I built multiple iOS apps and went through two start up acquisitions with my M1 MBA as my primary computer, as a developer. And the neo is better than the M1 MBA. I edited my 30-45 min long 4k race videos in FCP on that air just fine.
Before I was a professional software developer, I used a scrawny second-hand laptop with a Norwegian keyboard (I'm not Norwegian) because that was what I could afford: https://i.imgur.com/1NRIZrg.jpeg
This was the computer I was developing PHP backends on + jQuery frontends, and where I published a bunch of projects that eventually led to me getting my first software development job, in a startup, and discovering HN pretty much my first day on the job :)
The actual hardware you use seems to me like it matters the least, when it comes to actually being able to do things.
But I'm planning to do a big jump: Soon I will switch to a 2012 Mac Mini as my primary linux server!
T420s has loose USB ports and the power socket is almost falling off, so I plan to replace it by a 5 years old T14 G2 in the coming months.
I can afford the latest MacBook, but I'd rather not generate more e-waste that there is, and more importantly I feel closer to my users, and my code is efficient and straight to the point.
My non-hobby laptop is an old cheap Dell from 5-6 years ago.
The best laptop I ever had was a maxed-out Thinkpad P7x, and it came with the most meaningless job ever.
I can only compare that job to the one at a unicorn that gave me the latest and greatest MacBook. Not only the job was meaningless, the whole industry made no sense to me.
i have a computer that benchmarks literally 10x faster and with 32x the amount of RAM, but i miss that little thing that helped me build my career from nothing
I still remember retiring that computer. The first thing I did when I got my Pentium IV chip a year later was download Macromedia Dreamweaver. Did me well.
After all, the actual server ran the code, I just needed text editors, terminal windows, and web browsers.
Running neovim on termux was fine. Developing elixir was no problem, the test suite took 5s on my phone, and takes 1s on my laptop. Rust and cargo compiling was slow enough that I didn't really enjoy it though.
Meant that I could just pack up instantly and have an agent do review workflows while I was out and about as well in my pocket, and didn't really notice a big battery hit.
Not sure the difference other than weight, but I wasn't carrying it day to day when i could leave it in my hotel room.
So maybe different meaning for everyone. For me it’s getting away from technology and into nature.
The main activity was still the traveling, hiking and enjoying some calm time. But instead of spending the usual downtime reading or something else, I had a blast coding and experimenting.
[1] https://www.zdnet.com/article/how-to-use-the-new-linux-termi...
(Maybe the fans sometimes sound like they're a jet engine taking off…)
Finally just put an order in for a new 16" MBP M5 Max with 48GB memory only because it looks like they're going to stop supporting the Intel stuff this year and no more software updates. It'll probably be obsolete in six months with the rate things are going, but I've been averaging seven years between upgrades so it should be good!
So, the m5 with 48gb of ram will be amazing.
Obviously the LLM inference is super heavy, but the actual work / task at hand is being executed on the device.
Is still handling the load well, though at times, fans get quite loud, especially with all the background processes and VM setups.
Hope to get a new MBP this year, as being on Intel means lots of software that won't run on it (ie, Codex app for example, won't run on Intel Macs)
I have a 2010 MacBook Air that I still use when traveling.
The battery is completely shot, but it works fine when plugged in. And if I'm on the road, I don't use my computer until I get to the hotel anyway. And even then, it's just fine for e-mail, browsing, and even Photoshop.
Am probably giving newish iPad and magnetic keyboard a spin on my next trip mostly to see how it goes.
I also was using an Intel MacBook Pro with 16GB at the time. Doing the same thing there was much smoother and snappier. On the whole, it actually made me want to just the laptop instead since it "felt" nicer. (This isn't measuring build times or anything like that, just snappiness of the OS.)
The worst corner they cut is no keyboard backlighting. That saves them what, $1 BoM per MacBook Neo? Especially because now they have to put up an entire new keyboard production line instead of just piggybacking off of the Air keyboard production line.
No 8GB, compressed or whatever was not enough for MacOS when M1 was released. Even for simple outlook, web browser, excel type of workflows.
After 3-4 hours of work, the window manager process itself is consuming gigabytes of memory. Not even considering any browser or electron apps.
My M1 Mac mini was choking up so much that I had to trade it in. That was back in 2021. Today apps are even more bloated.
8gb has ALWAYS been fine in Apple Silicon Mac OS. RAM usage on a fresh boot is a meaningless statistic (unused RAM is wasted RAM). And they're just plain capable!
