I'm definitely aware, first-hand that AI can't do anything on it's own. You listed many tools. They sound impressive in isolation. However, you're just repeating what the AI sales people tell us....I'll give you new, shiny tools that'll make you SOOOOOO productive. Mark Zuckerberg says his AI can replace a mid-level coder. So I use one daily and it sucks (Claude 4.0). I've dabbled with ChatGPT and CoPilot and they were far worse...OK, so maybe I'm the issue and I suck...well...there's still no results...just promises and tools being sold. Anyone can promise an amazing useful tool. It's far more compelling to point to a business making money putting these tools to use...right now, the entire AI industry is just people selling you AI tools, it seems.

If vibe coding is the way of the future, why isn't software shipping noticeably faster? Jensen Huang is promising that an LLM can do what takes me a year in a weekend with a few prompts and no one will ever need to learn programming. So....where's the proof?

Why aren't my favorite publishers pushing out releases so much faster? OK...maybe they're the dinosaurs who can't keep up...where are the upstarts disrupting the market? If an LLM can replicate an existing program, where's the salesforce.com of AI-generated apps? Where's that upstart that will generate your custom ERP for you in a fraction of the time it takes to implement Peoplesoft, SAP, or salesforce.com?

I (think) I want AI to succeed, but I am personally disappointed because there are SOOOOOO many obvious use cases for a working LLM. For example, my favorite game...why can't an LLM generate new levels for me?...OK, maybe it requires supervision...why isn't my favorite publisher pushing out a new level a weekend or even a month for $5-10 each level? Why isn't some upstart releasing a tool that can take my sloppy legacy code written in COBOL or even Python/Java/JavaScript and porting it rust that runs faster than I could have ever imagined? Why isn't it optimizing my legacy code? Or...a service that promises to patch all vulnerabilities for a hefty monthly fee? It's SOOOO profitable and so easy if these things actually worked.

I think you and I know the answer...there's no useful LLM services tackling real world problems because they don't work. It's a fucking license to print money if they did. For now, all we get are shiny tools that may or may not work...with no guarantee. 3 years is an eternity in tech. We discussed Netscape in 1995 to massive commercial internet in 1998...another example is the iPhone being released in 2007 and how much the market exploded by 2010. There were many revolutionary mobile apps by 2010 that disrupted many markets and created new ones. Also, AI is ANCIENT. They were talking about this in the 60s. I had machine learning textbooks in the 90s. Google had a working LLM in 2018.

You mentioned legal recognition. That's based on theoretical AI. Theoretical AI is AMAZING. Real world AI? Not sure...definitely not useless...but...on the flip side...what is it accomplishing? I can see some disruption in creative industries....but entertainment is a small industry that allows mistakes....a Java compiler doesn't. The frustration is we're getting promises and so many people are using their imagination to picture what AI COULD bring...but then I try it out and see it falls far short of what I imagine (and I have very realistic, low-ambition expectations) and it falls short of what every tech executive tells me it can do.

As much as people are frustrated at Apple for Apple Intelligence, I respect them. They put the cart in front of the horse and announced this...assuming ChatGPT could keep up with their promises of innovation...they tried it out and saw it failed hilariously....and let's face it, the stuff they promised wasn't very ambitious and Apple has an attrocious history with software, so their standards are not very high...1st rate hardware and AWFUL software (think iTunes on Windows). Even Apple, with their billions and partnering with the industry leader couldn't make a useful product with it....so yeah, I am skeptical until I see it solve real-world problems...not just sell me toolkits and tell me to figure out a way to make it useful.

When people want to stop aging, few want to live longer, they just want to enjoy the time they have more. I'm right at the Gen-X/Millenial line. Aging sucks...even in your 40s. I think most don't want to live to 110. I think most want their 50s and beyond to not suck. I'd personally love to be a productive member of society until I die...instead of feeble and a drain on my loved ones.

If an effective anti-aging treatment existed, our economy would be in much better shape...lots could work, if they choose to. We could spend less on medical treatments. The elderly could be fit and active, reducing their chances of all sorts of diseases. I'd rather see grandparents die from recreational activities than degenerative diseases.

