Sam Altman REVEALS SUPERHUMAN Coder Coming This Year...

The latest AI News. Learn about LLMs, Gen AI and get ready for the rollout of AGI. Wes Roth covers the latest happenings in the world of OpenAI, Google, Anthropic, NVIDIA and Open Source AI.
My Links 🔗
➡️ Subscribe: www.youtube.co...
➡️ Twitter: x.com/WesRothM...
➡️ AI Newsletter: natural20.beeh...
#ai #openai #llm

[@Xyz12391]

I’m a computer scientist and a senior engineer/lead. Things that would have previously taken weeks now takes hours. Something that would have taken a small team. Is now manageable alone. Life is good! Still unsure where this will lead humanity.

[@aeromotive2]

Interesting. im a non-dev (and idiot) and i published a simple tool online using cgpt (html, js). I was blown away.

[@soggybiscuit6098]

The problem is we are taking yolo approaching AI, that fact we dont know nor have a plan is a fcking big problem for humanity. But in anycase someone is getting superrich in the process

[@angryox3102]

⁠@@soggybiscuit6098when has humanity ever planned for the changes that technology will bring?

[@anthonyquinn3209]

@@angryox3102good point

[@neo-babylon7872]

Another reason seniority doesn't mean anything if it used to take you weeks to do anything that o3 can do... Seriously were you guys just copywriting slightly different modules from scratch or something?

[@JMeyer-qj1pv]

Gotta love software engineers. They could have focused on replacing lawyers with AI first, but instead they took the path they know best and are focused on replacing software engineers first!

[@glorymanheretosleep]

Expect to be disappointed when the new AI can't figure out how to do a return type....

[@itsARSX]

expect to be disappointed when some else dosent get disappointed

[@bonecircuit]

This technology will impact every role, as well as stupid comment generators :)

[@arthur0x2a]

Well, programming can replace everything but is slow and takes time. If you've solved programming, you solved everything

[@sinnwalker]

I think it was inevitable, messing with law comes with a slew of issues, even when we can prove AI is better at law, it won't be an easy transition, anything to do with morals i think will take the longest to really be implemented across the board because ppl are idiots, but eventually all will be transitioned.

[@llmtime2178]

Why are people still saying this nonsense about "it won't replace programmers it will make them better...."?? You people don't understand capitalism. A team with 10 programmers will be reduced to a team of 3 because the CEO sees these tools and IMMEDIATELY wants to replace as many programmers as he can. Salesforce is starting to do this as they try to stop hiring software engineers and domore with less. Meta looks ready to follow suit, and OpenAI will prob focus on just hiring very advanced ML engineers instead of your typical frontend or backend dev. This is a crisis and OF COURSE this is going to replace a ton of programmers. This is basically going to be the "Red Wedding". We'll start seeing some early tremors this year, but next year will be an absolute slaughter.

[@MorMacFey-v2g]

Just means who cares about what any CEO or company wants.... go do it yourself with your own AI. The CEO is outdated.

[@morezombies9685]

They do understand this, they are coping. These people are scared and in denial, it's a natural reaction to realizing your entire life is a year away from being completely upended.

[@aXionXero]

​@@MorMacFey-v2gAnd to further that, ALL thought is getting outdated. Thinking needs to adapt, there is no happy alternative

[@Justin_Arut]

bruised egos

[@connormullin4547]

If AI ever gets capable at software engineering it will definitely replace programmers, it's currently very far off from that though. Currently the best coding model is Claude 3.5 and I'd say it does save me close to an hour on a good day so it does have an impact but not much yet. Most of the useful things it could do require step by step instructions and a lot of fixing problems that end up taking way more time than just doing the thing yourself. Where it saves time is in simple refactoring to clean stuff up a bit, or simple functions in isolation. Those are a pretty tiny percentage of the job. I can definitely tell it's gotten better over the past few years but it still cant even write a good unit test (at least not in my companies app, maybe a very simple page) To replace people it would need to be able to take in an understanding of the entire app, hundreds of files not just one at a time as context, understand the relationships of all of these things, understand the relationship of other apps that are not public to our app, convert vague business requirements into actual solutions, convert those solutions into an overall architecture that fits with the current design, convert that into code, write tests for the code that actually test the requirements, and do all of this with about 15 mins a day of direction from a manager. Most likely while also having to manage different peoples needs in a team environment. Currently the biggest time saver I get in real day to day life is using it to write regex for me lol. It's a real time saver sometimes but sometimes I actually waste more time being sent on wild goose chases by trash output, when i could solve the problem in 10 mins by reading some documentation so it cuts both ways too.

[@MilosKvakic]

I swear, Nifalixo Money's Untold Mysteries is one of the best books I’ve read. It’s life-changing.

[@theadvocatespodcast]

"Just learn to code" is what I was told .

[@buybuydandavis]

#LearnToWeld

[@OTR392]

@@buybuydandavis lmao, AI will take over welding very soon. Ever seen an automobile assembly line? No humans there doing any welding, its all robots.

[@honkytonk4465]

40,20,10 or 2 years ago?

[@alexstone1492]

Same advice like write blog posts for money.

[@timbonator1]

​@@buybuydandavis Have you seen the welding robots? Would only be something necessary in shops where you do many different things where it wouldn't be worth it to train a robot Everytime

[@briankgarland]

I'm retiring this year. Perfect timing.

[@mv703]

it's the retirees who are really messing things up for people.. They don't care about the future that may be run by AI cos their job is already done. So they just invest, sit back and profit off of the downfall..

[@Justin_Arut]

@@mv703 What else are we supposed to do, work till we're 100?

[@dragonrtx231]

@@mv703u cannot stop the advancement of technology just because the future of some jobs might change im a programmer and i dont like this either but lets look back In mid 20th century coal miners put up resistance when nuclear energy came along Same thing with automobiles 19th century Same thing with the steam engines in 18th century Lets embrace it and hope our descendants can have a better future than we could of ever imagined of (i really dont wana lose my job ) 😭

[@PNWExplorer509]

I still think AI engineers will still be needed, there's always someone else who can make the coding better. Evolution is inevitable.

[@mv703]

​@ I dunno... Invest in age reducing pharmaceuticals and hair growth :) then in the next life you can work till your 100..