I am jealous of my wife’s 13” M5 iPad Pro though, that oled screen is gorgeous, a wonder of modern engineering.
Well, the MacBook Air was also a lot more expensive than the Steam Deck?
Couls you please describe your dev process.
I setup a self hosted runner and then use that in my CI workflows. Then I disabled it from sleeping so it can clamshell forever and now it sits here in my living room silently workin' https://imgur.com/a/EaBICdo
Those apps don’t need every single byte of memory you see in Activity Monitor to be active in RAM all of the time. The OS swaps out unused parts to the very fast SSD. If you push it so far that active pages are constantly being swapped out as apps compete then you start to notice, but the threshold for that is a lot higher than HN comments seem to think.
Everything from apple to modern software is rotten to its core.
…in reply to someone who just said their experience is fine, and included details. If you just want to rant about Apple, have at it, but you’re going to have to do better than “nuh, uh” if you want to be convincing.
8GB really shouldn't be an option in 2026, it is just shortsighted and an insanely uneven build.
I could rant about Dell too. Or most other manufacturers (surprise, greed isn't apple exclusive). But Apple at least tries to keep the appearance of a higher profile.
Fair enough; though experience says 8Gb will run VScode, it would very much depend on the use case, I agree. OTOH, I would argue that anyone working VScode that hard probably isn’t buying 8Gb machines, but OP did say they’re running it so it’s up for discussion.
Can we please just move on? Maybe get your hardware checked if you’re legitimately still having these issues.
My M1 Air would slow down a little, but was still usable doing the same thing. And they both had 8GB of memory.
It is completely feasible, and the battery life - amazing. Even when running a whole pile of Kubernetes services.
Any modern Mac is more than capable. I had the baseline M1 Macbook Air that I did work on as well, just to see how that fared. Much better than this machine - 10x the price, but more than 10x the performance. This one is great as a "I don't mind if I break it or lose it" device.
Also a browser sneeze takes more than 4gb.
Using older hardware has helped me not accidentally build slow stuff. Although at some point I gotta upgrade and just add more performance tests :) but nothing replaces feeling it yourself.
The other I just owned the front end infra and was on the growth team. The rest of the folks were the stars on that one.
Edit: I guess I brought that up because I guess I don't know any more "real work" that that, ha. What is 'real work'?
But damn I like that design
i just got an m5 max with 128gb of ram specifically to run local llms
I developed some work that keeps tens of thousands of people alive every day on a $100 Acer netbook almost 15 years ago. The tools are always there, I don't think anyone thinks the work is actually impossible to do on a limited machine.
But... you can do the same exercise with a $350 windows thing. Everyone knows you can do "real dev work" on it, because "real dev work" isn't a performance case anymore, hasn't been for like a decade now, and anyone who says otherwise is just a snob wanting an excuse to expense a $4k designer fashion accessory.
IMHO the important questions to answer are business side: will this displace sales of $350 windows machines or not, and (critically) will it displace sales of $1.3k Airs?
HN always wants to talk about the technical stuff, but the technical stuff here isn't really interesting. The MacBook Neo is indeed the best laptop you can get for $6-700.
But that's a weird price point in the market right now, as it underperforms the $1k "business laptops" (to avoid cannibalizing Air sales) and sits well above the "value laptop" price range.
And, the whole shittiness of the experience will even distract you attempting real work: the horrible touchpad, the bad screen, the forced windows updates when you trying to start the machine to do something urgent, ads in Windows, the lack of proper programmability of Windows (unless you use WSL).... Add the fact that the toy is likely to break in a year or two. These issue exist on far more expensive Windows machines, how much more a $350 machine.
Leaving Windows machines and OS behind for more than a decade has been a continuing breath of fresh air. I have several issues with the Apple devices and macOS (as I have with Linux too), but on the whole they are far better than Windows. The only good thing about Windows that I miss on Macs is the file explorer and window management, not sure why Apple stubbornly refuses to copy those.
If Windows/Linux/x86 is non-negotiable and that’s your budget, I would never in a million years recommend anything brand new. This is when you go pick up a $350 used midrange ThinkPad on eBay. It won’t outperform a Neo in terms of CPU and battery life but I guarantee it’ll be a better experience than the garbage routinely sold at this price point.