I started my career during that fad. It wasn't the complex things you mentioned...a simpler explanation is that India doesn't have infinite programmers. Wages rose and Indian professionals are no longer 1/5 the cost of an American. The same talent shortages in the USA applied everywhere else. But similarly, those low-cost programmers made the project a lot more expensive for the reasons you mentioned...as well as the biggest factor IMHO...anyone you want?...they're talented enough to find a job at Google/MS/Netflix/OpenAI/etc...who's willing to pay a lot more than any offshoring firm. You get someone great at first and they're quickly poached.

Everyone is scared of AI taking their job...legitimately so...but think about it from the pro-AI side...where's their victory? You brought up games. Where's a game that was built with AI or a maybe a great game that was built in record time by a small team using AI that looks like a AAA game? Why aren't games shipping faster? If the big studios haven't embraced them, then surely upstarts would be building some mind-blowing games that were built in a few weeks that look like they took many years in a AAA studio.

3 variables? So which is it?...the sugar? the trans fats? the processed meat? What does processed meat mean? Hot dogs?...no problem...but what about tinned fish? what about smoked salmon? "processed" is a non-specific term.

I eat psychotically healthy. I am around people who eat like shit. I honestly don't see any correlation. I've had days where I was on vacation and ate like shit...again..no correlation. I didn't feel better or worse, beyond heartburn...which I honestly get more from eating healthy than eating some absolute garbage. I can feel a massive difference in drinking or not drinking, but junk food vs eating fruits and vegetables and lean protein?...nope...not really.

I think there's a stronger correlation between stress and health...and poor people tend to be under greater stress and eat shittier. The older I get and the more experience I have eating healthy, I am starting to think the difference between junk food and healthy food is very minor. Quantity matters more than quality...yes, junk food is easier to overeat with, but if you can avoid that, I honestly can't tell you the difference between the weeks I eat like shit because I am staying with relatives who live off junk food or at home when I live off salads and fruits and lean meat or whey protein.

When I was young and poor and eating very mediocrely...I felt guilty. I couldn't afford lots of fresh fruit and vegetables and had to eat a lot of processed calories because I was going to college and working multiple jobs...because I was constantly on the run, I sometimes had to live off stuff that didn't require refrigeration and when I did eat, I'd often eat something high in calories because I was broke...so would go longer between meals. I drank sugared soda as a means to go longer between meals.

I apsired to someday be able to eat healthy. Once I graduated and had a consistent schedule and general stability, I ate perfectly...no more sugar in drinks, barely ate dessert, stopped eating chips, etc.....didn't experience a single difference whether my meal was a cold cut sandwich or salad.

Today...whether I eat perfectly or eat like shit...can't tell the difference. I'll wager that these differences would disappear once you account for socioeconomic status. I will bet those who live off hot dogs daily have a lot more stress in their life than those who never eat sugarey drinks or processed meats.

By your logic, all code would be written in the poorest parts of Asia or Africa. Why does anyone hire software engineers in Silicon Valley? NYC? Seattle? etc. It seems like the most expensive cities in the world have the most programmers. When money is on the line, cheaper is not better. All it takes is one customer breach to kill your sales for a long time or open you up to some very EXPENSIVE lawsuits. In your example, a time card system...what happens when someone breaks in and tweaks their hours? When real money is on the line, it needs to be highly functional and secure. Also, shitty code written by shitty programmers is much more expensive to maintain. The difference between an expensive coder and a script-kiddie/vibe-coder is that I know how to solve the problem with 1/5 of the code and in a manner that is easy to maintain 5 years down the road once I've left the project...and it won't get hacked.

That's also why people hire me in a very expensive American city rather than some village in India. That's why my employer and nearly every other one grabbed every semi-talented programmer from every rural village in India and shipped them to the most expensive cities in the world.

You're technically correct that people care about profits...you just did the math poorly...you didn't factor the costs after something debuts. The same applies to AIs

100% of your code is generated...OK, did it actually work? Can you teach me how to make LLMs generate working code? Claude 4.0 only generates code that compiles about 50% of the time for me.