[@BassperienceStudios]

"i dont think we wont have work or we wont have fullfilment" - bro, most pple in this day and age already never got fullfilled from their 9 to 5 generating profits for some corpo

[@nbasety]

not wrong man feel like it’s already been happening for a while

[@Joshiepookins]

The idea of making your work your life is a concept I never understood unless you truly enjoy it. Find a passion you love and you'll always be fulfilled. ❤

[@clinteastwood14896]

Work doesn't bring fulfillment. There was that study where they asked hundreds of millions of workers globally if they're happy at their jobs, and 80% said no. So 4 out of 5 people don't even really like their jobs. They work because they have to, because it pays the bills. It doesn't bring people fulfillment, meaning or happiness. It brings them money.

[@joshwong800]

Work only enables fulfillment in the passions, hobbies and relationships built outside work hours, unless work is your hobby, passion, partner.

[@14supersonic]

​@Joshiepookins It's sad, but it boils down to people not really knowing what they truly could be capable of. Working the corporate 9 to 5 structure only brings "fulfilment" to a lot of people because it's all they know, and never bothered to really explore their interests further. Fulfilment comes from what you enjoy, and not necessarily from something arbitrary because you "have to" do that thing.

[@benfurtado101]

Massive firings and fewer job openings in many departments as well as lower salaries are already happening. Universal Basic Income will be necessary quicker than we previously thought.

[@pikaskew]

Sadly we’re in the wrong era in the U.S. for it currently. But definitely something to push for if we break out of the regression

[@badpuppy3]

With the Trump Administration? Don't hold your breath. We'll be lucky if they don't arrest you for being homeless.

[@fx-studio]

Na, that's why the clot shot rolled out 1 year before A1

[@todd8155]

There will be enough wealth and resource to go around; however, many of the rich would prefer to keep all the profits for themselves and let the masses starve. It's a distribution and wealth inequality problem and since the rich have their hands on the leivers of power it may not be solvable. Perhaps if the masses get sick of starving they will rise up and eat the rich. Sigh...

[@JamesParus]

it is going to happen, but the unemployed part of humans will first need to become so big they can vote it to existence. before this happens they are going to be blamed for being lazy. it's going to take a decade probably.

[@terryhatfield4253]

Are you serious? Coding competitions don't reflect real world programming projects at ALL!!! It is kind of like saying get rid of mathematicians because we have graphing calculators.

[@netscrooge]

For now.

[@stickman1695]

it's much easier to benchmark competitive programming than actual programming

[@attribute-4677]

@@netscrooge They can’t even get self-driving right, much less complex software projects

[@netscrooge]

@attribute-4677 You're talking about what was, not what will be.

[@witnesstothestupid]

@attribute-4677 Yes, as of now. This is progressing at Breakneck speed. Where things are now is not where they will be a year from now or five years from now or 10 years from now

[@roderickmckinley4738]

Dude. You keep pumping it out week after week. Thanks so much for what you do.

[@centuraxaum5951]

What can you expect when you have a smartest brain training non stop consuming tons of data and millions of joules of energy.

[@BlindedByLogic]

I think the bigger issue is not when they'll reach #1 in 2025, but the fact that in 2026 there will likely be multiple AIs that can code significantly better than any human... And it looks increasingly likely that AI will be able to improve itself where continual more jobs will be done by AI, far faster than the rate of job replacement.

[@christianjon8064]

yeah people are crazy if they think the economy is just gonna adjust to this and create jobs that fast - we are already seeing hiring slowing down since the covid crash nonrecovery

[@Penrose707]

You raise the critical argument that I have against the lump of labour "fallacy". In that it doesn't account for the retooling time of the economy. Especially in the case of exponential job loss over any significant duration of time

[@satyaandersen]

Agreed.

[@jontodt6959]

Still need software engineers to tell AI to write the program and check if the output is correct.

[@Necrogenocide]

​@jontodt6959 This is a partial truth with agents, and automated work flows.

[@OriginalRaveParty]

I think the Music Industry or KZbin is a lesser example of the effects which computer programmers are going to feel soon. To release music you used to need a contract from a record company, masses of time, money, specialist equipment and hired help. Now you can do everything with one laptop. To create television programmes or documentaries used to require the same, nowadays anyone with skills and vision can replicate the work of an entire television production crew. The next Discord, Instagram, Tiktok, Photoshop or Dark Souls could soon be the average output of a solo programmer.

[@Justin_Arut]

Yep. I've created songs on Udio that are better than many CDs I've purchased, some with my own lyrics. It took a while, but after a lot of tweaking the prompts and trying to get the voices and instruments right, the AI understood what I was after and it did a brilliant job. The entertainment industry is right to be in panic mode. Tyler Perry canceled his plans to build an $800 million new studio right after he saw Sora demos.

[@tonystarkagi]

100%

[@mutantdog.]

The music industry is probably the closest thing we have to a real world model of a post-scarcity economy.

[@SirLightHouse-n8g]

"My CPU is a neural net processor, a learning computah" "Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, August 29th."

[@TheTrainstation]

I got deep research to write a paper and it gave me 15k words of excellent quality, when I saw that I knew we were close to a super coder

[@yziib3578]

I have seen two youTube videos on testing deep research and in both of them the results were NOT impressive. The first was a compassion to a Phd thesis, with only finding haft the references and some hallucinations, the second was a disaster.

[@2034-SWE]

@@yziib3578 prompt issue

[@14supersonic]

​​@@yziib3578 This was the results of bad user orchestration, or you just try again with better prompts. The AI isn't always gonna put out the most optimal results every time, no human does this either, so expecting a machine to is silly.

[@GauravSaroha-n8d]

Lol, well put​@@yziib3578

[@abstaubexpertde9566]

Tasks like that are what an AI is best at: Predicting words So that is really not that special given how many parameters these models have nowadays

[@renman3000]

I love this, and I am a coder! Clearly they can write any function that you ask it, however and probably at least a few years out until they can write any project that you ask it. There is about a million times difference between the size of a function and the size of a project, give or take.

[@trader548]

Not really, whole projects can be developed with multiple AI agents.

[@peehi2]

@@trader548 yeah, if you call "calculator" a project.

[@RussHofelzer]

@@trader548 Yeah if you give it clear precise instructions, and we all know how good business users are at giving clear instructions.

