We're getting close to local AI being really useful. Before the M1 is dead we might have already seen a huge shift. In my mind, there are two types of macs: those that have the bandwidth to run local models (Max & Ultra) and those that don't (Base & Pro chips).
The generation doesn't mater as much to me as the memory bandwidth. When that day comes, I'd rather have a M1 Max 64GB over a M5 Pro 64GB.
Local AI already is really useful. This has been true since Qwen Flash Next and Qwen 3.8 27B rolled out. The struggle with exactly 64GB of unified RAM is that you end up having to stream the engram table from SSD because you will need almost all of your 64GB to run QFN. A 27B dense model still runs a little slow, even on an M1 Max Mac. However if you have even an older gaming GPU with 12-16GB of RAM and enough system RAM you can run this model (QFN) at a very respectable token rate and intelligence level. It is a Sonnet 5 class model you can run at home. There are new quantization techniques that make these models very good at agentic tasks.
And now if you stretch you can run DeepSeek Flash 4.1.
And if you were willing to roll up your sleeves you could build your own AI lab at home and run this model and models like GLM Flash 5.3 and DeepSeek Flash V4.1 at lower quants. The budget GPUs (like V100) are kind of not budget now, but still <5K to acquire 128GB of fast VRAM that can compete with a 5090 and actually has a lot more interesting options available IMO.
Can share a good way of running Qwen 3.8 27B (or similar) on 16GB VRAM? I've been trying but I end up with either a heavily quantized model or a small context or a slow speed.
Dense models are not well suited to this. Depends on your system, but most inference engines let you put the KV cache on the host. They should let you use "Q8" KV cache format as well which takes half the storage of bf16, the default format. You essentially are stuck at having to fit the entire dense model into VRAM because it must compute across the entire model every token. Without it being in VRAM it is slow. MoE models such as Qwen 35B with 3B active parameters are much friendlier to limited VRAM systems. This is why you see systems with 16GB of VRAM running something like DeepSeek Flash if they have a bit of spare system RAM. Usually not at any good token rate, but tolerable.
Dense models also don't like quantization as much and are less tolerant of it. There are some very interesting smaller models like https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4... Dory that I would check out. You need system RAM or to stream a larger MoE model from disk and it won't be fast. You need to specialize the models some to get quality results if you have very limited system resources.
I asked Claude to come up with 7 problems to solve that were not part of any benchmarks. Besides speed, I asked it to put together a scoring that would reflect models that perform well on the 7 problems and in a reasonable time-frame (RWS, the definition is in the artifact). I valued models that produced good answers in a reasonable time better than models that had very high throughput but kept spitting useless stuff. Overall, IIRC most of them still had a >30% chance of producing wrong/non-working replies.
I then tested Strata ( https://github.com/niko1221/strata ) and the results are not even close. You can basically ignore everything in that table and use Strata instead, you can use bigger/less-quantized models and have way better results. Basically any otherwise-decent model degrades noticeably at Q < 5, and KV cache at Q4 also contributes a lot.
Still, a 20$/mo Claude subscription is way cheaper/more efficient if you code 2-3h in the evening than running stuff locally, unless you really have hardware than can hold models that get most things right at the first try. You can iterate way faster and cover a lot more ground in 2-3h using a Claude subscription, or OpenRouter if you prefer to change models. You get faster replies that are 99% correct at the first try, whereas using the local model you'd be spending 30% of your time fixing things that don't work and prompting it to do better.
Nowadays I use the local model to delegate tasks that would burn a lot of tokens, like sifting through big data archives for indexing and such.
QFN Q4 K XL Unsloth is basically identical to the bf16 model. It’s KL divergence is very very small. It is competitive with the full model. Also I benchmarked iq3 xxs on Strata for 3 days against deepswe and it scored better than the published benchnarks for the model. If the gsq rco quant prettain tasks fit well it is a great model quant. My data shows it takes 20% more tokens basically than q4 xl. A4 xl is within noise on all my repeated task benchmarks over the public model serving and my own runs of q8. I published the deeswe results. You have to be very specific about model choice and know these models and their quirks.