OK, so what is it good for? ChatGPT was released 3 years ago. Think of how much the internet advanced 3 years after Netscape Navigator 1.0 was released. What is something commercially valuable these LLMs can actually do that we can objectively see? (no, Mark Zuckerberg promising his developers are so much more productive with them is pure bullshit...as he has presented no evidence and based on subjective "vibes").

The wealthiest companies in history have poured trillions into and hired the best minds and turned it loose on the collective public imagination. What do they have to show for it? Is there anything we can objectively measure?

A WORKING AI can do a lot of things cheaper than MANY jobs. However, nothing I have seen can reliably generate Java that compiles...like obvious, basic syntax mistakes, like missing braces or semi-colons placed randomly. Java may not be your jam, but if can't do that...what can it do? That's an easy use case and perfect for AI.

I use Claude 4.0 daily at work with a mandate by my employer...their quote "those who don't embrace AI will be replaced by programmers who do." OK, kewl....I want to be more productive and spend more time solving problems and less time looking up how to do obscure things with common libs. I give the AI lots of tasks, ranging from "write a unit test for this" to "optimize this complex for-loop" (when I see it can be converted to lambda). In every case, I can see there is a solution, I just want to be lazy and I don't want to have to remember how to use groupingBy or setup a unit test and think of the obvious edge cases of nulls. I've tried it with CoPilot, ChatGPT, Claude 3.7, Claude 4.0. I think maybe 20% of the tasks I've given it worked. Java is a strict compiled language with LOTS and LOTS of open source code to train on, not to mention it's a dominant language in computer science education for many many years. The code compiles about 50% of the time. It often can't even match braces and will introduce semi-colons in the middle of statements.

OK, so you don't give a shit about Java??...fair point...but...as mentioned earlier....if it can't master the simple rules of matching braces, how can it do anything useful?

If a human being is being paid to do something, it needs to be correct. No one would hire a truck driver with a 99% accuracy record...that only crashes the truck 1% of the time. So to my knowledge, beyond scamming people, there's no industry that pays someone to get things kinda right.

Today's LLMs are not useless, but they're very close to it for all modern tasks...and there's little evidence of dramatic improvement on the horizon. The pace of innovation is EXTREMELY slow. ChatGPT was released 3 years ago.

Take the internet, in contrast. Netscape Navigator 1.0 was released in 1995 (technically Dec 15, 1994, but no one used it until 1995).

Think of how much the internet advanced between 1995 and 1998!...you went from something that NO ONE bought anything on to suddenly every retailer was online and there was an active e-commerce scene...in fact, it destroyed the music industry shortly thereafter....and in 1998, anyone could easily see rapid improvement.

We don't see that with LLMs. We see improvements with fake benchmarks, but there are SOOOOO MANY EASY EASY EASY things that a working AI could revolutionize. If this was the real deal and not INCREDIBLY OVERHYPED...we'd be seeing it actually take over industries outside scamming the elderly. We'd be seeing amazing new programs written by AI...or maybe even existing ones made a LOT faster and leaner by AI optimization. Think of how many creaky old neglected programs are running around that a working AI could make faster.

Beyond software, there are just so many things an AI could easily do...so I turn the question back to you...what have LLMs actually done today? What is one useful thing they've taken over? And don't say make people more productive...that's subjective and people have a STRONG incentive to tell their boss that shiny new toy they paid a fortune for is making a difference....something we can objectively prove.

I think your statement doesn't hold up. I wish it did, but how does canceling science programs or vaccine programs help him or stoke his ego? Scientific research is something most leaders, even despots, take pride in. He was very proud of warp speed. If your statement was correct, I would imagine him barely lifting a finger and taking all the credit for what's happening.

He's actively sabotaging the USA and I don't know what he gets out of it. I know what Putin and our adversaries get, but not him or his family. For his major, yet baffling, decisions, I can't see a path between those and financial enrichment or prestige for his legacy - short term or long-term.

