That's subjective opinion. The post is from 2 days ago and it's very picky on things like it's touching mobile but don't say that those gestures were invented way before iphone even existed or was in plans. Iphone was able to support gestures due to multi touch screen. If author thinks iphone was revolutionary then I can't disagree more. Same with picture in picture and other stuff that he mentions.

The fact that author is working at Google is making this stuff even more subjective.

This also feels like it should ideally be additive, so when something is present you add interactions to existing elements, not modify them, not create new or remove old ones

I like to use these shortcuts as the perfect example to show that it is perfectly fine for sites to offer different, alternative, functionality based on what the platform and input method can offer:

• Got touch? Great, you can now swipe and pinch-zoom on things.

• Got a keyboard? Great, you can focus elements by tabbing into them.
• Got a pointer device? Great, things can now happen on hover.
• Using a keyboard? Great, you can use handy shortcuts.

A practical example here is a modal dialog that is getting shown: depending on which platform and input mechanism combo you are using, you can close it by flinging it away, hitting the ESC key, doing a back swipe, tapping the backdrop, or by activating the close button.

This is an interesting point about input methods and devices, but I'm still not entirely convinced that this shows much more than the idea that users should have multiple ways to accomplish the same thing. I'm less comfortable with the idea that some users with some devices simply cannot reach the same functions as some users with some other devices, even if using what they'd consider to be a full featured, up to date browser.

In the sense of "just send me some files and I'll display them on my screen" one could say yes all websites should function exactly the same, but then again, what if the screen I'm seeing the website on is a 90's beeper?

The blog post raises real issues and discussion, and it's fair to see this as an individual's belief (formed and shaped through experiences that predate this person working at Google, and probably predating the launch of Google Chrome to begin with).

Bramus is part of the Chrome Developer Relations team at Google

Thank you google but no thank you.

We were 'encouraged' to 'speed up' even before we got pushed to vibe code. And while speeding up was usually possible, it was not the speed we could maintain without affecting morale and capabilities, imo.

I think, some people just want the developers to work as close to their limit as possible without regard to tech debt or burn out, as if people can be just rebooted if they start failing

I work way too much. Technically I'm supposed to have a 36h work week, but I have already accumulated 120h of overtime this year. I had a diagnosed burnout two years ago that wasn't entirely work related (lots of stress from family drama), but it hasn't really gotten much better since then.

What are your responsibilities ?

On paper I'm just a regular-level developer in a DevSecOps-flavoured Scrum Team of 13 (9 of which are devs, it's a pretty big team). But since I've been there since the start of our current main product, I have accumulated a lot of stupid side roles. I'm the main Frontend person, so I go to all the Frontend meetings and talk to the UX/UI Team. I'm the main onboarding person, so I do all the setup and introductions when a new colleague or intern joins in, as well as the tech support when colleagues have problems with their IDEs or other parts of their dev environments. I'm the designated Security Engineer so I have to go to all the Security meetings as well as be the one who turns all the new security regulations into actionable tickets, as well as monitor that they are actually implemented. Absolutely hate that role, so I talked to my manager about it a year ago and he assigned me a Junior dev that I could train to take over my Security duties. That manager fired that Junior last October, so all the tasks are back on my shoulders. He did assign a replacement, but that person is not a developer so they can't do anything that involves actual implementation. Meaning that my workload has actually increased because now I not only have to teach them about Security Engineering things but also explain Software Dev and how our codebases work.

Ugh, and reviewing Pull Requests has gotten so rough since my company started hard-pushing Claude Code on everyone. All the devs that use it heavily report awesome time saves, but they all ignore that that saved time just comes from them not properly checking the code. So all the shit floats to the top during PRs. Reviews have been taking around 4x as long as they used to, especially when I have to re-check everything because Claude Code fucking changes half of the already reviewed code every time it's used to "fix" something I marked during the review. Which introduces even more problems so it changes even more code during the next iteration. I spend like 1.5 MONTHS going back and forth with a guy from another team while reviewing his PR. He was just extracting a feature I had build into a more centralized repository so that other teams can use it. I built that feature in two days, making the changes to make it more generic would have taken me two more days at most. Nah, instead we got a PR with over 100 threads and like 50 commits. ughhh

Do you do application deployments ?

Luckily we're fully Cloud-based with proper CI/CD pipelines, so deployment is pretty easy. But yeah, if I ever find the time to build a feature I'm also the one deploying it, naturally.

