every year.
And even if the hardware lasts, the local experience usually doesn’t. A heavily quantized local model running at tolerable speeds on consumer hardware is still nowhere near frontier hosted models in reasoning, coding, multimodal capability, tool use, or reliability.
The economics just don’t make sense to me unless you specifically need offline inference, privacy guarantees, or low latency for a niche workflow. Otherwise you’re tying up $10k upfront to run an approximation of what you can already access through a subscription that continuously improves over time.
You could literally put the difference into index funds and probably cover the subscription indefinitely from the returns alone, even accounting for gradual price increases.
But what if you were going to buy a laptop anyway? Obviously you can't do anything with less than 64 GBytes these days, so the question is just whether you go for the jump to 128.
In the UK, it's currently an extra £800 to get a 128 GB vs the 64 GB equivalent. So that's more like 3 years of Claude - I think? - assuming current prices stay the same.
Or: you might just feel like £800 isn't an unjustifiable amount of money (one way or another), and tick the box, on the basis that it might just work out. As the saying goes, in for 459,900 pennies, in for £5,399...
I rebuilt the entire fastcomments moderation UI 2yrs ago with webstorm on my 16gb thinkpad. 64gb is nice but not needed. I wonder if every dev didn't use an M4 Pro if software wouldn't be so resource hungry...
I didn't click through the full UI to get the lead time or anything; I just looked at the options presented on the UK site. Maybe there's a stock of laptop types here that have all sold out elsewhere? Or maybe they were just teasing me, and I'd have been hit with a 6+ month delivery time if I'd gone all the way.
You are assuming I'd only get it for that. That would probably just be the straw that broke the camels back, but I'm already thinking about a purchase even if that doesn't work out.
You can buy a used GPU for under 400 dollars if you already have a desktop and run qwen 3.6 a3b and for a majority of frontier tasks get by just fine. Why do you need to spend 10k on a laptop, we are swimming in ewaste.
Have been using Qwen 3.6 27b recently along with various other models the last month and it is very capable for writing code at a level I haven't need to use a subscription for 95% of what I throw at it. Been using it to write extensions for Pi to expand tool kit without much fuss as one example. Is it as fast or SOTA? No, but you can't ignore how functional it is on hardware you own. Where it can begin to struggle is giving too open ended prompts or investigating complex technical issues. At that level its knowledge is not high enough to solve those problems on its own.
Ultimately it comes down to your entropy source. I always generate and insert in a loop for this reason, if there is a collision, I therefore handle that gracefully.
People are blown away when I tell them that, in the last 6 months, my job of coding has changed entirely, and that I now write very little code, but instead manage agents who write it. It is still engineering, and I still very much care what that code is, it's interfaces, how it interacts with the world, how it is tested, etc. etc., but it has taken me a while to get used to the idea of me not writing the code. I'm still not sure how I feel about it, although I am getting more done, and it has helped me keep better focus on "the big picture". That is tough to do when your day is spent in the weeds.
It isn't about the code (they probably don't even know what that is), they focus on the fact that my daily job tasks have changed entirely (iow, I used to do 'something', 'what' doesn't matter, and now I dictate to an AI to do it). Most people can't fathom this unless they change to a completely different profession/job.
I don't code (well, I write scripts for my own personal use and use Emacs), but I follow with interest these reports from software developers. Why? Because my profession is similar screen- and keyboard-based intellectual work, and what has come for devs will probably come for many other careers.
It's worth mentioning that one reason LLMs can clean up on coding tasks is because of the volumes of available data. Not only has the world produced copious volumes of code, they continue to produce copious volumes of code, and some code can even be generated synthetically (ie, not from an LLM, albeit at high cost).
Other domains are not like this. There will probably never be enough poetry out there to make an LLM do anything but be a poor imitation of a poet. This data is extremely hard to generate.
Urm, isn't that a terrible example? LLMs are better at poetry and rhyming then most people trying their best...
Also you're thinking with an extremely short time horizon there.
All jobs which are centered around computers will be impacted the same way programming is today in the medium term. It's just a question of time until the data is gathered and tooling is adjusted. Because every company that's currently employing people to do something with software will start to use that data as training material, and a few years later they'll be swallowed along.
So that means eg bookkeeping is still safe... I sincerely doubt it'll be like that in 2035. It'll probably still be somewhat safe in 2030 though
I suspect most industries aren't quiet filled with people like the author though, most people treat their job as... Well, a job.
This is a wild take. Of all things, you think a large language model will be stumped by poetry?
I would wager everything I had that an AI could already win a blind poetry contest, especially if any effort was put into fine-tuning it. Unbelievable effort has been put into optimizing them for coding given the financials; next to none for poetry.
Have they made a way to move those tiny icons in the lower left (aka "activity bar") to larger icons on the upper left like VsCode? As it stands, I can barely see them on my 4K screen and selecting them with a mouse cursor is like a pixel hunting contest. No go for me until they offer a way to change that. Beyond that it seems like a decent editor, but if I can't switch modes back and forth, that is a deal breaker.
UPDATE: Looks like they haven't yet, bummer, and doesn't seem to have much traction either. They redirect to discord, but AFAIK that doesn't have a way to make a feature request directly?
While larger icons might be ideal, note that (according to those linked discussions) those icons also have associated keyboard shortcuts, which are displayed when you hover over each icon. If you memorized those shortcuts, you would no longer have to “pixel hunt” to switch modes.
Interesting. I started working on this same idea a couple of years ago as a way to bypass CGo. Eventually I moved on to something else. Glad someone else is working on this. How does the generated Go performance compare to the original WASM performance?
That's going to depend on what you mean for "original Wasm performance".
What were you using to run Wasm instead of this?
I can compare with wazero, which I was previously using, and say performance stayed mostly in the same ballpark. Things that crossed the Go-to-Wasm boundary very often became much faster, things that stayed mostly in Wasm became slightly slower, as the wazero compiler is pretty good.
wasm2go also does not support SIMD, so if your Wasm module uses/benefits from SIMD, you'll notice.
"plansturbation" is a real industry, there are tons of successful YouTubers that sell millions of dollars in tutorials, courses, books, etc on how to setup your productivity harness
I'd like to use this, but I don't want to refactor all my services when they change the request/response types. Interested to know the timing of 1.x. It seems to be moving pretty fast atm - hopefully that momentum keeps going.
As I understand, protobuf has compatibility (it stores field ids), so new service can read request from older client, and vice versa, so you do not need to refactor anything. Also, it is made for long-range communications, and is inefficient for inter-process or inter-thread messaging.
reply