joshstrange

It's the complete opposite for me, my interest in side projects has exploded.

It was never about writing every line by hand, it was about solving a problem and making my life better. LLMs let me do that at a speed never seen before.

I was and am and avid Home Assistant fan but the number of automations/scripts I now have is higher by a magnitude than ever before. Few things would reach the level where it made sense to spend my time writing and debugging a workflow but now I can speak my automations into existence.

Yesterday I got a candle warmer, I wanted to set it up so when I turned it on it started a timer to turn off and I wanted a push notification to see if I wanted to extend the timer when it was close to finishing. All things I knew were possible but HA YAML is miserable (all YAML is) and the UI is not much better (it's a hard problem, I get it). I prompted Claude and had a working implementation in <10min. Additionally it asked if I wanted to auto-turn off if I left the house (yes!) and if it should auto-start the timer if ever detected the candle warmer was on without a timer set (power loss, HA restart, manual interaction with warmer).

My HA automations have never been more bulletproof and if they do something wrong it's as simple as telling Claude Code and having it fix the problem. For example, my motion activated lights (motion sensor -> HA -> smart switch) were working great but I realize that while in the shower my main bathroom lights would sometimes turn off. 1 prompt later and Claude grouped up my motion sensor in the shower to the one in the bathroom so if motion on either was detected the lights would stay on but the shower light would only turn on if its motion detector detected motion.

HA is only one small part of my LLM usage but it's representative of my overall point.

My biggest issues with Passkeys is how inconsistently they are implemented and how opaque they attempt to be.

I understand SSH keys, I've been using them for decades, I know where they live, I know how to secure them.

Passkeys are murky as fuck. Is your PW manager supported? Do they sync? Where are they stored? How can I move to another PW manager if I want to in the future? Can I have more than 1 passkey per site? And the list goes on.

I _know_ some of you out there can answer some/all of the questions above but it's mostly on a per-site basis. Passkeys take too much of the control out of my hands and I don't like that.

Even more than that, I hate how they are trying to be pushed on me at every turn. Login -> Want to save a passkey (but they never call it that, they use some other confusing euphemism)? I click "No" and then it proceeds to pop 1Password's UI, then I dismiss that and it opens Chrome's passkey save UI, I dismiss that, and then it opens the OS's passkey UI. It's incredibly disrespectful and unclear.

I never use anything but 1Password but somehow everyone (OS and Browser) try to reach their grubby hands in. This is what scares me, I don't like having to be on high-alert to not accidentally save a passkey in Chrome or Safari and not realize until I'm on a different device and notice it's not in 1Password.

Lastly I trust the developers implementing passkeys... none, I trust them none, zero, zilch. I don't trust them to pick the right defaults, I don't trust their recovery options, and I know they will always pick the configuration that benefits them and not me.

No, for now I'll stick with my as-long-as-you-let-me-make-my-password random passwords which I never copy/paste into random website and be perfectly safe, thank you.

I guess I see very little "soul" in most marketing/advertising in the first place. It was slop well before LLMs entered the scene. If my local hardware store wants to put up a poster for their summer sale I don't care if it's black comic sans on white paper or if it's a AI-generated advertisement. Either way I'm not looking that closely at it and neither version were going to show up in an art museum or recieve design accolades.

There is so little intention in anything (again, even before LLMs) so I don't begrudge places for using LLMs to try and produce similar quality to what larger chains/corps are able to commision.

I see this as similar to things like people wanting "clean ingredients" or "sustainable packaging" or "healthy food" in restaurants/bakeries/etc. They _say_ they want that but they are not at all willing to _pay_ for it. To be clear, I do wish those things were the baseline, but they are not and unfortunately it is near-impossible to compete with other businesses that don't care about those things. At the end of the day consumers see "$15 hamburger" vs "$10 hamburger", the better/fewer ingredients, the better-paid labor, the recyclable packaging, etc simple don't matter. This is _not_ an argument to underpay workers, kill the environment, and fill your hamburgers with sawdust, it's simply a statement on how hard it is to compete when consumers very rarely reward companies for doing things "right" no matter how much lip-service they pay to it.