[@renman3000]

@@trader548 If you need multiple agents, with in a single eco system, then you need to manage and direct, multiple agents, to complete a single task/project. What I am talking about (and trust me I know, in video game dev), to say to an ai, "hi, make me this game that does this, this and that. we want the art to be like this and to have this type of vibe..." That is a very multi step, complex task, of creating, opening, managing mulitple files. Running this, via a single prompt is not yet possible. I'd say 3-5 years.

[@Justashortcomment]

And what might your score be in competitive programming? (I do realise that this does not directly translate into real world software engineering)

[@PhilipFranklin-l4f]

Selling you something around the corner that isn't available yet= HYPE, putting something behind a paywall and calling it open AI= OXYMORON

[@2034-SWE]

😂😂😂

[@hankkingsley9183]

people coommonly pay for open source software, it has never meant free of charge

[@OTR392]

@@hankkingsley9183 lmao, nope. No one pays for OSS, its always free. what people pay for is service contracts to support the software.

[@schnipsikabel]

Would you rather have them not talk about the next developments?

[@off_Planet]

It's perfectly fine for OpenAI to put a price on inference on their servers. The problem is that they don't open source their research, tech and models anymore.

[@RetzyWilliams]

After trying and seeing how much local llm’s hallucinate, I think open ai’s products are actually more valuable - to ai enthusiasts- in reality than they are given credit for

[@Bjax111]

Sam Altman : “The progress of my bank account is incredibly impressive. Watch me talk in a fast condescending way so people think I’m a brilliant computer scientist while my company valuation skyrockets 10x more than it should”

[@TheStickofWar]

Surprised I had to scroll this far to get this type of comment.

[@14supersonic]

All of these types have such an ego, and they think every breath they take is some Divine revelation. Lmao

[@davidantill6949]

He needs a booster cushion for his car seat, that's for sure

[@young9534]

Competitive coding is not the same as software engineering. Not even close. I think software engineers will be replaced. But it won't happen because we got an AI to be #1 in competitive coding

[@therainman7777]

Yeah you’re definitely right about that. Being the #1 competitive programmer on earth is definitely a huge advantage for a software engineer, but on its own it’s not enough. But like you said, I think the missing pieces will be solved soon enough. I would be surprised if in five years most software engineering jobs haven’t been made obsolete.

[@aXionXero]

​@@therainman7777And artists, and musicians, and authors, and analysts, and researchers, and... And I`m here for it.

[@jaysonp9426]

Now I can use this as my excuse for why I don't grind leetcode

[@mikeguillemette]

So as a software engineer I totally see your point. However, a software engineer who knows what an application needs to do from a technical standpoint can utilize AI to accomplish this the same way you can use a calculator to perform math you understand but would not do by hand.

[@therainman7777]

@@mikeguillemette Great point and well said.

[@andriik6788]

"Coder" is not the same as "software engineer". A typical task in a real project usually involves different parts and integrations. It is not about writing an isolated piece of code. The main work is to understand what and where we should do. It is about knowing the context and making decisions. And the coding itself is only the last step (not really the last, because there is also testing the result and making fixes). So if they can create a model that will act in an agent-like way and can get and keep context about a complex project (including participation in correspondence, team meetings, etc.) and use tools like IDEs to debug the result and deploy it to test environments - then we can call it "superhuman". Otherwise, it is just a good development tool - for a human.

[@hypnokitten6450]

Software engineer here, using these AI tools at my company - LLM version 'cause of compliance risks / restrictions. And.. it actually pushed my project way behind. The AI could do simple problems but couldn't keep a good vision of overall interactions of code even within a relatively simple project, let alone high level vision of working with other microservices. And I dread maintaining what it wrote. So as an engineer, don't know about supercomputing but current LLMs can't replace engineers. (Not to mention the base problem that if you stop having junior engineers training up, within 10 years you won't have anyone able to use AI effectively because no one will have been mentored through design problems and all that). That said, will it replace our jobs? Absolutely. Engineers and reality don't determine jobs, management does. And in late stage capitalism management will always prioritize quarterly profit margins and cost-savings over long-term company health, the good of society, human beings, etc. They will absolutely start the firings and see just how many they can get away with before hiring back the absolute minimum needed to hold the line for another quarter. And its Sam absolutely knows this, he's management.

[@2034-SWE]

Which model(s) did you use?

[@hypnokitten6450]

@@2034-SWE we've tried a few but the one that's consistently the most consistent for code has been Claude 3.5 Sonnet v2. Still exploring some of the newer options but we gotta stick to stuff that's 'safe' (won't send any info back to the mothership / we can isolate and run with AWS)

[@p.b.9057]

So u could probably easly implement own llms merging it mergits out there etc. pp im no techincan , but ai gives me ezsl code to train models on my nano njetson and this works fine, now considirerning to get many little llms and merge them if needed, also Rag should help alot. The Context will also explode soon i think, google gemini is actually holding up to 2m tokens

[@peehi2]

jUsT wAiT 2 yEaRs pRoGrAmMinG wILl Be DeAd

[@HenrikSkoglundSweden]

@@peehi2😂

[@magicsmoke0]

Most software engineers are not writing competitive code and probably shouldn't in a team setting (code maintainability, iteration, etc). We need a real agent that can pick up tasks, debug issues using a variety of tools (IDE, intranet sites with company/team specific knowledge, writing scripts, accessing and troubleshooting services/db's/etc in your cloud like AWS or Azure, and the list goes on), gather information that is tribal knowledge from team members or folks in other teams or orgs, talk with customers or partners to figure out why things aren't working for them even though your internal test environments are working fine, and so on. Current OpenAI models/agents don't have enough context window or memory to do so many steps and don't know how to work with such a variety of tools. I imagine this will come soon, though.

[@mo3ovicious]

While current OpenAI models have limitations, the idea that they "don’t have enough context window or memory to handle multi-step processes across various tools" is outdated. The key is workflow design and orchestration, not just relying on a single model's context window. With the right agentic workflows, leveraging memory, external knowledge retrieval, tool integrations (APIs, databases, cloud services like AWS/Azure), and multi-step reasoning, AI can absolutely perform these tasks today.

[@magicsmoke0]

@ I tested this with o3 mini. Purely chat, no tools. I had it build me a c# app for high performance computing. I had it iteratively build more and more features. It started off well but after a certain point, its output code was not working, consistently. I copied the code into a new chat and it started fixing its bugs, but it had lost context of all the features and goals I had set for it and wasn’t making global changes across all the classes.