This is not about if a frontier AI provider is cheaper. It is about having control over your data and use of AI.
> QFN Q4 K XL Unsloth is basically identical to the bf16 model.
Yes, but like original poster I have 16GB of VRAM and 64GB of RAM, I was referring to non-Strata benchmarks :)
I did quickly run a couple things on Strata but haven't gotten around yet to benchmark bigger models/customize them with Strata, but will do rather soon. So far Strata/Swift IQ2_XS performed really well compared to non-Strata/Qwen3.8 quantized variants.
> This is not about if a frontier AI provider is cheaper. It is about having control over your data and use of AI.
Yep, totally get it, and to be honest getting to run ablated models sometimes can also be useful.
> Also YMMV in “faster” — I have less than 5K into my home AI lab shenanigans and it’s pretty fast :)
My rig is... ~1.5k, maybe 3k if we count the CPU and GPU upgrade since 2019 :-D I really wanted to grab one of the latest mini-PCs (Halo, Spark, and other minor assemblers) for local setup, but to be honest I can't justify a 6k+ purchase for me and my wife alone. Perhaps in 3 years these mini-PCs will get to the ~2k ballpark (perhaps used), and although older by then, I might be able to run things 24/7 locally as well.
With 32 GB more of RAM you can run ISTA GSQ RCO IQ3 XXS fast. You can make it work now but it is tight. You can stream experts from disk with builds now. You will get an okay token rate. Try it out if you have some disk space. It is a very good local model. I benched it at that quant. Sonnet 5 class at coding benchmarks.
Get a couple or V100S GPU. Figure out cooling. They will run it /fast/ with the right engine. :)
The easiest thing to do (takes 5 minutes) is to use LMStudio and then search for the model inside LMStudios UI.
It will show you on the model selection IF a model fits and if you use a "Qwen 3.8-35B-A3B" type of model, it should just work out of the box. A3B stands for Active which means you need a lot less fast ram.
You will run into issues fast though: If you have big tasks for it to do, your pc/mac will get warm very fast and noisy and you will cosume relevant more energy.
My setup is a Ryzen 3600, 90GB DDR4 allocated to a VM (don’t even really need that much, just had it available, special thanks to the year 2024), and an RX 9060XT 16GB ($500 GPU).
I get around 40-60 tok/s depending on the quantization level I choose.
As you can see, the CPU is nothing special, don’t even need DDR5.
Yes it runs on all levels of hardware - with the latest Strata my 5090 and 64GB DDR5 gets 200-240 tok/sec decode, 6500 tok/sec prefill. Around $0.09/M tokens in power usage (have excess solar capacity).
Jealous, lol. But I did spend a grand total of $500 for my setup. The rest of the server was built purchased with the use case of running game servers.
AI harnesses like Claude will set it all up for you, you can just say "I want to run a localllm on this machine" and it will find the best one for your hardware and even help you fine tune it etc
You will end up with a sub-par result. The space for small models is more specialized and specific and you should optimize to the specific task and or tasks you want to do. It will get some local model working, but there are a lot of interesting small models in the 4-10B space to look at and play with for your specific tasks. If you expect a full agentic coding agent it's just not going to happen without a larger model like Qwen 35B and ... that is not that great by today's standards, even Muse Glimmer is better.
After arguing and back and forth with Claude (not straight away), it found a kv streaming fork of llama https://github.com/achyudh/llama.cpp and an interesting ascii only model Qwen3.8-27B-ASCII-Condensed-UD-IQ4_XS.gguf . It's running, i'll see how it goes.
That's only part of the issue - prefill is horrible prior to M5 class chips.
So it depends on what you're doing. Using it for a fun oracle without feeding it too much context per turn? Sure, an older Max chip is great.
As a coding agent doing meaningful work on a sizable project? Compared to a cheap as chips cloud option, it will take an order of magnitude more time to complete.