I've watched a close relative have people say "you should do X thing!" where X is something like "fewer ingredients" or "low calorie" or "healthy"-type things, only to see sales crater after an initial boost. People try it, they feel good about themselves, then they go right back to buying the thing packed with sugar.

There is a wide gulf between "what people say that want" and "how they actually behave". LLMs are currently in a category where people can have visceral reaction and claim "I'd never buy something from a company that does that" but I can tell you right now that won't last. Be anti-bad-design, be anti-bad-code but don't be anti-LLM because that opinion isn't going to age well (and yes, I know _you_ simonw, aren't anti-LLM, I'm speaking generally).

This.

I said it elsewhere in this thread differently but people want to pretend the options are:

1. Paying a local designer to make a bespoke beautiful poster

OR

2. Use AI

It's not, the money is often just not there to pay someone. So you get whatever the intern or the owner can slap together in whatever tool they know best (which is rarely the best tool for the job).

It's completely understandable that they would reach for LLMs to help improve their posters.

So many people are anti-AI/anti-LLM just because it is AI/LLM, not because it's objectively worse than the alternative. It's tiring.

It's the Electron vs Native argument all over again. Sure, we'd all love hand-crafted, beautiful, low-resource native applications for everything, that takes a lot of time and/or money. It's a decision between Native and Electron, it's a decision between Electron or nothing (or at best only 1 OS supported).

Just so I have this clear, you rate a restaurant badly based on the pictures of the food and not the taste/flavor/etc?

These places are doing their best to compete in a cutthroat environment with often razor-thin margins but you are willing to write them off and actively try to harm them for reaching for a tool to try and meet consumer's ever-increasing expectations?

Just like in the native vs electron debate it's not "Beautiful hand-crafted native app" or "Electron wasteful trash", it's "Electron or nothing".

I don't begrudge people at all for reaching for LLM/AI to help cover over what they are incapable of doing or unable to pay other people for. And while "anyone can take a picture" taking good food photography is difficult.

Ok, I'll say it. I like the new menu better.

I do not understand the author's aversion. Sure we can talk about "picture of food" vs "what was served" but have you been... anywhere? This is the reality of all food advertisement. I've never in my life been served a plate of food that perfectly matched the picture unless it was a user-uploaded image. Getting a decent idea of what the dish looks like is valuable (especially for unfamiliar cuisine, obviously not unfamiliar to the author but I'd wager unfamiliar to most patrons).

Being anti-LLM/AI for the sake of it is boring. Show me mistakes on the menu or a level of unreality higher than the food industry at large and I'll care but I'm not seeing the problem here.

KOReader support (not just KOReader sync) has become my new line in the sand for an e-reader. I'm currently using my jailbroken Kindle PW5 and loving it.

KOReader paired with BookOrbit is my sweetspot and I love having all the data under my control. If my Kindle died today I'd probably grab a Kobo and put KOReader on it.

This project is very interesting to me but they seem to want to "own" the whole software stack, it seems like it won't support things like KOReader? Or am I mistaken?

Yeah, and this is the direction I'm considering going. I just wish they had a subscription plan where I could still use the web search but also get MCP/API access included (not unlimited).

The last 6 months I've averaged right around 300/searches while paying for the unlimited plan (the lower plan allows 300 searches/mo which is just barely too small for me). I wish they provided usage data back further. I'd be interested to see if I was much higher in previous years.

I've been paying for Kagi for almost 3 years now but with the advent of LLMs (really in the past 3-6 month) my usage has fallen off a cliff. I really wish they included some level of API/MCP access so I'd like to continue to support what they are trying to do but every month when I get charged I think "Is this still worth paying for?".