$350 USD can get you a decent laptop with a SSD, 16GB RAM and something like an Intel N100 or N95. And they pretty comparable to a decent Intel Skylake CPU which are still pretty usable.
https://www.amazon.com/NIAKUN-Computer-Processor-Keyboard-Fi...
https://www.amazon.com/AOC-Computer-Processor-Laptops-Window...
Yes, the Neo has a faster CPU but it also has less RAM and less storage and costs more and has less ports. Besides ray traced games what can the Neo do that the others can't? They'll take longer but they'll get there.
And if you're willing to go used? That $350 goes a lot further.
8GB on Apple Silicon is far better than 16 GB on Wintel, and I don't event trust the quality of 16GB of RAM on a bottom of the barrel Windows machine.
Would you prefer a machine that is still good 7 years from now with less ports, or one with more ports that you have to replace in 2 years? Yes it is more expensive now, but over 7 years it is an absolute bargain.
Why would you have to replace it in 2 years? How do we know Apple will even be offering updates to Neo in 7 years? Will 8GB still be usable in 7 years really? 8GB is barely on the fence already.
I wouldn't be surprised if Apple drops the Neo from software support in less than 7 years.
Sigh. I mean, even absent the obvious answers[1], that's just wrong anyway. You're being a snob. Want to run WSL? Run WSL. Want to run vscode natively? Ditto. Put it on a cheap TV and run your graphical layout and 3D modelling work. I mean, obviously it does all that stuff. OBVIOUSLY, because that stuff is all cheap and easy.
All the complaining you're doing is about preference, not capability. You're being a snob. Which is hardly weird, we're all snobs about something.
But snobs aren't going to buy the Neo either. Again, the business question here is whether the $350 junk users can be convinced to be snobs for $600.
[1] "Put Linux on it", "All of your stuff is in the cloud anyway", "It's still a thousand times faster than the machine on which I did my best work", etc...
But as to the 4G quip, that's showing some ignorance of where the market is. The value segment is filled with devices like this: https://www.amazon.com/HP-Stream-BrightView-N4120-Graphics/d...
That's a 16G windows box which will happily run multiple VMs for whatever your deployment environment is, something the Neo is actually going to struggle with. The Jasper Lake CPU is indeed awfully slow, but again for routine "dev" tasks that's just not a limit.
You would obviously refuse out of taste, but if you were actually forced to use this machine to do your job... you absolutely could.
I used to think this way about Apple and its jarring to read with it 10-15 years behind me.
It reads as aggro and oddly tribalistic / sports fan-y.
(what people? who thinks its slower than an M1? who thinks you can't code on it? what will you coding on it prove to these people that the benchmarks they read can't? with all that, why get so invested you're buying a machine you don't want to use day to day? what does "handicapped" mean in this context?)
Only sharing b/c I never understood why people would roll their eyes at me, and apparently I finally reached my own graybeard moment, and I am now rolling my eyes at both of my selves :)
Having said that duckDB is awesome. I recently ported a 20 year old Python app to modern Python. I made the backend swappable, polars or duckdb. Got a 40-80x speed improvement. Took 2 days.
Just about every physical world telemetry or sensing data source of any note will generate petabytes of analytical data model in hours to days. On the high end, there are single categories of data source that aggregate to more like an exabyte per day of high-value data.
It is a completely different standard of scale than web data. In many industrial domains the average small-to-medium sized company I come across retains tens of petabytes of data and it has been this way for many years. The prohibitive cost is the only thing keeping them for scaling even more.
The major issue is that the large-scale analytics infrastructure developed for web data are hopelessly inadequate.
But that follows A and A+ which were extremely column oriented and date to early 1990s or even late 1980s ; and to various APL implementations going back to the 1960’s
Columnar DBs were very much a thing among APL users (finance and operations research) but weren’t really known outside those fields - and even in those fields, there was a period of amnesia in the late ‘90s/early 2000’s
Claude suggested to just use DuckDB instead and indeed, it made short work of it.
Outside of the king's ransom you now have to pay for it, you can fit 99% of problems into RAM.
I benchmarked I4i at ~2GB/s read, so let's say I7i gets 3GB/s. The Verge benchmarked the 256GB Neo at 1.7GB/s read, and I'd expect the 512GB SSD to be faster than that.
Of course, an application specific workload will have its own characteristics, but this has to be a win for a $700 device.
