Old news now I think, but good news. Except for my Apple Watch I have given up using Siri, but I use Gemini and think it is good in general, and awesome on my brother's Pixel phone.
Because Apple Silicon is so good for LLM inferencing, I hope they also do a deal for small on-device Gemma models.
With very good tooling (e.g., Google Antigravity, Claude Coding, Open AI’s codex, and several open platforms) and not caring about your monthly API and subscription costs, then very long running trial and error and also with tools for testing code changes, then some degree of real autonomy is possible.
But, do we want to work like this? I don’t.
I feel very good about using strong AI for research and learning new things (self improvement) and I also feel good about using strong AI as a ‘minor partner’ in coding.
I agree so strongly with Antirez. For me (just a personal take!) the main benefit of AI is helping me improve myself via: better understanding technology that I am currently studying; a springboard for fleshing out new ideas; as an AI coding assistant. For deploying LLM-based applications I have a strong bias for starting with smaller models, looking for a good fit for cost/performance/environmental impact vs. what I need for a specific engineering task. Pardon the plug, but I have a work in progress book "Winning Big with Small AI" that can be read online at https://leanpub.com/read/small-ai
Well prepared article, and the window resizing pisses me off also!
For practical reasons I am stuck inside Apple’s macOS garden, but I wanted to share a few things that at least make me feel content using macOS:
First, I have at least two VPS systems so via mosh/ssh/tmux I always have Linux dev environments, the ability to use throwaway VPS for sandboxing, etc.
Second, when actually working on macOS I stick with tools that make me happy: Emacs and terminal windows, a uv-based Python enviroment and tuned-up Common Lisp, Haskel, and Clojure dev environments.
Anyway, I am just sharing my ‘macOS therapy’ - hope it helps someone here.
I hope things work out for Tailwind. I think it is very decent of Google to do this. Obviously Google takes some heat for their business model but when I was invited to work at Google in 2013 I thought the company had a definite vibe of trying to do the right thing in several dimensions (e.g., renewable energy for data centers).
Most smart TVs have advertisements and spyware that yields additional profits. Same with some electronic devices: Apple devices and Windows laptops sold directly by Microsoft have less advertising and spyware, but at a higher price.
Years ago I got so fed up with the smart TV experience that I bought a $200 dumb TV at Walmart, only had one HDMI input and terminals for a local antenna - hooked an Apple TV into it and had such a good experience.
It is easy to lose sight of how much money is made by collecting data on people and advertising.
The simplest answer is to buy a quality TV and not hook it up to WiFi, and use another smart platform. HDMI CEC works pretty well to discard the garbage TV "smarts" and replace it with something Android-based, Apple TV, or something else HTPC/free open like Jellyfin or Emby.
I have a Sony Bravia with an Android stick and Samsung QLED with an Apple TV. Less ads-ish and spying, but not totally out of the walled gardens. Already have a Plex lifetime and shoved stuff on a RAID10 NAS, so I'm okay with it as-is. I like that remote UPnP-basted casting works, at least with my login. Maybe Jellyfin or Emby have slight advantages in some areas, but it's the devil you know™.
Depends what flavor you get. Much like Android itself different manufacturers bundle different crap with their hardware. I used to have an Nvidia Shield and it was a wonderful vanilla implementation. But I've since switched to Apple TV.
Computer monitors have been getting a lot better while being cheaper, with no ads or services. You can get a high resresh rate 4K ips for about $200 nowadays. Display tech is just advancing faster than other tech at the moment
Huh, interesting. My experience has always been that computer monitors have been more expensive than TVs, even when panels are ostensibly the same. I've attributed it to comparative volume in TV consumers and desktop computer consumers.
At this point (as opposed to a decade ago) there's arguably no difference between a TV and a monitor anymore outside of packaging and the bundling of a remote and input defaults.
How does this work with respect to using a remote? I know something like a Roku remote would work display-wise, but you usually program it to use the signal that the your brand of TV responds to. That way you can use the Roku/whatever remote to turn on the actual TV and control audio. Speaking of, how does audio work for this set up?
HDMI standards allow plugged in devices to control the power state of the TV. e.g. my Apple TV will turn the TV on when I press a button on the aTV remote and will turn the TV off when I turn the Apple TV off.
Audio is a separate challenge, I'm not sure what you'd do there. Do computer monitors have eARC outputs? None of the ones I have do. Again if you had an Apple TV you could pair it with a HomePod (or pair of them) to avoid the issue but that's a niche solution.
Samsung already makes a bunch of "smart monitors", putting there the same software they use on TVs. Not sure about other manufacturers, but would be surprised if they don't catch up soon.
Many laptops with Windows preinstalled came with all kind of bloatware to 'enhance' the user with software they 'need'. Desktops too, but with laptops (and smartphones) it is more noticeable due to battery.
> Most smart TVs have advertisements and spyware that yields additional profits.
Something I just realized is TV companies can very accurately put a price point on a specific buyer - household size, TV watch time, content being watched, TV lifetime usage, etc and calculate how much the buyer is worth in their eyes.
> Windows laptops sold directly by Microsoft have less advertising and spyware
Really? It’s a little hard to believe. I’d think the easier thing to do is to put the same adware everywhere instead of segmenting out the MSStore-sold devices. Do you have a citation for this?
Manufacturers add ‘apps’ and other spyware. My understanding is Microsoft direct sales don’t have the extra apps, etc. That said, I have not been a Windows user for over 15 years.