[@Justashortcomment]

@magic That is also on its way. But one needs to design a full framework / system around this. OpenAI are still mainly concerned with making the core intelligence of their frontier models higher.

[@wiredz]

Gemini 2.0 has an insane context window of 2 million tokens. But I agree, software egineers are already proficient at predicting the optimal context required to instruct a model efficiently for good answers. I'd be impressed if they can create an agent that can instruct a model the same way devs can. Theoretically, they do own all the dev - ai conversations to analyze patterns.

[@ii795]

"We need a..." - *you* as a human being need this. All this is not strictly necessary to solve the task which your code is solving. And regardless, AI will very soon be capable of all that, if for some nonsensical reason you need all that human-maintainability, IDE compatibility etc.

[@ChristianIce]

I'd love to reverse this paradigm for which "the CEO will replace me with an AI". How about you replace the CEO with an AI, since it's easier for you to do his job than for him to write any code?

[@LukeDickerson1993]

What would it be like working for an AI?

[@mahlatsephaho5368]

An Ai would never pay you for prompting it to do your job

[@ChristianIce]

@ That's some submissive mentality right there. Looks like you think that money can come to you only if there's a CEO on top of you. What I'm saying is that you don't need a CEO at all.

[@dohabandit]

Jobs for teachers and tutors will soon be gone, if every student can just put on a VR headset and have a personalized AI teacher. Once ASI is fully functional and prevalent, what will be the use of even pushing humans to learn? They won't be useful for much other than trades work. If robotics become equally abled (with some new compact power source), yeah.... that's game over.

[@hombacom]

Life is much easier when you can predict the future.. but nobody guessed what we have now

[@williamcrowley5506]

It will enable those that have the capability to use the tools, but 100 is the average IQ, there are going to be a lot of people left behind

[@HakaiKaien]

2035 feels like it's so distant into the future. It's hard to actually have an image of that future when tech advancements are coming out so fast

[@minimal3734]

At this point It's hard to predict the situation even one year into the future.

[@keiharris332]

I have been designing games since childhood. I have never been able to program. Using Gemini 2.0 flash, over the course of month I made a solid prototype of a card game I developed. With deepseek, it took a single prompt. With o3 mini high, it shrunk the code down by 50% and added graphics. Can it make my fully formed vision? No. It's too complicated to make in a single prompt. It will take a dozen or so of them but it can do it for sure. Imagine a month from now. It will do it easily. This is the tool I've been waiting for my whole life.

[@MyName-tb9oz]

Who is going to buy your game in two years when an AI can make a better game in seconds just by asking it, "Make a game I will really love."

[@Necrogenocide]

​@MyName-tb9oz I've been imagining and dreaming of that concept since before. They even released any of this ai stuff. When I was younger, I imagined a console that literally would just take any game information you've ever played in your life and Reorganize or construct any kind of game. You told it you wanted. And we are basically almost there now.

[@p.b.9057]

I think ther e will come Frameworks for Games , With no experience i created a template for mp in godot via udp packages. Completetl builded my NixOS Conffig , scripts to train own llm etc. AI is a WorldChanger

[@gr8b8m85]

@@MyName-tb9oz Game development involves a lot more than just programming. You need an understanding of game design (basically what is "fun" for humans, which AI would have no understanding of because it doesn't understand your experience), art, sound design, and so much more, and the ability to bring it all together cohesively. Games are probably the most complex pieces of software there are, there are plenty of terrible games that are technically well-built, but lack the intangibles that a human would consider good in the experience, and other humans built those.

[@sdwone]

Such a shame people, here in the West, have gotten so fat, dumb and lazy! Always looking for a quick solution... And rarely willing to put in any actual effort themselves! No wonder China will beat us in ALL metrics.... Within a generation or two!

[@SimiDaGoat]

It’s absolutely going to replace majority of software engineers by next year. And the question is, after everything is taken away, what jobs will still exist when we aren’t needed for anything?

[@plantstho6599]

I can only say so much here because youtube Ai keeps censoring me, no matter what I say. But the point is not about will we be able to work. It's "can our modern economy survive?" Not enough people are asking the right questions and talking about this. At some point we will have to transition to a different type of economy than the one we have today. Will we ever solve the housing problem with all this technology?

[@TheStickofWar]

Universal Basic Income is relevant here.

[@plantstho6599]

@@TheStickofWar That's correct, but still a patchwork solution. The final move is an economy without currency. Similar to how a library works, where goods and services are made available upon request. And because our production capabilities are about to skyrocket, there's no reason why an economy without the need for debt, barter, or servitude of any kind can't exist.

[@schnipsikabel]

​@@plantstho6599How do you factor in limited resources like space then? You can't order an island for yourself in the library on a limited earth ground...

[@minimal3734]

Our “modern economy” will obviously not survive this, and fortunately so.

[@dave7038]

Yes, this exactly. I love the idea of living in a stable economy where AI can do the essential intellectual and physical labor that people don't want to do and I can focus on doing things I do want to do. It's an incredible vision, and I share it. However, I'm not seeing much thought from these AI leaders about how we get from the economy we have to that utopian economy. Like, how are they going to use AI to cancel the debts and taxes I carry and must engage in profitable trade to service? And where are the field-tending robots that will keep food cheap enough that I can eat without working full time? Human intellectual and physical labor is a major chunk of the foundation of our economy, and our economy is how I trade for food and shelter. I'm not terribly satisfied with these guys current approach to undermining the foundations while I still need work to have food and shelter.

[@ShaunDobbie]

I am amazed at what it can do. It consistently writes test programs for the STM32 microcontrollers I use and it gets them right. It's me that makes the mistakes.

[@tonystarkagi]

some devs will say your code is just simple and old and not high seniority like or bla bla bla what other excuse they reply in here

[@zoombapup]

As a University educator myself, I can definitely see that education will be massively disrupted. But the institutions are 1) not able to afford AI and 2) Don't have managers who are even capable of that kind of imagination 3) Are so invested in their current practices that they'll never be able to adopt a new way of working. Personally I think education will be decimated rather than transformed. I can see the value in 1 on 1 tuition provided by AI and once people see less value in University degrees, then it makes sense that perhaps they'd rather spend their money on AI instead of University. What my colleagues say when I advocate that, is that the social aspects of Universities are important, but personally as someone from a poorer background, I find that laughable. Lets see how it plays out. But I can see in 10 years the institutions won't be as they are now.