It's hard to find a comparable AWS instance, and any general comparison is meaningless because everybody is looking at different aspects of performance and convenience. The cheapest I* is $125/mo on-demand, $55/mo if you pay for three years up front, $30/mo if you can work with spot instances. i8g.large is 468GB NVMe, 16GB, 2 vCPUs (proper cores on graviton instances, Intel/AMD instance headline numbers include hyperthreading).
It’s staggering. Jaw dropping. Bandwidth is even worse, like 10000X markup.
Yet cloud is how we do things. There’s a generation or maybe two now of developers who know nothing but cloud SaaS.
I watched everyone fall for it in real time.
You're either underestimating how big cloud instances can get or overestimating how much it costs to rent a cloud instance that would beat an M1 Max at any multi-core processing.
According to Geekbench, the M1 Max macbook pro has a single-core performance of 2374 and multicore of 12257; AWS's c8i.4xlarge (16 vCPUs) has 2034 and 12807, so relatively equivalent.
That c8i.4xlarge would cost you $246/mo at current spot pricing of $0.3425/hr, which is, what, 20% of the cost of that M1 Max MBP?
As discussed recently in https://news.ycombinator.com/item?id=47291906, Geekbench is underestimating the multi-core performance of very large machines for parallelizable tasks -- the benchmark's performance peaks at around 12x single-core performance. (I might've picked a different benchmark but I couldn't find another benchmark that had results for both the M1 Max and the Xeon Scalable 6 family.)
If your tasks are _not_ like that, then even a mid-range cloud instance like a 64-vCPU c8i.16xlarge (which currently costs $0.95/hour on the spot market) will handily beat the M1 Max, by a factor of about 4. The largest cloud instances from AWS have 896 vCPUs, so I'd expect they'd outperform the M1 Max by about 50-to-1 for trivially parallelizable workloads. Even if you stay away from the exotic instances like the `u7i-12tb.224xlarge` and stick to the standard c/m/r families, the c8i.96xlarge has 384 vCPUs (so at least 24x the compute power of that M1 Max) and costs $3.76/hr.
The tooling — K8S with all its YAML, Terraform, Docker, cloud CLI tools, etc. — is pretty hideously ugly and complicated. I watch people struggle to beat it into shape just like they did with sysadmin automation tools like Puppet and Chef a decade or more ago. We have not removed complexity, only moved it.
The auto scaling thing is a half truth. It can do this if you deploy correctly but the zero downtime promise is only true maybe half the time. It also does this at greatly inflated cost.
Today you can scale with bare metal. Nobody except huge companies physically racks anymore. Companies like Hetzner and DataPacket have APIs to bring boxes up. There’s a delay, but you solve that by a bit of over provisioning. Very very few companies have work loads that are so bursty and irregular that they need full limitless up and down scaling. That’s one of those niche problems everyone thinks they have.
The uptime promise is false in my experience. Cloud goes down for cluster upgrades and any myriad other reasons just as often as self managed stuff. I’ve seen serious unplanned outages with cloud too. I don’t have hard numbers but I would definitely wager that if cloud is better for uptime at all it’s not enough of an improvement to justify that gigantic markup.
For what cloud charges I should, as the deploying user, receive five nines without having to think about it ever. It does not deliver that, and it makes me think about it a lot with all the complexity.
The only technical promise it makes good on, and it does do this well, is not losing data. They’ve clearly put more thought into that than any other aspect of the internal architecture. But there’s other ways to not lose data that don’t require you to pay a 10X markup on compute and a 10000X markup on transfer.
I think the real selling point of cloud is blame.
When cloud goes down, it’s not your fault. You can blame the cloud provider.
IT people like it, and it’s usually not their money anyway. Companies like it. They’re paying through the nose for the ability to tell the customer that the outage is Amazon’s fault.
Cloud took over during the ZIRP era anyway when money was infinite. If you have growth raise more. COGS doesn’t matter.
Maybe cloud is ZIRPslop.
If your application won't ever require more resources than a single server or two, then you are better off looking at other alternatives.
If the metal dies in a catastrophic way (multiple nodes at once and loss of quorum, catastrophic DC outage, etc.) you spin it up in AWS.
> Here's the thing: if you are running Big Data workloads on your laptop every day, you probably shouldn't get the MacBook Neo.
> All that said, if you run DuckDB in the cloud and primarily use your laptop as a client, this is a great device
Where I live, our government-funded clam research programs are mostly shutting down. Very sad.
Did a PoC on a AWS Lambda for data that was GZ'ed in a s3 bucket.