Honestly, most of Trump's actions do not make much sense based on his stated goals. However, they do strongly align with Putin's goals of a weakened USA. I don't know what's up with Trump. I'm sick of guessing his motivations. However, as many have stated..."I don't know if he's a Russian Asset...but if he was, I can't imagine what would be different." The USA has a long history of investing in science and research and most Americans benefitted from the results...directly or indirectly...including the working class which voted for Trump...lots of construction and manufacturing jobs created either directly or indirectly from this research.

His actions are out of line with the needs of the working class, out of line with traditional Republican policies, and I don't see a path to how they improve America in any way. He has great ideas hidden in there, like reciprocal tariffs...but his implementation was so poor and unserious, that they didn't come near to accomplishing the stated goals. I have given up explaining him. I just can't wait for his reign to be over and the one bright side is that he's a narcissist and only interested in his own glory and being treated like a king...so this movement will likely die with him. His kids lack his charisma and from what I can tell, he has not groomed any other successors...he doesn't want to share the spotlight.

Way to let me down! :( ANY answer but "she's wearing pants" would have been far more fun and make me like her more!

Well, things are always about degrees. Both a bicycle and a motorcycle are transportation tools...we regulate a car differently than we regulate a bicycle or a motorcycle. Same goes with nukes vs dynamite.

In a way, I view this like recreational drugs...in the hands of someone with their life together, drugs aren't that harmful. If they have severe depression, it's a recipe for addiction. I have a friend who is in love with her chatbot. She was a functional person on anti-depressants and in therapy 3 years ago...now she literally is in love with the chatbot and obsessed with it. Make no mistake, the chick has issues...always did...but with such a personalized experience, she's now largely stopped engaging with her husband, misses work...alienated her friends, etc. It's not ChatGPT's fault, but her loved ones are VERY concerned.

Is it an issue?...not really sure...it's curious...worth examining, but I would gather data first and react second. I don't think there is anything all that magical about generative AI...it's definitely a quantum leap forward in personalized interaction...and as others have stated, it will love you when no one else will.

Sorry, your old boys network view is very dated. Make phone calls? WHO THE FUCK IS PICKING UP THE PHONE TO TALK TO STRANGERS? No tech company will hire you. But hey, who wants to work at Netflix/Google/Apple/Microsoft/Amazon/OpenAI/nVidia/etc? If you're coming to the job market not willing to adapt to modern times, I question your work ethic. You're not virtuous...you're the laziest fuck I've ever heard of...shaking your fists at people doing what is expected of them today instead of what people did in the 80s and early 90s. No one cares about effort, only results...working smarter is always better than working harder.

Also, you're assuming people aren't asking everyone they know. Everyone I know is. Here's the thing...you only know people you know...and people in top jobs?...they don't socialize a lot....big tech employers do what they can to isolate their employees. It's not like the late 90s where you'd meet them at conferences.

Everyone I know from top employers I know by accident, or they're former coworkers....I was very active in the open source scene local user groups and conferences...I met a lot of people from shit employers there. Once I got a job from a top employer, they stopped paying for me to go to such things and I had to work on proprietary frameworks, so the stuff at conferences and user groups became a lot less relevant....like everyone else who upgraded their employer, I stopped going.

Translation: We're not failing...we're becoming more agile! We either have too many people or are expecting reduced revenue, so we need to do some layoffs and cut hiring...but instead of this being bad news which could affect our stock price, let's instead say we're DOGE-ing the shit out of this company and becoming super high tech and automated. We're not correcting for macro-economic forces, the majority of which are far beyond our control...nope...no mistakes on our part...we're riding the AI wave to glory!!!

Unless SF has some amazing AI 1000s of times more accurate than anything the world has seen, there's no way the current AI can vibe code enough to fully hand off any task. So either they hired an army of useless people in the past with no real-world function...or they needed to cut headcount + added some AI here and there and are intentionally misleading investors and staff to imply the AI is why they can operate with less employees...so nope...not a layoff...people...please buy our stock!!!....grads...please apply to work here!!!...current useful employees....don't seek work elsewhere, we're not a sinking ship!

Marc Benioff is famous for being a lowlife and Salesforce doesn't have a reputation for being a top shop in the industry. They're known as "cloud IBM"...but IBM today, not in the pre-Internet era...so NOT A COMPLIMENT.