Prefill isnt a huge problem with an MoE, but ya, dense models are a non starter before the M5. You are still better off using DeepSeek than a local MoE, but locally running something gives you possibilities if you are on an airplane without internet, or if you want to jump guard rails, or if you want to run a home device server, etc…
SSD token caching (like oMLX does) goes a long way for prefill. Typical miss rates are under 5% in long coding sessions, which is mostly the output of a turn's tool use.
>The generation doesn't mater as much to me as the memory bandwidth
I don't know man?
The generation kind of does matter. I'm not gonna lie, matrix multiplication units speed things up considerably. That hardware confers a material change in capability.
Maybe from M5 up, it doesn't matter? But you're definitely going to notice the difference between M1 and M5 for instance.
Still on my M1 Max 64GB from release date. Love it. I don't do local AI but this laptop runs so cool and fast I have no reason to upgrade. They really knocked it out of the park on the design.
It feels just as fast as day 1 and I've never had any speed issues. Say what you want about Apple and their pricing and ecosystem but I've been using their hardware for over 20 years now and it's always been top notch.
If you buy their support/AppleCare, your support and warranty is excellent too. They don't skimp. I've seen them go so far past the extra mile it's unreal.
This is also my experience, but when I saw how few tokens/second with a large model it gets locally I was bummed. I’m wondering if I’ll be obligated to upgrade in the near future
How many emulators are you running at once? You can also change how much memory is dedicated to each Android device (if you’re using Android Studio at least); often times I can lower each Android emu by a couple GB from default and it’s fine.
Running 2 is enough to slow it down to a crawl unless I'm not doing anything else and just letting agents work.
I've definitely tried optimizing with lower memory as you suggest but the margin is so small given normal usage of the computer at the same time that it really doesn't make a difference unfortunately.
So that's 2 emulators + a few terminals of pi/Claude, Firefox open, vscode and usually docker but nothing too heavy there.
I jest...but I'm disgusted by the price gouging going on. Glad my hardware is relatively recent and I had mostly maxed out already when I bought it a few years back. The same setup today would be closer to $7k with equivalent parts.
Puh i don't know, there might be plenty of new/better chips coming out for AI next year or in 2 years.
Isn't it also rumored that the M7 is reworked for AI?
And Xiaomi announced their AI Cube with 1.2t of bandwidth. Just because Nvidia is protecting its business by giving consumers either compute or bandwidth and Apple being something in between, doesn't mean we can't get both a lot better balanced.
Yeah, the price is so high and things are moving so fast I don’t wanna risk committing to something that becomes obsolete in a year. I wanna see the rumoured M7 first. Hopefully hardware demand doesn’t make waiting a stupid decision…
Waiting is never a stupid decision because you can always save a good deal on apple refurbished. I got a mac on bestbuy refurbished a couple years ago shortly after it was released. It had like 4 cycles on the battery and I saved hundreds.
Yes the latest chip is better but the rumor is that the M7 has either more/better parts for LLM math OR that it has significiant changes to memory bandwidth too.
It seems so wasteful to spend premium on the hardware for local ai to be shoved into macbook form though. You'd probably save a lot building a compute node you shove under your desk that you connect to with some more comfortable but cheap thin client, maybe a refurbished low spec macbook pro (inb4 I say air or neo and people say they cannot work on inferior screens).
Then again the people working on stuff involving this much local model compute demands probably make enough where they don't care about the couple grand they'd save building their own node. I guess apple will get the bag from being the nicest prebuilt out there in this case.
> I'd rather have a M1 Max 64GB over a M5 Pro 64GB.
I am one of those: bought the M1 Max with 64gb thinking in running VMs, afterward bought a lighter Air with less memory and ended up using the M1 Max as a kind of luxury for LLMs and multicore computing.
Running Gemma 4 26B A4B and Qwen 3.8 27B on a M4 Pro with 24 GB - Qwen 3.8 runs fine but is slow and taxes the machine a bit, Gemma 4 on the other hand, is a better experience.
I guess this means I’ll be hanging onto my M1 Pro for another year and seeing what M7 has to offer.