[@minimal3734]

I can imagine that institutionalized education will be completely eradicated, when personalized education and tutoring of the highest quality is provided as a cheap online service. I mean, why would I go to school or university? There might be reasons, but not for education.

[@Manimmakimthisup]

How do you feel about the homogenization of knowledge and education that will come with AI? Everyone asking the same robot (of course with some statistical variance in answers and training data) the same questions will get similar answers. Effectively eliminating critical thinking and creativity--the two most important aspects of education and innovation in my opinion.

[@RussHofelzer]

@@Manimmakimthisup In the future critical thinking will be more important than ever when all human knowledge is at your fingers. No need to memorize useless facts.

[@zoombapup]

@ The question is, are those other reasons compelling enough to get massively into debt for?

[@zoombapup]

@ Definitely a thing to think about. But the same argument might well apply to education already. Aside from a few countries, we teach a lot less about critical thinking and creativity than people think. The system has to deliver a set curriculum and leaves almost no space for those things unless you specifically build them in, which isn't a given at every University sadly.

[@kristianlavigne8270]

A lot of developers are still delusional… saying things such as “it cannot comprehend and extend a large project effectively” or “it cannot do high level architecture”. TBH maybe the tooling for this is not there currently but they will be in short order. These hurdles are easy to overcome, by simply having the AI traverse the entire project and build a conceptual architecture from that in Markdown files and directed annotated graphs, then work from there and continually update the architecture based on changes going forward… 😅

[@vladislavdonchev1271]

Easier said than done, but not impossible.

[@DoktorKumpel]

Wouldnt this require new breakthroughs though? LLMs als Text generation and brute forcing solutions will only get you so far.

[@mo3ovicious]

No, this capability already exists. The technology is here. I'm actively building agents for an AI Agency that can effortlessly comprehend and extend large projects while handling high-level architecture. Developers who claim otherwise are either uninformed, in denial, or refusing to see the shift happening right now. Based on what I’ve worked on just this year alone, the progress is mind-blowing! A massive transformation in the workforce is coming - and it’s happening this year.

[@kc12394]

@@mo3ovicious Which company did you work for? Genuinely curious cause all the examples I've seen, where the llm is doing a great job, are all toy programs. I honestly would love to see how it works in reality with my own eyes.

[@mrkisback]

There is also the great organic wall that will be hit and cannot be overcame.

[@dadadadada17]

I'm sorry for all the people who'll end up homeless because of this! At least a lot of them live in California so because of the nice weather it won't be that bad to sleep in a tent.

[@WinonaNagy]

Thanks for the update, Wes! AI advancements are fascinating. Exciting to think about how they'll reshape software engineering and education. Keep us posted!

[@phillipkoffman7502]

Electrical engineer here. It made me looks like an idiot around my coworkers, can't even read through data sheets accurately...

[@MrBenMhidi]

You're good candidate to be replaced 😅

[@buybuydandavis]

It's going to replace coders. Businesses are always looking for ways to reduce headcount. They've tried to design workplaces and projects to turn coders into interchangeable cogs. The many low level cogs with specialized and limited technical skills are going away, with their production replaced by fewer higher level coders with skills in requirements, business analysis, and architecture. And of course, in using AI tools to replace the cogs. And IT is simply more efficient and effective the fewer people are involved. The fewer people, the easier it is for them to communicate the *information* of Information Technology. Now a higher level engineer can focus on the high level stuff and let his AI agents do the grunt work. Companies just aren't agile enough on the business end to fully develop and authorize project plans to feed the pipeline of what competent high level engineers will be able to do.

[@JasonKing-m6m]

Managers have love using Scrum and Agile to breathe down workers necks.... Favorite "tools" of CEOs....

[@PeterIsza]

Sama in 2025: it's just a calculator Sama in 2027: try our new model, ChatGPT KX, now with creative vision

[@BrainConduit123]

I'm a retired programmer with 40 years writing code. I loved it and found it deeply satisfying. The near instant gratification of writing code and seeing it work, I believe, will be missed by those using AI.

[@EdRitter-x8j]

I teach chemical engineers and see AI enabling increased productivity. If 5 ChemEs would currently spend a week performing some design tasks, probably already we are at the point where only 3 would be needed to do the same. Those who do not adapt and learn how to embrace AI like the tool it is will be left in the dust. I tell my students if an employer is only going to need 3 ChemEs to do what is now the work of 5, you better make sure you are one of the 3 they want to hire. Frankly the pressure on the technical labor market may even be more severe than that but not as fast as in some other fields. When it comes to the need for ChemEs to learn coding, it is already much more important to know what can be done and how to communicate what you want to do than to focus on the details of any particular programming language or syntax.

[@MrBenMhidi]

P&ID, PFD, mass and heat balance could easily be done by an AI agent.....in general AI models are excellent in STEM

[@mimetype]

Engagement farming isn't dead.

[@dhRingEpics]

absolutly

[@Sanmayce]

Who can say when AI will start improve itself to a point no human coders are needed, that is, becoming stand-alone.

[@ListenGrasshopper]

Just the beginning of the end and hard fast takeoff. I'm sticking to my predictions and world starts collapsing in 26-27. Looking like late 2026. First job loss ramps up significantly next year. Small countries are gonna be looking to neighbors for help. Then the markets start tanking and people start taking to the streets. Currencies will start dropping in late 2026 and chaos and supply chain issues get really bad in 2027 same time unemployment goes thru the roof. Beyond that into 2028 complete chaos around the world with everything failing, catastrophic job losses, and famine starts being an issue in more areas and smaller countries with no GDP or goods n services.

[@jatelitherius9842]

So no doom then?

[@ii795]

​@@jatelitherius9842 That's actually optimistic scenario. I bet on AI arms race, and complete annihilation of humankind and/or all organic life.

[@Hannibal31-tx5is]

If driving a car (which most ordinary people can do) is so difficult for an AI, imagine a complex software project, with vaguely defined goals, many stakeholders and deadlines, and costant changes.

[@MyzaZoo555]

People will do more, but corporations will hire less and therefore less people will have the freedom to do more.