It was able to replace about 400 C# LoC with about 10 lines.
Amazing little bit of kit.
I wish more companies would do showcases like this of what kind of load you can expect from commodity-ish hardware.
- c8gd.4xlarge - this has a single 950 GB NVMe SSD.
- c5ad.4xlarge - this has 2 x 300 GB disks, which I put in a RAID 0 array. There are no c6ad.4xlarge instances, so this is the closes NVMe-enabled approximate to ClickBench's most popular choice, c6a.4xlarge.
I also added results from my local dev machine, a MacBook M1 Max with 64 GB RAM and 10 cores.
Here are the results:
| machine | cold_run_avg | cold_run_sum | hot_run_avg | hot_run_sum |
| -------------- | -----------: | -----------: | ----------: | ----------: |
| macbook m1 max | 0.48 | 20.68 | 0.43 | 18.60 |
| macbook neo | 1.39 | 59.73 | 1.26 | 54.27 |
| c8gd.4xlarge | 0.51 | 22.04 | 0.24 | 10.36 |
| c5ad.4xlarge | 1.29 | 54.14 | 0.55 | 22.91 |
| c6a.4xlarge | 3.37 | 145.08 | 1.11 | 47.86 |
| c8g.metal-48xl | 3.95 | 169.67 | 0.10 | 4.35 |
I know this is moving the goalpost, however, it's quite interesting that both of these cloud instances with instance-attached storage are still outperformed by the M1 Max (which is 4+ years old) on the cold run. And they would quite likely lose against the latest MacBook Pro with the M5 Pro/Max on both the cold and the hot runs. But that's an experiment for another day.
Props for identifying the issue immediately, but armed with that knowledge, why not redo the benchmark on a different instance type that has local storage? E.g. why not try a `c8id.2xlarge` or `c8id.4xlarge` (which bracket the `c6a.4xlarge`'s cost)?
That couldn't be more accurate
the laptop is gonna have some local code, maybe a lot, but if I'm doing legitimate "big data" that data is living i the cloud somewhere, and the laptop is just my interface.
Also, at on-demand prices, three months of continuous usage of a single c8gd.2xlarge will pay for that MacBook Neo. The MacBook Neo has a larger SSD than the AWS instances. To be fair, the MacBook Neo has seriously nerfed external IO bandwidth, so the c8gd.2xlarge will outperform it in networking. That being said, I think that any other Mac in the current lineup will utterly smoke c8gd.2xlarge if you are willing to use Thunderbolt-connected network adapters.
Given how little power modern Macs use, a little closet full of Macs with a decent network switch will easily run on a single 20A circuit and will perform better than quite a few thousands of dollars per month of AWS products. Sadly, you’re kind of stuck on MacOS (which is not actually a fantastic server OS) and the management tools are poor. Oh, well.
The Neo is neat and for someone who mostly does surfing and standard office work kind of stuff I suspect it’s a pretty great little laptop for way less than Apple usually charges.
But it’s not going to compete with an M5 anything.
Or am I missing something?
I ran TPC-DS SF300 now on the c6a.4xlarge. It turns out that it's still quite limited by the EBS disk's IO: while 32 GB memory is much more than 8 GB, DuckDB needs to spill to disk a lot and this shows on the runtimes. Running all 99 queries took 37 minutes, so about half of the MacBook's 79 minutes.
> Command being timed: "duckdb tpcds-sf300.db -f bench.sql"
> Percent of CPU this job got: 250%
> Elapsed (wall clock) time (h:mm:ss or m:ss): 37:00.96
> Maximum resident set size (kbytes): 25559652
Their numbers are a bit outdated. M5 Macbook pro SSDs are literally 5x this speed. It's wild.
That's decently fast but not especially remarkable, most Gen4 NVMe drives can hit 6-7GB/sec.
https://www.apple.com/newsroom/2026/03/apple-introduces-macb...
"The new MacBook Pro delivers up to 2x faster read/write performance compared to the previous generation reaching speeds of up to 14.5GB/s..."
Those speeds on the Pro/Max are impressive though, more in line with Gen5 NVMe drives. Those have been available in desktops for some time but AFAIK the controllers are still much too hot and power hungry for laptops, so I think Apple's custom controller is actually the first to practically hit those speeds on mobile.
That's not tldr, that's just subheader.
I’m guessing so many devs started out on 32gb MacBooks that the NEO seems underpowered. but it wasn’t too long ago that 8gb, 1500mb/sec IO & so many cores was an elite machine.