This is the first time I’ve ever kept a computer this long and never really felt like it needed to be upgraded. I suppose this is a bit of a gift and a curse for Apple.
More on the curse side, tbh. Mostly because RAM and flash chips have exploded in price and fabs get now booked by NVIDIA et al where previously Apple was the biggest dog and could demand better conditions.
But well, they'll survive :) Their M-series was a stroke of luck considering AI.
Srouji is killing it on whatever he has on mind or is asked to do. Everyone is riling up on GPU and Cloud AI costs. 99.99% of what people use AI for today already processes fine on an M1 from 5 years ago.
I'm running qwen 3.8 flash next with strata (q3 quant) on laptop with 5090 and 128G RAM and OMP harness. At 90t/s speed it is pretty much good enough already.
Probably not. The ROI for local AI is pretty terrible at best it's like 20 years to see a cost advantage. Most enterprises are going to be fine using hosted models. The two main reasons to use local AI are 1. education and 2. if you cannot use hosted AI for your use case (safety guardrails, payment restrictions, etc.)
Considering the insane memory price surge it's going to take years (or an eternity) before most people will consider upspecing their machines to run local over paying $20 per month to have the honor to loan a slice of Altman's hardware.
The only hope we have is the AI bubble bursting, but that will undoubtedly come with its own set of macroeconomic surprises.
The ~$1T/y AI capex isn't remotely sustainable. Memory fab capacity will go up, AI datacenter provisioning will go down, prices will become reasonable.
What is it with fabs where there is just no competition springing up outside the existing memory cartels? Is the knowledge of how to build and run a fab strictly controlled by these companies? Is all qualified talent just immediately poached and put behind NDA and noncompetes? It just makes no sense. Capitalism would suggest these conditions with so much unstable equilibrium in this market would see increase in supply to take advantage of the pricing. The fact that this isn't playing out as such suggests there are some very powerful thumbs on the scale of this "free" market.
It takes about a decade to bring one up and during EVERY wave of a booming memory demand supply crunch there has been a subsequent round of major bankruptcies from suppliers who decided to scale and had no demand when they finally got online.
Supply is not elastic and responds very much slower than demand. There's no conspiracy or corruption needed for this result.
The conspiracy is real, it's even been proven in court, but as things stand, it's also necessary because of the factors you describe. Investors will not touch the DRAM market with a 10-foot pole, so it has had to rely on self-funding. While there have been state subsidies too, price fixing has proven to be the most effective way to survive the extreme volatility and fund the massive capex necessary for each generational leap. They have gotten a bit more sophisticated about it since the court case 20 years ago, but as far as I can tell, it's basically the same as it's ever been. The lack of effective competition is a function of the market itself, not any tilted scales. If anything, tilting the scales on purpose with greater state involvement is the only way to change the dynamics, but few states have the capacity and competence to carry that out effectively.
Micron's stock price did go way up, after they stopped selling DRAM to the consumer market. It has been flat for several months though.
For comparison, SK Hynix stock is flat YoY and Samsung's stock did go up this year, but has come back down a bit (they're also much more diversified). Those are the only other companies outside of China that actually make DRAM at scale. There's also CXMT, which is making inroads, but its stock is not traded on any open markets, and the company is heavily backed by the Chinese government.
Mini PC with Occulink, EGPU dock, 750w power supply and a refurbished 3090 is 2500. It can run Gemma4:31b at 1000+ tok/sec prefill, ~40 tok/sec generation (which is slow, but a fully connected model is way better for all tasks)
if you want more specialized partial activation models like Qwen, you can get 100+ tok/sec generation.
This beats anything that Apple has by a long shot. You don't need to run large models for usefulness. Large models can get things write the first time, but thats their only advantage. Running small models that are good enough like gemma4:31b fast allows you to do faster iterations.
And if you really want, you can also do multi GPU setups if you want slightly slower model loading with usb4/thunderbolt ports.
TLDR; local ai has already been useful for quite some time if you arent brainwashed into believing that Macs are the best for running local inference.