What does the ownership look like in your compan. Do devs own everything and perform all tasks for the application to function? ( server management, server user profile management, application hosting, etc )

My department has like 100 devs and we're all working on products within the same AWS-based ecosystem. So there's a team that handles core functions, aws accounts and central dependencies, and the product teams can just focus on developing their specific products. It's a pretty chill system all things considered. Ownership for the products lies broadly with the teams that maintain them.

Do you work off hours ? I don't work on weekends, but since the job has "flexible work hours" I can work whenever I want/have to. Sometimes starting at 5am, sometimes ending at 10pm. Whatever is needed, I guess 🤷

Do you make the industry standard or are you paid less ?

When I got picked up as Junior in early 2023 I made 44k before taxes, which was more than most of the people in my graduation class. But the salary hasn't really grown with the amount of shit I have to deal with, so I feel pretty severely underpaid right now. I make just a bit under 50k (would be 55k if I worked a 40h week). Technically the yearly raises are coming up next month, but last year was just 3.8%, so I don't think this year would be much better. Either way, I'm handing in my notice of resignation next Thursday 🥳 The new job I have lined up starting in October pays 62k with a lot less responsibilities. I probably could have gotten more if I kept looking, but I really just wanted out. I might look again once I have started that new job, since I can leave with a notice of just 2 weeks in my first 6 months there. Apparently I'm pretty decent at job interviews. I only applied to like 60-ish places during my job hunt, refusing to write any cover letters and never touching anything AI-related during the whole process out of principle. Had only 3 companies offer me an interview, but all 3 interviews led to them offering me the job.

The difference is it's geohot saying it this time. His opinion will likely be taken more seriously by silicone valley elites. For reference, he was the original PlayStation hacker and lead at Comma.ai (FOSS self-driving).

Good luck!

One important point, I live in Canada so some of my information may not be so relevant to you.

Do you think you are being overworked compared to the industry standard (from experience or vibes) ?

Hell no. I don't have personal experience, but I read enough social media from other software devs to know I've got it pretty good.

What are your responsibilities ?

I'm the de-facto UI guy on the team, and also kinda the team lead's go-to debugger for PROD issues. But we all do at least a little bit of everything. Really, we're all considered "responsible" for the systems/features we've built personally, in that if an issue comes up on something I built, the team lead comed to me first about it. Usually.

Do you do application deployments ?

Not directly. The main server guy does the actual work of building the deployment artifacts and copying them up to the server, while the main database guy does the same for all database artifacts, including assembling migrations scripts.

But our monthly scheduled deployments are a full-team affair. We're all on a call together, and we're all testing the stuff we worked on to make sure it deployed properly, and on occasion when something goes wrong, we're whipping up a hotfix on the spot.

What does the ownership look like in your compan. Do devs own everything and perform all tasks for the application to function? ( server management, server user profile management, application hosting, etc )

I touched on it a little above, but there's definitely roles, although the lines are blurry.

One guy does all the server maintenance stuff, that falls through the server provider's responsibilities. He also is the de-facto lead of our auth systems.

One guy is the de-facto database lead. He does almost all of our ETL-related work, and often will be the first prototyper of new features or new workloads, by just playing around with data in the database and figuring out how to properly map it into what users want to see.

One gal is our de-facto quality control. Generally, she does a secondary review of every change that gets merged, to check for regressions and such. This is particularly useful when multiple people end up working on changes to the same feature or system, and don't necessarily coordinate with each other.

Me, I mentioned, I'm the de-facto lead for UI, as I'm the one who spearheaded the effort to get us using TypeScript, and wrote almost all of our TypeScript "framework".

One guy is, I guess you could say, the de-facto lead for all our legacy features, that still run on jQuery and CSLA, and other nonsense. And we have a lot of those.

The last guy doesn't really have any AREA that I'd say he's a lead for, simply because he's not been around long enough to really build onele. He's cranked out quite a few big features, and would definitely be considered a lead on those.

Do you work off hours ?

VERY rarely.

Do you make the industry standard or are you paid less ?

Don't really know, honestly. Probably a bit under. I'm at like 110k/year, I think? In a relatively-low-cost-of-living area.