Non-MS manufacturers get offers from e.g. McAfee to pre-install a nagware version of their software for a kickback. I have an ASUS ROG laptop, and even if I run a full Windows Reset, I get a prompt to install McAfee during OOBE setup, right after being prompted to subscribe to office/copilot/365/onedrive/game pass/etc.
While the over-arc of this message is good (avoid packaged and processed food) I personally don’t like the advice that these are not top tier foods: non-GMO organic whole wheat (i.e., not soaked with pesticides), brown rice, and other pesticide-free whole grains —- all in moderation.
I also don’t like the emphasis on meat protein. Small amounts of meat protein a few times a week are definitely healthy for most people, but organic (not soaked in pesticides) beans, lentils, etc. are almost certainly a healthy way to consume extra protein.
I sense the ugly hand of the meat industry in realfood.gov. I think if more people understood how (especially) chickens and pigs are tortured in meat production, it would help people who are addicted to excess meat cut back on their consumption to just what they need for good health.
EDIT: the documentary movie The Game Changers (2018) is an excellent source of information. The scenes interviewing huge muscular vegetarian NFL football players really put the lie to the ‘must have meat’ addicts. That said, I still think small amounts of meat protein are very healthy for most people.
I totally agree on all three accounts: unprocessed foods are great, organic wheats are good, and the concerning focus on abundance of red meat. I think we are going through a fad of "we need to gobble down as much protein as we can". I agree it's reasonable we need more, and especially older adults at risk of falling. I am concerned that there are so many junior residents that I work with that are throwing back protein shakes because they are "optimizing their macros". So many of these protein powders have added sugar and are contaminated with heavy metals! I will commend the guidelines for supporting lentils, beans and other pulses.
While I appreciated Meta's early open weight models - large benefit for society, I am in agreement with Yann LeCun recently. Two things can be done: judicious use of LLMs, especially energy efficient models and infrastructure, and research into new architectures and technology.
This article struck a personal chord with me: I bought a new MacBook a week ago and installed minimal software on it, specifically I did not install VSCode and I don’t miss it.
I use Emacs exclusively on my new laptop. I have about 40 years experience with Emacs and except for a treemacs automations, I am using my regular setup.
VSCode is a great project but I just didn’t feel “happy” while I was using it. I feel happy using Emacs and I only use very minimal LLM integrations with Emacs, preferring to separately running gemini-cli occasionally, or using a variety of LLMs (especially strong local models) with one-shot prompting.
Likewise, I feel happy when using Emacs in a way that other editors do not. Emacs was made for a different era of development, with different views on what productive programming looked like. Rider, VSCode, and etc are all post-NetBeans editors and it shows. Editing text buffers isn't the focus so much as refactoring projects is; and agentic AI development slots easily into that refactoring process. With Emacs, it _feels_ purposeful, manual, and dare I say it, artisanal.
Also started using emacs (doom) a couple of months ago, after realizing that jetbrains and vscode are going to be AI-shittified, and there is no turning back.
At this point, I would recommend to every coder worth his salt to just jump to vim/neovim or emacs, these editors will be around for the next 1000 years and you wont need to fight against some BS features and you wont need to switch ever egain.
The 1-2 month learning curve is worth it!
I was a long time Emacs user, spent way too much of my life in ~/.emacs.d/init.el. I don't use it for anything other than magit any more. I just tried it again, first by upgrading my packages in package.el. Of course, everything is still locked up when I `package-menu-execute` to upgrade packages. I guess in a thousand years it will still be mostly single-threaded, with almost every action locking up the UI thread.
Believe me I have tried. And already have made my config. Took me weeks, and is still no closer to getting to be up to par with what I get with helix out of the box.
It also is just super slow on windows unfortunately.
I’ve fallen in love with emacs again now that I can have an LLM tune up my config. I love emacs, I don’t love lisp. Maybe LLMs are helping me with that, too.
But I so happy with my config now. Simplified and modern.
I‘ve been happily using Doom Emacs on Linux as VSCode replacement. When I switched to macOS, I found the experience to be rather slow with significant input lag, though. (And yes, I did use native compilation.) Has something changed in that regard, or did you just accept it?
Many Emacs functions are sluggish on my work Mac, but I found out that this is because the Cisco endpoint security software stops and checks every binary that runs, every single Goddamn time, which means that things which shell out, like M-x compile and anything in magit, are noticeably slower.
I had the same issues. I mainly use Doom on my linux machine, but on mac it is distractingly slow. d12frosted/homebrew-emacs-plus seemed to reduce the lag, but I've had other issues with it, so have gone back to vim.
I'm also a long time emacs user (>15 years) but got tired of the endless config fiddling, with some packages breaking over emacs versions, other packages which were cool at the time slowly getting stale and need to switch to yet another similar incarnation of the same idea.
And most of all having to recover the config every time I use a new computer or just connect to a new VM.
I'm building an alternative, and I haven't opened emacs for a month now
I've been Emacsing for 30+ years at this point, but I'm frustrated at its performance in the 21st century.
By which I mean both startup time (yes, I know real Emacs people never leave the editor. I'm Not That Guy) but its single-threadedness leading to painful blocking pauses when using eglot + rust-analyzer, etc.
I bought a new basic laptop last year, with enough RAM, and little do you know, the VS Code did not feel faster on the new device. So last week, when I noticed that Zed for Windows is "stable" now, I've uninstalled VSC.
VS Code is still the better tool (imho) but I can't stand it.
Because Apple Silicon is so good for LLM inferencing, I hope they also do a deal for small on-device Gemma models.
reply