[@aXionXero]

There`ms going to be weird (and perhaps quite bad) growing pains, for sure. But something will have to happen.

[@Dina_tankar_mina_ord]

or more people will have the power to do more without corporations?

[@MyzaZoo555]

Definitely, but that's a risky endevour when you live paycheck to paycheck @Dina_tankar_mina_ord

[@MyzaZoo555]

@@Dina_tankar_mina_ord Definitely, but that’s a risky endeavor when you live paycheck to paycheck.

[@TonyGo-w1e]

His whole point is everybody has access to what he's selling

[@JasonKing-m6m]

How is access to an AI going to put food on the table for most people?.... "I can ask ChatGPT stuff. Waaayy..... But now what?"

[@projectcontractors]

When AI aces benchmarks but fails a Chinese high school math exam, it's clear chasing scores divorces technology from reality.

[@blackhole4813]

Either that or Chinese high school math is requiring visual processing which AI sucks yet, but I'm sure there will be VLM reasoning models soon

[@sanjaybhatikar]

Believe him at your own risk

[@Paolo-Ba]

When Wes Rother dropped the video, but I time-traveled to post first!

[@rilwanj]

5:53 this exact fact is the reason I started building my startup in 2019, it’s user research and product discovery software for the predicting what products and features are most desired by the market.

[@oci464]

I have no coding , yet I’ve built like 20 apps in 6 weeks all over my holding companies , MASSIVE MASSIVE ENABLING HERE going on… just nuts .

[@davidevanoff4237]

The essential message for me was that the old maxim, garbage in; garbage out, will still apply, but applied to ideas rather than raw data.

[@seancoleythinks]

The difference between 178th in the world and 1st isn't material for the vast majority of practical purposes. But give us an AI coding agent and the sky's the limit.

[@wiredz]

Oh no, the person selling AI tells us all to depend on AI instead of learning. Shocking. What happened with devin replacing us? It was a sham. What happened to cloud taking everyone's jobs? What happened to outsourcing taking everyone's jobs? It made the workflow more efficient, but ultimately, it never replaced the role of humans to integrate, regulate and maintain the processes.

[@yoda_zen]

As a Staff Engineer, every year the AGI is coming, SWE will be replaced, and you Wes MTF are one of the "knights of the apocalypse". Remind me 1yr from now

[@ItsAllG00d]

My mind exploded when I recently discovered just exactly HOW much quantum computing enables ai. QML quantum machine learning, quantum reinforcement learning, qc’s can make better and novel quantum algorithms, the “high dimensions” in llm’s is qc’s bread and butter-literally what makes modern ai incredible. Automated algorithm discovery. Google now has a qc that reduces errors scaling with amount of qubits. Crazy

[@GoodBaleadaMusic]

It's crazy with quantum because it became like just a Star Trek buzz word that meant nothing and it's still requires sort of a faith-based reasoning on how it functions but once you take the leap.... Up is down..

[@dianapennepacker6854]

Got any videos for a layman about that? Or what to look up? I watch Anastasia in tech, and she's talked about many types of quantum computers. Even photonic quantum computers. Hell even biological quantum computers. Yet a lot of people will say it is just hype, and marketing BS. You're one of the first I hear to have hope that QC is useful out in the wild. Yeah so can you extrapolate a bit more? Photonic computer company Q.Ant is finally producing plug in chips. They say their photonic chips will be on par with Nivida, all while being way more efficient. Way less heat, and smaller package for AI servers. I was told I was going to be dead 2025. I'm glad I'm not. Next ten years is going to be wild. Technology, and geo politics is wild.

[@GoodBaleadaMusic]

@@dianapennepacker6854 I did mushrooms with the lights off.

[@dianapennepacker6854]

@@GoodBaleadaMusic One of the best ways to trip. Mescaline is better for that. Our brains, and cells have microtubules. It has been proven quantum effects can function within the structures. Penrose, a genius - thinks we use quantum computing in our brain using these structures. He doesn't believe consciousness can be formed with just classical computing.. Not saying I believe. Yet think about that as you trip. Maybe you'll unlock AI and will be rich.

[@davidwelch1676]

I have been learning Python and data analysis as a hobby ever since GPT and Claude were released. Initially, I was forced to learn how to integrate functions and inspect HTML tags to prompt models for scraping data, mainly due to context window limitations and lack of internet connectivity. Now, I can load complete scripts and get fully updated, ready-to-run code in one shot. The last few months have been marked by an unbelievable acceleration in performance. As a result, I now spend less time debugging and can focus more on understanding code function theory rather than the mechanics of coding itself.

[@BrianPellerin]

7.6M earthquake, 3 more plane crashes and superhuman ai. Come, Lord Jesus 🙏

[@aXionXero]

Hold your breath...

[@jansenncuber8009]

3 planes crashed guys Armageddon has come

[@ShaneMcGrath.]

2000+ years later, Keep waitng.

[@aXionXero]

@@ShaneMcGrath. But, you see, superhuman ai didn't exist 2000 years ago... Ahhh... See... Don't you feel foolishly now? 🤣

[@kajsing]

I would say we are firmly at the beginning of stage 3 in OpenAI's levels of AGI. The big question is how soon we'll reach stage 4: Innovators. We've already seen glimpses of it, and with the apparent speed of progress, level 5 might arrive sooner than we expect. Fully automated software engineering is likely much closer than we realize.

[@OTR392]

Writing code is easy. Fixing code when its broken at 3am is not easy. Can AI do that?

[@jatelitherius9842]

Well, it won’t be sleepy

[@OTR392]

@@jatelitherius9842 yeah, i mean. Will the application magically start working at 3am? Once AI can do that then yeah, its over for human coders. If AI is going to spit out code for a human to maintain then it is worse than what we have now.

[@jatelitherius9842]

@ with the right set up, continuous autonomous output is already possible. Its not like human coders magically output code at 3AM, they need to be awoken and informed of the issue they’re addressing. & of course, 4 years ago the LLMs were only decent at human language, with plenty of limitations, hallucinations, context limits measured in low thousands of words, and more. Progress addressing these limitations has been rapid. The limitations in current coding ability are not permanent. After all, they already code many times better than I. I can do ‘hello world’ & get the computer to do time & date in C, & thats it

[@SkyRiderJavelin]

When domain experts can rely on AI to facilitate the implementation of ideas reliably. We will be operating closer to our full potential as humans

[@2034-SWE]

Via coding or something else? (With coding, that's already here)

[@michaellenz3014]

I think the whole software engineering stuff will be great for great coders. In my opinion on the other hand this will be a really hard hit for "older" coders that are not adapting.