I could just pay for their API directly instead of my subscription but I wish they'd meet (some of) their customers where they are.

I have not, I'll have to check this out. While I like my smart home stuff off my network (Z-Wave/Zigbee) I do have a soft spot for ESP32 (and it's own SSID/Vlan). I currently don't "manage" any bluetooth devices but I'll have to take a look!

4 color

I guess I should just get over myself but color e-ink (I have one) drives me nuts with its refreshing. At least the one I have can't partial refresh so it's a ~20-30 second flashing-hell every time you want to draw anything different. To be fair, I have these updating once a day and when I'm asleep so "who cares" but doing any active development/testing with one is maddening.

EDIT:

Holy crap, the ESP32 vs BT comparison is crazy! I would never have guessed it was that stark. Though it makes sense, I have mine deep sleep for all but <1min every day to wake and update and with config they should last months, but _years_ with regular updates throughout the day? So freaking cool!

I'll probably be looking into some BT screens. It looks a little more DIY than the reTerminals I have but maybe this is the final push to get me, back, into 3D printing (enclosures for the board/screen/battery).

I've been buying (2 so far) Seeed Studio reTerminal e-ink screens, flashing them with ESPHome, then using Claude Code to write backends to render out whatever I want to put on the screens (a la TRMNL, I also have a TRMNL). Being able to "speak" into existence what I want is pretty freaking awesome.

Note: Yes, I know what I'm about to say could be done using TRMNL (as part of it)

I was able to set it up so that it rotates book quotes every day with a picture of the book cover next to the quote. That, on its own, is far from impressive. What made me giddy was that it shows book quotes that I've highlighted in KOReader on my jailbroken Kindle. KOReader syncs up to BookOrbit (open source ebook manager/reading tracking tool). The whole thing happens 100% local to my network.

BookOrbit is running in a docker, my little server to render the book/quote runs in a docker, my Kindle talks to BookOrbit, my ReTerminal talks to HomeAssistant (ESPHome). It's 100% local and it's silly but it's mine and I love it. As I'm reading I highlight something and a day, a week, a month later I see it on the display on my dresser or in my office.

I want to put these all over the place in my house and I already modified it once so each screen will get a different random quote so I don't have to worry about duplicates. I'm considering getting 1 display per book shelf where I have a full series and configuring it so that it just shows quotes I've highlighted from that series. The sky's the limit.

I fully understand no one cares, I get it, this is super-duper-custom, "just for me", and that's what I love.

It's a shame because the Steam Deck is the best gaming device I've used in many years. My Xbox One and Xbox Series X are both lackluster compared to my Deck IMHO. And that's without me ever "docking" it, just using it handheld.

I won't buy another Xbox, the Series X was a massive disappointment and AAA games all suck. I found I was just playing decade+ old games on my Series X and while load times might have improved on the newer hardware, it wasn't enough to justify the upgrade. I regret my Series X purchase.

I think the jury is still out on K3/Q3.8 and if they are equivalent to Opus or Fable. Benchmarks have gotten incredibly murky and I've tried models that are "the same as X model" and been unimpressed. I've tried other models with Claude Code and with tools like OpenCode or Pi and nothing has really come close to Claude Code using Anthropic's models (mostly Opus 4.8).

I'm not saying these other models are trash, just that I'm not quite ready to put them on equal footing to Anthropic or OpenAI models.

I think the future of LLM coding (and more) is probably routers that decide on a per-task basis which model to route to. Know when to use Fable/Opus and when to fall back to DeepSeek/Kimi/Qwen, even all the way to local models. Of course that's not in the frontier lab's interest but it feels like there is a lot of low-hanging fruit there. I hate that right now it's pretty much "just use the same model for planning/execution/etc" (without standing on your head).

I'm not sure I could I have read a more detached-from-reality telling of the current OpenAI/Apple story.

Literally, this person has no clue what they are talking about.