> "Man you Apple Fans really live under a rock... ...refurbished 3090 is 2500"
Before you make assumptions, there are valid reasons for the choice I made:
I paid $1k for a refurbished M1 Max 64GB macbook that can run Qwen 3.8 27B at 24tok/sec (looks like they are about $1.5k on ebay). In my case, Apple was a lot cheaper (used) if you're looking at price-per-token. Yes, it's almost twice as slow as a 3090, but it's also half the price /performance ratio.
Qwen 3.8 27B Q4 with 64k:
3090: ~40tok/sec
M1 Max : ~24tok/sec (MLX)
In fact, the Max can even do 265k context which isn't possible with a single 3090 last I checked (without dropping to Q3-Q2)
Go ahead and run Gemma4 on your M1, and you will find that its unusable at 12 tok/sec.
And this is also not touching prefill, which is the biggest thing that you need fast, and Apple sucks in this regard compared to nvidia cards. With gfx cards, you can throw big codebases at the local models and it will happily crunch through them in very little time.
Sure, spending 1.5k on a Macbook enables you to run a local model, and if you really have a good reason for running local (like running heretic/abliterated models) for tasks you don't mind waiting for, it works. The difference is, spend 1k more and let you never have to pay for tokens, and you can run inference all day long without your computer thermal throttling, like your Macbook.
And I don't need to make assumptions when you say idiotic stuff like "We're getting close to local AI being really useful", like you have any idea what you are talking about. On par with being an Apple fan though.
A $2500 Mac Studio with M5 Max running Qwen3.8-27b or Gemma4:31b in oMLX gets about the same speeds you quoted... and it comes with a warranty and has enough memory not to choke at longer context lengths.
Gemma4:31b, which is the better model with more complex architecture (and thus can't be compiled into a megakernel), is gonna be way slower in prefill, and ~50% slower than a 3090 even with oMLX. You can't get away from much slower shared ram vs vram, and less parallelism compared to 3090.
The other advantage of an eGPU setup, you can simply add/swap performance. Dual 4070s give you 24b of ram and you can still get about 80% of the performance of a single 3090, and actually you get faster prefill due to more parallelism.
So in terms of money/usefulness, Mac always loses.
At this point, I really want all AI benchmarks to use the best coding/writing model that normal people can run locally.
For now that seems to be Qwen 3.8 27B that requires ~28GB of RAM (Q4) which is pretty doable for anyone with a beefy GPU or slightly-above-average mac.
I want to know how close open weights (open models) are to all these leading commercial models. Yes, yes, local models aren't as good, but that distance is closing as local models are already better than what we had just over a year ago in many cases.
Same with Text-to-Image generation (Flux 2, Z-Image-Turbo, etc..) or text-to-speech (Parakeet, Whisper, etc..).
RAM is insanely expensive and inefficient. Try to keep important info in L3 cache. Every additional hop from the CPU transistors adds exponential overhead.
Everywhere we look we see people making the best choice for them. Big apps using Electron vs native. Ports of 250MB games like Starcraft now taking 5GB of "compatibility" libraries to run on modern PCs.
I completely agree that people need to be aware of Latency Numbers Every Programmer Should Know (https://gist.github.com/jboner/2841832), but in reality we don't all have the budget, time, or personal to do things the right way when it's the company ordering us around vs our own projects where we can do things "the right way".
Honestly, I'm glad companies are wasteful. If large companies were super conservative and agile then our economy would be in really, really bad shape. It's thanks to these inefficiencies that small startups can jump in and capitalize in ways the many bureaucracies of enterprise can't.
Then again, the dev in me would LOVE it if everyone rewrote everything in Rust or Zig and we put wasteful scripting languages to death for good.
That reminds me. Can we standardize cypress tests to load the website via a restricted speed as one of the tests? Some of of our modern webpages are still slow on the latest iPhone/Pixel/Samsung over 5G.
8,600,743 bytes of assets and frameworks later I'm finally able to see your contact-us address.