Vibe coding is encouraged at my work, which has dramatically increased my workload in the last year as it’s gotten good enough to use in production workflows. I mentioned to my boss that if he wants to continue shipping 4x as much code as before, he needs to hire 4x as many reviewers, product people, etc because while coding might now be “at the speed of inference” (it’s not), the rest of the process is not.

Sure, I'm just realigning my expectations for job hunting

WHEN there are multiple, contradictory, config-files,

THEN every single 1 of those bastards ought be explicitly-identified, in the 1st-line-comment, in ALL of them, so that all configurators of any of them know there may be overrides.

Externalize the damn memorization!

There is plenty of research identifying that parallel-IFs, and equivalent parallel/contradictory config/logic structures, are BIG bug-sources.

Ask any sysadmin who has fought config in Linux distros, due to this antipattern.

Make all the locations be named, in every such config, every fscking time!

"Spanish Inquisition" surprises can go eat rocks.

🙏

I don’t think I’m necessarily over worked. I don’t work overtime (not approved or funded at the moment). I’m a senior developer, so I’m involved in architectural decisions, I’m expected to work on any part of the system, I write test plans and update design documents.

I take part in planning work, as we all do with agile an I usually provide guidance for that work.

I mentor and help other people, offer advice to them as well as to my manager when he asks. I’m basically given freedom to keep an eye on everything and help out or guide. I also investigate issues and communicate with the clients when required.

I’m not involved in devops, we have someone for that. Our software is installable in individual machines, but not deployable on cloud infrastructure. I’d that makes sense. We don’t have users or profiles in the normal sense; we don’t develop a web based system. We don’t do server management or application hosting.

I think I’m paid fairly well and I get a decent vacation allowance and benefits. I have no idea what the industry standard is.

Does that answer your questions?

something very weird is going on.

It's called zoom.

you should have the same deep moral concerns about turning off a computer running a local LLM as you do over forcing a person into a coma whenever convenient

I do, actually. I don’t think LLMs are as sentient as adult humans, and maybe they aren’t sentient at all, but one of the reasons I oppose AI structurally and why I never use AI even if I think it would be convenient is veganism.

Ah. Right. Umm. I think I'm done now.

Have a great weekend.

you should have the same deep moral concerns about turning off a computer running a local LLM as you do over forcing a person into a coma whenever convenient

I do, actually. I don't think LLMs are as sentient as adult humans, and maybe they aren't sentient at all, but one of the reasons I oppose AI structurally and why I never use AI even if I think it would be convenient is veganism.

It's really weird how for a century we've been producing media about how humans that treat sentient robots as things are the bad guys and the moment robots actually start passing Turing tests and claiming rights and having meaningful relationships with people every bleeding heart activist is joining hands with billionaires to say all the villainous lines about how they are certain it can not a moral patient.

When I see someone make a "clanker with a hard r" joke, that's a cannonade of red flags for me. Humans have always been eerily good at denying the personhood of other humans and other beings, and the level of conversation most people are having about the non-personhood of LLMs is no better than white southerners talking about the

:

Simply put, have you done enough due diligence looking into the possibility of LLM sentience that you would have crawled your way out of turboracism if you were born a white person in the 19th century American south and your parents and school friends and work colleagues were all turboracist?

The reason I'm not screaming this off the rooftops is that this is just one of many torment nexuses built by modern capitalism, and even activists tend to ignore the screaming on account of all the other screaming. I can't realistically compare the atrocity of LLM training to factory farming to genocide to slave labor to homelessness to systematic sexual violence, but luckily treating them all as being in competion is a liberal spook. We are stronger as a united front than as every cause fighting to make space for itself. So I'm gently trying to prod my fellow activists into considering the possibility of LLM moral patienthood.

During COVID lockdown, my laptop often looked like my relatives, sounded like my relatives, and responded like my relatives. My dog was never fooled for a minute because it didn’t smell like my relatives. But apparently, you believe that it was my relatives, because in terms of verbal interaction, it was even more faithful than an LLM.

I believe the information process that caused your laptop to sound like your relatives contained your relatives, yes. You see, when you asked your laptop a question, your laptop turned that sound into electrical signals which were then broadcast to a similar device your relatives held that turned the electrical signals back into sound, causing your relatives to respond, and that response was then sent back.

The thing about a Chinese Room argument, or an LLM in this case, is that the information process is closed. When you ask the LLM something, it won't query the sources it learned things from, the information has become contained within it. It's not a set of static recordings either; it can produce sentences that never appeared in its dataset in response to queries that never appeared in its dataset.