Zero mention of OpenAI (ex-Apple) employees telling current Apple employees how to steal documents and prolong their time at Apple to walk out with as many secrets as they can.

And then pretending OpenAI wooed Jony Ive from Apple... That didn't happen, Jony left Apple many years ago (good riddance). His last few years at Apple were a disaster. OpenAI can have Ive and Meta can have Dye, both were net-negatives in the end.

The author phrases this whole thing as sour grapes for poaching talent. No, that's not what, appears, to have happened here. Potentially the higher-ups at OpenAI were ignorant to their subordinates were doing but OpenAI refusing to respond when Apple reached out to address the matter loses OpenAI quite a bit of credibility.

OpenAI didn't break an "unwritten code" their employees (if the reports are to be believed) engaged in corporate espionage.

That doesn't work with subagents/wakeup reminders/etc. I've done this then come back to it working again.

We really need a /pause or similar. I know I'll lose my cache benefits but sometimes you realize a task is going to burn too much, too fast and it's not the highest priority. You still want to do it, just not right now.

in a country where legal bills can climb really fast

Honest question: Are there countries where this is not the case? I'd be interested to read more about how that manage that. If it's some sort of "protecting the little guy"-type thing or a general suppression of legal costs. Or maybe I'm reading too much into your comment.

Small nit.

they did buy Jony Ive and are presumably giving him everything he wants to build a platform for them

If they hired Jony Ive to build a "platform" they will be very disappointed. He has no experience in doing that. They hired him to design a device, probably comment on the UI (if there is any, though I don't think he is qualified to direct either UI personally).

Aside from that, yeah, they royally screwed up here. Either by hiring unsavory people who think this acceptable behavior and/or by not managing/supervising them.

I've said it before on this topic: this goes _way_ past non-competes and the like. If you learn a novel method for doing something you are free (in my book) to recreate it at another company. You are not free to steal code/designs/etc verbatim and you are absolutely not ok to encourage people you are poaching (poaching is fine itself) to steal secrets/ideas on their way out. Also the whole "lying to a manufacturer to say Apple gave OpenAI permission to use the same proprietary technique" is really gross.

I'm right with you on how companies can hide behind layers in ways individuals can't, I'm still salty my address is public for on the Play Store, but the text you quoted "com.companyname.andriodapp1" is just the bundle ID. It can be anything, it doesn't have to have any connection to your company or your app at all, it's just a unique string. It's only convention that most people use a reverse domain name.

It's fun sometimes to see what that app was called before a pivot or what the internal name was since you can't change it (once published) without losing your users.

See also: LLMs or Uber (two of many such examples).

It's incredibly disheartening to know that one of the easiest ways to success is breaking the law and hoping it doesn't catch up to you before you have enough money to be above the law. That crime does pay. That lawsuits ruined people's lives for downloading MP3s and some of the biggest companies on the planet stole everything and walk away with a slap on the wrist.

This isn't an anti-LLM comment, it's just depressing how we live in a system with 2 very different sets of rules based on how much money you have.

It looks nothing like an Apple product and no, it's not part of the io/Ive partnership.

Fun game, I got up to level 5 before dying (with very little/no warning?). The mining so _slow_ to start and really not that much faster with a mining laser. I'm not sure if there are computer-controlled players or if it's only real people. Very few people shot back at me and while I hit level 5, there wasn't really any difference I could tell between that and level one.

Fair point, I just have oodles of cables (literally 10-20+ per combo of micro/mini/standard-USB-A/B/C <-> micro/mini/standard-USB-A/B/C each) and I like having them all organized by capability so that when I need one I can grab from right bin.

Before getting the tester I just kept buying new "known good" cables whenever I needed one since my cable drawer was just a huge unknown.

Because recycling takes relatively little effort for me and, aside from costing extra, doesn’t really negatively impact me. Giving up LLMs is a completely different story.