I did a lot of dev work on a glorified eePC Chromebook when my laptop was damaged. You don’t need a lot of ram to run a terminal.
I’m hoping NEO resets the baseline testing environment so developers get back to shipping software that doesn’t monopolize resources. “Plays nice with others” should be part of the software developer’s creed.
:shrug: as to whether that makes the laptop or the giant instance the better place to do one's work…
My good old LG Gram (from 2017? 2015? don't even remember) already had 24 GB of RAM. That was 10 years ago.
A decade later I cannot see myself being a laptop with 1/3rd the mem.
If it didn't, Apple has other laptops today with more RAM.
2025-09-08 : "Big Data on the Move: DuckDB on the Framework Laptop 13"
"TL;DR: We put DuckDB through its paces on a 12-core ultrabook with 128 GB RAM, running TPC-H queries up to SF10,000."
https://duckdb.org/2025/09/08/duckdb-on-the-framework-laptop...
It would be a surprise if more than 0.1% of Macbook Neo users have even heard of DuckDB.
Which means that this article is probably just riding the hype.
People buy Macbook Neo because they "just need a laptop" or are budget conscious.
I imagine a student would get their hands wet with Postgre before looking at DuckDB or similar.
It would be a surprise if they do heavy workloads with DuckDB. In which case it's definitely worth investing in a more powerful computer.
Also there are countless reports of bricked M1 8GB MacBook Airs that are bricked because the SSD used up it's write cycles
https://www.macrumors.com/2021/02/23/m1-mac-users-report-exc...
Do you have a source for these "countless bricked SSD's"?
https://m.youtube.com/watch?v=MZuv4TIjk-I&pp=ygURZGVhZCBNYWN...
If Apple would build their laptops serviceable like ThinkPads I would buy one today.
They’ve slowly been moving towards making it easier to repair individual broken parts. I’m very happy to see that a new keyboard doesn’t require replacing the entire top case. That was just crazy.
I just thought it was neat. It’s a phone chip, we’ve never been able to do stuff like this on an Apple phone chip before. No one was porting this to the iPhone to run there.
In my mind this is purely a curiosity article, and I like that.
There is always a trade-off of cost/convenience/power, and some folks are going to end up the the Neo end of the spectrum.
With I/O streaming and efficient transformation I do big data on my consumer PC and good old cheap HDDs just fine.
I’m really surprised just how competitive it was in their benchmark. I was expecting “sure it doesn’t compete but it works and you can use it”, not “it beat an Amazon instance, though not a really powerful one”.
I guess they’re using a different definition?
very much so…
You have phones that are faster than cloud VMs of the past. You can use bare metal servers with up to 344 cores and 16TB of ram.
I used to share your definition too, but I now say that if it doesn’t open in Microsoft Excel, it’s big data.
As you say, single machines can scale up incredibly far. That just means 16 TB datasets no longer demand big data solutions.
Many people like to think they have big data, and you kinda have to agree with them if you want their money. At least in consulting.
Also you could go well beyond a 16TB dataset on a single machine. You assume that the whole uncompressed dataset has to fit in memory, but many workloads don’t need that.
How many people in the world have such big datasets to analyse within reasonable time?
Some people say extreme data.
Google has big data. You are not google.
Or one could define it as too big to fit on a single SSD/HDD, maybe >30TB. Still within the reach of a hobbyist, but too large to process in memory and needs special tools to work with. It doesn't have to be petabyte scale to need 'big data' tooling.
8TB is a couple hundred hours of 4k RAW video assets.
No.
>Do I reject a world where all of the above is necessary to realize value from an entry-level MacBook?
In theory, yes.
Reality check: it works remarkably well for AI agent orchestration. The unified memory architecture means the agent, browser, and streaming can coexist without the memory wall you'd hit on x86. But running OBS alongside everything else does make it laggy — I've got an M5 MacBook Air (32GB) incoming and I'm planning to swap the Mini for a 64GB model to give more headroom.
For anyone considering Apple Silicon as an AI dev machine: the sweet spot is 64GB unified memory minimum if you want to run an agent + browser automation + anything else simultaneously. 32GB works but you feel the pressure. The M-series efficiency means you can leave it running 24/7 without worrying about your power bill, which matters when your AI agent literally never sleeps.
> Don't post generated comments or AI-edited comments. HN is for conversation between humans.