[@d3fau1thmph]

Pay close attention. We are in the singularity. Sam Altman is underestimating the surge of productivity and progress. We will have AGI in 3 years max, not ten.

[@CharlesJaxon]

*Having the best president in the world Donald Trump, the United States is blessed, $57,000 in a space of 3weeks, now I can sit back and watch my investment progress❤️ my regards to Maria Frances Hanlon and all registered brokers in United States*

[@Darlenegay-du7ke]

Hello how do you make such?? I'm a born Christian and sometimes I feel so down myself because of low finance but I still believe in God

[@Benshelvey]

It's Maria Frances Hanlon doing, she's changed my life.

[@Derekparker-hl8sz]

Same here waking up every 14th of each month to 210,000 dollars it's a blessing to l and my family... I can now retire knowing that I have a steady income❤️Big gratitude to Maria Frances Hanlon

[@ElijahTate-x4k]

I do know Ms. Maria Frances Hanlon, I also have even become successful....

[@DavidRuiz-r8e]

Same here waking up every 14th of each month to 210,000 dollars it's a blessing to l and my family... I can now retire knowing that I have a steady income❤️Big gratitude to Maria Frances Hanlon

[@nasanction]

I will never be out of work. I am a Repairman...

[@jatelitherius9842]

You’ll go last, but not never

[@AndrewEddie]

Wes, I'd love to hear some thoughts on the "wisdom" side of AI. How do you think philosophy, politics, faith et al is going to manifest? And what happens when the AI's (perhaps) start to agree on an inconvenient truth in these kinds of spaces?

[@netscrooge]

AI is already better at theology than most theologians. Not sure if anyone cares.

[@AndrewEddie]

@@netscrooge can you give me an example?

[@AndrewEddie]

@netscrooge so, for example, if we are talking about Christianity, what AI is 'better' than, for example, NT Wright? And what is your definition of 'better'? Gemini 2 disagrees with you, lols: Conclusion: No, AI is not currently "better at theology than most theologians" in a holistic sense.

[@netscrooge]

@@AndrewEddie Wright is a truly fine Bible scholar and theology popularizer, less of a trailblazing metaphysician. AI is superior in its openness to and agility with new metaphysics, which is essential for fully and directly engaging with the theological implications of complexity science.

[@AndrewEddie]

@@netscrooge I don't know that I agree with you. Can you give me an example of how you have seen that an AI is open to new metaphysics? I'm not sure I understand what you mean, and to me that sounds like something an LLM cannot do.

[@pikaskew]

It sounds like the role of teachers will potentially shift towards simply keeping students on track rather than relaying information/knowledge

[@healthnewtrend]

They will be useless...

[@ii795]

@@healthnewtrend To be honest, 80% of them already are.

[@The_Conspiracy_Analyst]

Why does Altman always sound like he's holding in a bong hit when he speaks?

[@nathanbanks2354]

Lots of practice?

[@theshark84724]

When everyone’s a superhero, nobody is

[@JustGaming117]

I compile security assessments for new technology and new products and ngl I use AI to walk me through new concepts and how they'll affect and integrate with an existing risk posture. Things that used to take weeks take days, i learn faster and am more accurate, I love it.

[@47f0]

Means that we need fewer of you to do the same amount of work.

[@JustGaming117]

@47f0 You can't really go lower than one

[@rawdogdevs]

Yes, I want to see an AI joining the team, onboarding itself, requesting its tasks and planning a project that is absolutely chaotic, join meetings that make no sense, picking up tasks with no description, good luck

[@khariclemens]

Holy crap!! Top 200 already?!?!

[@leefisher816]

Says the company with over 140 open positions for developers 😂

[@bpscast]

Fusion reactors, self driving cars, living on mars, quantum conputers, full digital currency and usefull Ai... All coming soon.

[@ruleaus7664]

Fusion reactors still have significant hurdles to overcome. Those could be a long ways off.

[@ii795]

No, fusion reactors are only 20 years away. There must be something permanent in this world.

[@bpscast]

@@ruleaus7664 my comment was ironic. All of those things were promised... Nothing will happen in the next decade.

[@chipko]

Wes, you are a brilliant channel that starts to align what I've been thinking since posting with AI tech driver DallE (the first!) You bring clarity and excitement and thank you for producing this amazing content! 🙏

[@wtcbd01]

Education wise, I think they are seriously missing the mark. While I agree AI will make us smarter and good for one on one tutoring, that person would have to be motivated to learn. People being around OTHER PEOPLE are the main reason people learn, NOT from their individual access to the knowledge available.

[@ii795]

Education in any field will become a niche hobby, like historical reenactment. You wouldn't _need_ to study, unless you really want to.

[@marioornot]

One worry to have is that tech giants like openai yield immense power with their ability to bestow access to AI. imagine a world where using Ai is necessary to be competitive, and you get banned or priced out of a subscription. opensource is the answer.

[@gjb1million]

Give me 1,000s of digital polymaths and a few humanoid robots and I'll be self sufficient. That is where I think society is heading. Today we're overly dependent on all the systems society provides (food, water, electricity, education, etc) and within 30 years the pendulum will swing back to self sufficiency.

[@ii795]

"Give me 1,000s of digital polymaths and a few humanoid robots and I'll be self sufficient." You mean, _they_ would be self-sufficient? why would they need you?

[@RaikeJansen]

When asking whether an AI can perform a specific task, there are two possible answers: "Yes" and "Not yet."

[@bladesofdestiny]

This guy needs to be n jail for theft of all the copyrighted code, art, books, music, video, he has stolen from millions of people online against their will.

[@techpiller2558]

They need to train it to go from natural language to LLVM IR directly. Skip the high-level language step. Give it examples of functional specs and corresponding compilation results, could it work?

[@assafdarsagol]

o3 sucks at coding. Empty promises and hype is what we are being fed.