I’m happy to “do my part” but I go in eyes wide open and I won’t be guilted into doing it if it negatively affects me if there is someone/thing having a much larger negative impact without consequences.

You want to fine/charge LLM data centers? Be my guest. Ask me to stop using LLMs? Yeah no.

Calls to completely stop using LLMs are beyond absurd at this point.

Being anti-AI feels the same as being anti-war.

Ok.

This entire post seems to hang on climate change as the reason to be anti-LLM but nothing I've see seems to point at LLMs being any worse than literally anything else. It feels like the oil companies pushing recycling knowing that the individual effort of recycling does not begin to offset the damage done by corporations. This constant reframing as if it's the individual's responsibility (and fault) is tiring, verging on nauseating.

My use of of "AI" or lack of use will not influence anything. Period. To suggest otherwise is to ignore reality. It's the same silly arguments we hear about the USA/China "pausing" AI advancement, as if that's a thing that could actually be done.

I recycle and I'll gladly support bills/laws around forcing corporations to pay for their effects on the environment (even if that means higher costs for myself) but pretending we can stick our heads in the sand and ignore AI and that will have an effect? Pipedream. You are literally only shooting yourself in the foot.

Lastly, this reeks of the same thing as almost all "boycott" calls to action do, which is "This thing I don't like, let's boycott it!"... Ok, you understand it's easy to _not_ do something you _don't_ like right? That's not a boycott. That's just you living the same as always while feeling morally superior (trust me, I've been that person before). "FIFA is corrupt! I won't support them!", spoiler, I don't care for Soccer/Football so it really doesn't matter.

If I remember correctly, that application/your computer is simply reading what the device _claims_ it can do whereas this device tests it.

That app is probably a good place to start but I wouldn't trust it fully.

You're welcome: https://www.amazon.com/dp/B0DYJL5Z67

This device let me categorize all my loose cables (and throw out the truly terrible ones). It was worth every penny.

It [Finnish Company] says it agreed to pay an additional $5 million above the quoted price in order to avoid exposure to future price increases.

Initial delivery times slipped by two months and Dell then said that due to an increase in the cost of component parts, XTX would be required to pay an additional sum of between $62 million and $70 million. Dell has also suggested it might seek to impose further increases under the agreement between the companies.

Dell, in its responses, said it was not willing to lose tens of millions of dollars on the transaction. It said other customers faced with price increases had gone to other suppliers but had come back to Dell and agreed to the price adjustments.

Scott Hill said Dell expressed that it was “embarrassed” by the situation.

Is there any other response besides "Screw Dell"? Dell had no qualms taking the company's money to "lock in the price" but as soon as they realized they weren't getting as good of a deal they wanted to renege. You don't get to eat your cake and have it too Dell...

Also

It said other customers faced with price increases had gone to other suppliers but had come back to Dell and agreed to the price adjustments.

Just gross. "Other people tried to go elsewhere and came crawling back", I hope Dell gets smacked hard for this behavior.

Another HN post: https://news.ycombinator.com/item?id=48865019

Some pretty damning stuff:

OpenAI also instructs new hires on how to avoid scrutiny when they leave Apple. For example, Mr. Tan warns them not to tell Apple that they have taken jobs at OpenAI, so they can stay at Apple as long as they can.

Apple says it discovered a pattern of OpenAI recruits emailing themselves confidential information when leaving Apple, including Tan.

OpenAI apparently used confidential Apple hardware information when approaching Apple suppliers, and tricked one company into using a "specific trade secret metal-finishing technique" for an OpenAI device by claiming it had Apple's permission to do so.

Liu allegedly kept an Apple-issued laptop after departing the company and exploited a vulnerability to download dozens of confidential Apple documents while he was working at OpenAI.

Non-competes and the like are gross but what's described here isn't just "bring your expertise to OpenAI" it's "here is how to steal secrets on your way out" which is even grosser.