If you have a 24-64GB mac, consider running Qwen3.8 27B locally at night. It's a bit slower to run locally, but if you're sleeping it's less of a problem.
Depending on your memory, you'll need to use the weaker Q4 versions but they still perform well.
It ranks higher than GPT-5.3 Codex (xhigh) or Claude Opus 4.6 (max) so is great for pairing with https://github.com/kunchenguid/gnhf for nightly experimentation, cleanup, or recommendation lists for in the morning.
I have all sorts of local compute, and local models fairly capable the Frontier models still way more capable/faster and local electricity consumption is something else. Good thing it is getting colder around here.
I often pair them up, and I have an Astra or Sol work as a supervisor and reviewer while Qwen 27B FP8 or Qwen 3.8 Flash Next implements things.
I mostly do it as an experiment, just to see what kind of level of autonomy I can get, and they are slow to getting a decent reviewed outcome despite Qwen27B running at 100+ tps and 3.5-4K prefill rates and Qwen3.8 Next at 40 tps and 1-2K prefill.
I've been also using similar approach more with OMP, not just the straight Pi harness. And OMP seems to be slower because it has more guardrails.
OMP has an interesting feature where you can assign a better LLM as an advisor, wehere it just sort of monitors the progress and injects guidance. And it definitely helps, but one has to be careful. It actually turns out to be expensive if the cache reads are expensive. I learned it the hard way. Where on Fireworks' API, the cache rates for GLM 5.3 flash are quite a bit more expensive than for DeepSeek, and a simple runs ended up costing me three bucks in oversight.
So a better way is to have a Frontier model running a separate tmux pane and just directing it to Wake up every 10 minutes, take a peek at what's going on, review the milestones give feedback and then sleep. This turns out to be pretty decent cost saving strategy when quote needs to be stretched.
Paradoxically, OpenAI tightening up their quota allowance once they released Astra actually pushed me into all these sorts of experiments, and it's actually been interesting. I've been exploring all these smaller flash models, and it's been nice.
I do like using local LLMs for chore type tasks that are just mostly information gathering, post-session reviews, stuff like that.
There also exists a $600-800 GPU that can run Qwen 27B 3.8 @ like 60 t/s for around 200W of energy.
I find Qwen Flash Next quite competent as well. 27B is a solid worker like you said. If you batch work and let them crank they do remarkably well. I am working on.
I started using Herdr and taught my agents to use it. So I use OMP loop and or goal, and it has a review cycles to wake up an Opus or Sol reviewer to make sure nothing is going off the rails with a local qwen flash next coordinating for me. I kinda prefer Sol, it seems like a more patient and thorough model, especially Sol 6, but Opus 5.5 is really good and its voice and attitude is not as grating as Opus 5 for sure.
I have a few V100 GPU running Qwen 27B and they do all the work overnight. Not quite the same speeds you have yet, but this is V100 machine and an old gaming machine with 16GB 4080 and a handful of 32GB V100s... all in less than 3K (ignoring that my gaming machine is 3 years old, but runs qwen flash next for free now as I game not a lot) for my little "we have AI at home" projects and there is a lot of interest in these old GPU now because they are rolling out of data centers now.
For my local work and personal projects... they just seem to be getting done in this setup. Every few days I sit down and do a big cycle with astra/fable/opus batch things up. I have projects that are basically "i want to see what happens" to "I want this to be good, I understand the code". Some of the throwaway projects that have just kind of magically finished more or less how I wanted have been great.
What do you mean by "local electricity consumption is something else"? Doesn't an M3 Pro for example draw about as much power as a bright incandescent lightbulb for a maxed out gpu workload (~100W)? That's less than a tenth of what a frontier model will use in the datacenter (which I believe are racks of BlackWell or Vera Lynn GPUs, each using 500W+).
Because M3 Pro is not a real thing as far as actual agentic workflows go.
I run dual R9700 boxes those idle at 150W Because of a Ryzen AM5, and I run Spark boxes, which are decent But still idle at 45-50 watts a pop. My favorite, to be honest, is a 5090 with a Qwen 27B because that one is good for quick hitters and flies, But again, the box itself idles at 140 watts.
