The real virus is articles like this.

I cracked when he said “I wrote this article for my parents”. Sure buddy I also post everything I write for my parents online

😂

Desperation. Every single AI startup depends on Big corp AI achieving "AGI" Is all drama and marketing at this point.

💯

Also posted on YouTube:

i know you probably didn't intend to but you provided some comic relief in your analysis lol Laughed quite a bit.

haha glad you enjoyed it!

Finally an adult in the room

👊

I thought the original article was quite interesting, definitely disagree with some of it as well, but I also think you may be minimizing the models’ capabilities a bit too much. Sure, you may tell a model that it should iterate until it reaches an optimal outcome, but i've observed that for long-running tasks that I'm not actively managing, it is able to make judgements that improve the final product without me specifying exactly what changes to make. And also, because of their training data, these systems clearly have some sense of what humans find aesthetically pleasing. I think it's highly important not to get too caught up in some of this terminology - arguing whether or not they truly "feel" or "understand". At the end of the day, they mimic extremely well, and that mimicry produces genuinely powerful results. Also, after reading some other comments, I would implore you try Opus 4.6 and Codex 5.3. Anecdotally, they really feel like significant steps forward. And in general, it's worth continually experimenting with these tools even if you aren't all that impressed - you would be able to evaluate them more credibly

generally my stance is that the nature of work always evolves so that humans remain useful

I mean certainly right now i'm very important to the work that AI assists me with. I sense the thing people fear most is that very reductively, much of what I do involves writing prompts and orchestrating which LLM can probably do pretty effectively. However I feel we might not give ourselves enough credit for what actually goes into the prompting, the orchestration, the choice of tools, etc

yeah, not giving ourself enough credit is a good line i think

i laughed so hard out loud to “oh shit february 5th” 🎯 it definitely reads like a 9/11 documentary 😂

😂

just finished the whole thing. i’m an ai freak like the writer too but i thought this was hilarious

😄

The weird part is I think he's right, but NOT about the content/conclusion. There's likely gonna be a disruption, but it's because of *AI bubble burst*, which will likely wreak havoc on our entire economy, due to bankruptcy of all major AI companies 😬

yeah people are gonna get hurt one way or another here

On the one hand I agree with you; the author is incredibly biased, overly sensational and misleading. On the other hand, the bar to hire junior developers / juniors in general has risen dramatically and trends towards rising even more. Enjoyed your commentary.

glad you enjoyed it! the bar is always rising is it not?

As history shows yes and the market always adapts. Personally, I find it continuously harder to delegate work to a junior than to an LLM. trying to adapt.

He also said ai wrote that “article” for him.

did he?

🤢

😆🙄

I stopped listening after you said Opus 4.6 is a minor change. I think you should watch this video to understand this "minor" change means

bruh

@MohamedSaleh__ Bruh is all you have to say?

@MohamedSaleh__ nate is engrossing but just reads off a chatgpt pro/deep research report. hard to know what he actually believes

💯

lol nice 🤝

You sir get a follow for addressing this nonsense Who do you reckon he’s in cahoots with?

🙏

Well said. That’s the problem with modern algorithms. The more general and shallow a post is, the more it spreads. Naturally, if something is hard to understand, it can’t spread because people simply don’t get it

🙏

Thank you for calling it out. The comments under that article are filled with human stupidity and AI slop.

🙏

GPT told me it was fear porn

yup

As someone who is familiar enough but don't work in it directly I came to a lot of the same end conclusions Spent an hour in a space arguing about this topic yesterday for this "article" to come out today and see people react the way they were I thought I was the crazy one...

yeah easy to fee like the crazy one these days

have you played with openclaw???

no 😬

i used to think mostly the same as your video (watched the full thing, you have good commentary) but after playing with openclaw, it’s quite believable the original article is spot on

how’s that

words won’t convince you, install openclaw and try it

what have you been able to do with it so far

wait until religious orgs start forbidding it. it's a cycle that humans love to ride with every novel general purpose technology. we'll be fine like always, nothing more dangerous to humans than humans.

yup

Agree. and the odd looking finger salad in the pic is my first clue.

ha just noticed that

It sounds like you have some bias that is based in your own fears of being replaced. I think you are speaking from a place of wanting to believe a different version of the future. Thats perfectly fine. I would write my own article if i were you.

Best laugh I’ve had all day.

😄

The like button is broken, thanks for sharing Mo. I was almost laughing the whole time I read his article and was like nobody had ever tried using an LLM. He just straight up deceived the people that don’t know anything

🙏

I have been upgrading from model to model for the same usecases and the leap in accuracy was shocking to me personally. I understand the marketing angle in the article, but I couldnt rule out the possibilities completely.

this is basically what hit developers last year. and we’ve already seen through it. now it’s just hitting other fields. you will see right through it soon :)

Thank you sir. Things got a little out of hand last night.

👊

Tbf bro I don’t think we’re near AGI, but do you not think that mastery of language, chained recursive reasoning + access to tools will be enough to replace swathes of the workforce? Because honestly we’re not far off that

tempting to think so but humans adapt and evolve too

Betting against humanity is always a losing hand. And again, we’re not talking about AGI here, very very interesting few years for the good ol’ Homo sapiens up next

So I’m a partner at a law firm in Australia. Much of what the article says absolutely rings true for our business. I absolutely understand that the article is a pitch, it’s advertising and should be labelled as such. Having said that, we had (the partnership) a conversation…..

what was the conversation?

Well done. You changed the world with a single video reaction to an article @atmoio. CC @clairekart

@clairekart 😂😆

not to mention he was heading up that reflection 70b grift last year many were fooled. that said, we will see how things pan out.

@contentcontent_ what does he say in here

@contentcontent_ He was caught lying

@atmoio Yes I’m not sure if ever fully admitted it to being an outright scam or if he fully apologised either.

People not realizing they are reading marketing written by an AI to get them to buy into AI more is so funny.

💯

Thanks for this. That article was so over exaggerated that I couldn’t force myself to finish reading it. Shocking how many people retweeted and thought it’s accurate

glad you liked it 🙏

ngl this felt like a Kony2012 moment....

😂 you remember that too huh

I particularly enjoyed the part where Shumer insists every attorney ought to send client data into public LLMs — a violation of client confidentiality and protected privileges between counsel and client. Extremely bad advice and career altering if followed.

good point actually

@Skinnerisms Corporate data loss prevention policies would require them to buy an enterprise plan, which does not have the same privacy concerns as uploading straight to a public LLM as enterprise plan users chats aren’t used in training data and can even be stored locally.

@atmoio I didn’t see anything in the Shumer article that pointed attorneys to check in with their firm/corporate infosec about an enterprise plan.

Loved the video, but I think you are misinterpreting some of what he's saying. I don't think AI will do your job better than you, but it does...like 70% of your job faster that you do, and only really needs your input for the 30%. I do think that the article was sensationalized, but the core of what he's saying is true. Most of the "free tier" available on these AI tools are quite limited, that the difference between them and paid versions are....quite significant.

Thank you for providing this take, I feel the same way. I’m a senior developer and I am constantly correcting models making completely ridiculous decisions. It’s output would be useless if I didn’t have the ability to identify and correct its shortcomings.

yup people that can correct llms will always be valuable