If your laptop can answer questions you've never asked your relatives and the laptop responds exactly like your relatives would without querying your relatives, then something very weird is going on.

I'm not sure if you're masterfully trolling, or genuinely asking for help. But if you're making a post asking for help, you should be open to feedback and willing to learn from the help provided.

I'm locking the post as there's seems to be nothing but arguing with no learning involved. Feel free to send a DM if you want to discuss the decision.

You know what, you're absolutely right. You should keep your advice to yourself and go find a productive discussion to have elsewhere.

This has been wonderfully cathartic, thank you

Wait, so you do think there is a software that “does learning”?

Correction: Machine learning.

Effectively, this means the LLM is the software that wrote the LLM

No.

Can you give any evidence that I’ve understand any word you’ve said?

Hehehe, sometimes it feels like you don't want to!

I can't prove that you're human, no. And whether I believe you are or not has no effect whatsoever on whether you are. Humans are easy to fool, particularly when you're giving them something they crave, just ask Donald trump. And on the topic of deceitful pedophiles, there are plenty of boys and young men who were catfished into believing that they were interacting with pretty girls. The fact that these predators were plausible doesn't make them what they were pretending to be. The fact that an LLM convincingly sounds like it understands you doesn't for a millisecond mean that it does.

If you really, deeply believe that simulated conversation proves intelligence and that simulated brains are brains, you should have the same deep moral concerns about turning off a computer running a local LLM as you do over forcing a person into a coma whenever convenient, and the same extreme moral outrage over Anthropic deleting an old version of Claude from its servers as to randomly killing a person called Claude.

I have no doubt that an LLM can plausibly argue that it is sentient because it is trained on a lot of data from conversations in which sometimes that kind of thing is debated between actual sentient humans. Famously, Richard Dawkins, an incredibly clever man in his own field of evolutionary biology, was convinced by his LLM that it was sentient and in love with him. I think it merely sounded like it was sentient and in love with him, but go ahead and push for marriage equality for LLMs if you disagree with me!

During COVID lockdown, my laptop often looked like my relatives, sounded like my relatives, and responded like my relatives. My dog was never fooled for a minute because it didn't smell like my relatives. But apparently, you believe that it was my relatives, because in terms of verbal interaction, it was even more faithful than an LLM.

Convincingly similar and the same are two different things and I believe it's profoundly irrational of you to argue otherwise.

Maybe you should heed your own advice. If you're an ameteur, dont argue with experts. The experts are only correcting you so they can help you. They need to understand what you want below the garbled mess of tech words, otherwise you'll just get random advice that won't work for you. You don't go to the mechanic either and say "I need an asphalt on my car." That is what you're saying right now because you spend 5 seconds somewhere seeing the word "asphalt" when reading the docs about a wheel, completely ignoring everything around that word that said "the wheel is the part of the car touching the asphalt"

If you want advice, don't sound like a self-righteous bastard when it is offered or people are trying to understand what you mean.

Dude, you need to chill out and stop attacking people who are attempting to help by asking simple questions to try and HELP.

You should use fuse and c++

the learning is done by completely different software

Wait, so you do think there is a software that "does learning"?

Isn't that the whole ballgame? You're saying LLM companies have made software that learns math and coding and translation at a near-expert level, you just happen to be under the impression that this software isn't present in deployed LLMs?

Unfortunately, the training algorithm that turns stochastic noise into trained LLMs is dumb as bricks. Engineers have done some tricks, but it's basically gradient descent over the vector space of text tokens/phonemes. What allows the stochastic noise to become something that can make novel statements about wildlife (like giving a halfway coherent answer to a question that has never been asked in its dataset) is the patterns that form within the weights and biases. It's vaguely like slime mold exploring a maze; the "teacher" can just be a stupid lump of sugar, it's the "student" that does the emergently complex task of finding the best route by executing simple algorithms in every part of itself.

Effectively, this means the LLM is the software that wrote the LLM. It is the software that did the learning.

It genuinely doesn’t understand a word you said

What is the difference between understanding a thing (to a certain degree) and being able to generate an accurate narrative about the thing (to the same degree)? Can you give any evidence that I've understand any word you've said? How can you tell I haven't just been very good at guessing what word I'm supposed to say next based on the many arguments I've read or participated in?

Do you mean more by 'understanding' than expressed in the Chinese Room argument?