So it's really the idle power that I don't like And it's too much of an inconvenience to power boxes down and power them back on, so they just end up running and sucking electricity. I am working on getting some sort of a smarter proxy setup where I would give boxes time to wake up and go to the cloud while they're waking up. So the whole point of having these things running in the background is that you do want them to be almost always available. So power consumption is definitely an issue.
for 32gb, this is my model of reference now, you have to run it with a patched version of llama and is still not available in lmstudio or omlx, waiting for that to streamline the experience a bit. But so far, the best i had till now.
I still haven't found a use case for Qwen3.8 27B that Qwen 3.6 35b A3b (MoE) is better at. I can get at most 40 tokens/second with 27B, but I get around 90 tokens/second with the MoE and it seems to be a more capable model.
I guess I should still keep experimenting though. Maybe I'm just not using a dense model correctly.
if you have 128 GB, you could use Qwen Flash Next at some reasonable quant, with the new SSD hack for only keeping some of the model resident in memory.
I'm not sure about a return to keyboard, it's not large enough to actually type on, and swipe seems to be a much faster alternative to pecking at the keys.
Maybe swipe typing isn't as popular as I first thought.
hi there, I've done pocket PC, palm pilot graffiti, palm treo integrated and dockable keyboard, landscape slider keyboards, stylus only qwerty keyboards, Swype the app, Gboard, heliboard, t9 physical keys, messageease, and im sire some others.
day 1 the plb is slower than swipe typing because it needs getting used to. But the trick is one can focus one's attention on the content being typed /learn to touch type on a physical keyboard that on screen keyboards cant do, because of lack of consistency of on screen kbs always shifting layouts on you, at least with coming and going but usually also ruining the dot menu for common punctuation depending on if the input field is trying to be an email or a number or whatever.
dont try to help me keyboard, just give me a ctrl character, consistent shortcuts, and don't lag half the time I enter some text input, so I've typed out 4 words before the phone realizes anything has happened.
so yes i think you are right that swipe typing has value and more people should use it. but its lipstick on a pig for most of osks not being reliably static enough to let muscle memory / touch typing grow.
my unihertz titan 2 also has capacitive keys, i turn them off because I want the kb as dumb as possible (though i add keymapper android app to give me vim style arrow keys), but swipe typing would be possible on my or this device with just software support if there were demand for it / someone got their llm to vibe it up.
Less on the device and more on the keyboard. On Android, you can install third-party keyboards like any other app, and there's a wide range in quality. The default Android one is decent, but not among the best.
I've been strictly a Sony Xperia user for almost 15 years now, and Sony had its own keyboard app early on, which they later discontinued. Swipe typing on the Xperia keyboard is still the fastest and most precise I've ever been able to write on a phone, nothing else ever came close. Then they stopped development and moved over to a Microsoft keyboard, I think. I've never found another keyboard that did swipe typing as well as Sony.
I leave it running at night. No danger of burning my token subscriptions and it has hours and hours to run slowly with a manager like: github.com/kunchenguid/gnhf
Htmx 4.0 is over 100kb / 2,000 lines of code. Isn't there a way to slim this down some? I mean, isn't the basic idea just a couple lines of code? I remember AJAX 2.0 or whatever loaders that were only 1-5kb wrappers around XHR back in the day. Now we have fetch() which is even more basic:
I don't mean to imply this is close to total number of Htmx features, but fetching islands of content is the core idea and I feel like we're way past feature bloat at this point for something that is supposed to be simpler than React (Preact is only 10kB)
Yeah but when you're reading the actual source code you're not thinking in terms of bytes, you're looking and files and line counts. Htmx is deliberately maintained as a single, no-dependency file of about 5000loc. The maintainers are of the opinion that it reduces complexity, and I tend to agree.
The generation doesn't mater as much to me as the memory bandwidth. When that day comes, I'd rather have a M1 Max 64GB over a M5 Pro 64GB.
reply