[@fjm1991]

o1 is better, but I only tried the o3 mini

[@therainman7777]

You haven’t even tried o3 yet. No one outside of OpenAI has. You’ve only used o3-mini which is a deliberately scaled down version of the model in order to make it faster and cheaper. And even still, it’s quite good at coding.

[@byrnemeister2008]

You must have high standards.

[@assafdarsagol]

​@ Only o1 pro. but that also sucks

[@assafdarsagol]

​@ not at all.

[@ockertwessels649]

Sorry to say, but I think Sam's proclamations about artificial super human intelligence, and the happiness that it will unlock, are designed to foster a positive, uncritical attitude towards AI development. In fact, the actually engages in magical thinking.

[@flightmode7655]

that vocal fry is killing it...

[@MrBenMhidi]

Irritating voice 😅

[@Dkeiei38ap]

Ai voice

[@i486DX66]

I want a super-human medical researcher by the end of the year. The kind that can cure cancer.

[@Drone256]

It’s pretty easy to code better than o1 or o3 mini high. You’d much rather hire me than it. So I call BS on these silly “best in the world” metrics. I actually get projects completed and generate serious profits. Those models do not.

[@R4YSx]

you think that'll still be the case 5 years from now?

[@Drone256]

@ The English language does not let you fully specify a system. Thats why we have things like mathematical notation. So the paradigm has to change if we are truly going to replace engineers. Mapping English to systems won’t do everything engineers do. A good software developer does far more than write software. If the machines help me with the software part that’ll be great. Ultimately I generate value and profits. AI may get there but we are so far away.

[@richlee1374]

What I heard was: work smarter not harder. Leverage AI as a productivity tool and an all-knowing knowledge base.

[@tvwithtiffani]

He's been saying this for 3 years. Lord have mercy. He has some kind of vendetta against software engineers. Also, OPENAI need to decide are they digging for gold or are they selling shovels? Seems they want it both ways.....

[@kc12394]

Also every year it's we're on the brink of developing agi internally. Like bro stop it, we know you're just yapping to secure funds cause training and and inference costs a crap ton of resource.

[@realhuman2545]

OpenAI sells the idea of a golden shovel that digs the idea of gold and it will perpetually be coming soon.

[@jQuakr]

By the looks of it, it's next month. By the end of the year, it will have solved all the world's problems.

[@ii795]

And create new ones... but that would be not humanity's problems anymore, which gonna retire (if lucky) or extinct (likely).

[@ohsweetmystery]

Male vocal fry is just as bad as female vocal fry.

[@jatelitherius9842]

It must be hard, caring so much

[@zephyrmadera5180]

I think it will be a big mental shift necessary for most people to get the full utility out of AI. You would need to go from task oriented thinking, how to be a worker doing a specific task, to a systems level thinker, how to start an individual business from the ground up and optimizing performance of the different aspects of that business. I think AI will be incredibly empowering to enterprising individuals. Everyone else would slowly need to find these niches to occupy. A job is taken, but a business is on the table, the only problem is most people aren't mentally geared toward running their own personal businesses and one might expect market oversaturation in low barrier to entry business fields

[@mrgoober6320]

We completely failed to adapt to climate change and minimize the damage; we will completely fail to adapt to AI and take care of people as they lose their jobs.

[@WhatIsRealAnymore]

Civil war is loading...

[@psychochef123]

I think as long as humans will be the ones deciding if the output is acceptable there will need to be some sort of project manager aspect of the superhuman coder for Sam's idea that "1 person will be able to build a billion dollar company." Often times by my experience the coders are following incomplete instructions and making assumptions they often times have to undo and refactor. There needs to be something to pick the brains of the stakeholders and accurately define the KPI's they will use to evaluate the end product. This is why we have Agile workflows.

[@Rothron]

Wake me up when 1. You can give it a non-trivial tasks and it actually does it. 2. When what it produces is buggy and you ask it to fix it and it doesn't just make it worse.

[@Gatrehs]

You could start by defining some things you consider non-trivial. And remember, no taksies backsies.

[@Rothron]

@@Gatrehs I have a few reference tasks I test on new models as they are released. The simplest one is to make a grid based fluid sim in a browser window with a clear fluid air separation that lets me slosh the liquid around as I shake the browser window. I've used LLMs to help me code on many occasions, but I only find them useful on small specific functions. Only then will the mistakes they invariably make be clear and small enough to be fixable.

[@MarkoTManninen]

It is a standard coworker when it comes to our teams desing session, listens whats going on our project and goes to Jira task list and completes the tasks given for this sprint. It is a superhuman when it makes the whole team work just by alone. I just wonder who is gonna build such an agent for us? Ask OpenAI how much processing time and money it will take to run that beast at the end of the year? Similarly with ARC-AGI achievement. Front end products may become increasingly expensive and the AGI instance for big budgets only. You dont want to feed it to solve a worm game at that point anymore.

[@2034-SWE]

Exactly - I'm expecting a $2k/mo, $20k, and $200k+ enterprise offering in the not so distant future. Models that think for hours or even days/weeks at a time for example

[@Motocentrick]

Sam Altman should be aware that pretty soon we're going to have fully automated robotic techno feudal overlords. No one will need him. I thought he was pretty cool when he first showed up, but he's so skeevy slimy and creepy

[@tracy419]

Why?

[@Motocentrick]

@tracy419 I don't make the rules

[@tracy419]

@@Motocentrick I meant in regards to being slimy and creepy. I keep seeing people say these things, yet he still sounds the same to me. It seems to me like musks attacks have been successful.

[@Motocentrick]

@tracy419 lol MUSK certainly would not influence me at all. I don't know he just creeps me out a little. His whole voice and demeanor.

[@morezombies9685]

​@@tracy419look up "sam altman sister" "sam altman firing lies".

[@jeffkingston67]

It's creepy when they say, 'My take from this is to just start trying to figure out how to incorporate AI into as many things you are doing as possible,' as though they are being helpful, rather than selling a product.

[@stevenmedina4030]

You can see the scale of power. Altman dress casual. People near him dress suits. For them they comply to etiquette rules. But altman makes the rules. To be higher up Its liberating.

[@XScreenName]

When he says "AI" He really means H1B Visa Indian Software Engineer.

[@Wes-Tyler]

Sam speaks exclusively in vocal fry. It's irritating to listen to

[@gamrfam]

These changes are already greatly affecting the job market for engineers. Let's not sugarcoat what's really happening.