Tokenmaxxing Is a Problem — AI Is Shipping Bad Code and Nobody's Talking About It

Hinzugefügt: 2026-05-12

[00:00:00]It seems uh we need to take maybe take a  step back on the AI and try to find a way of um integrating it with how the humans will  use it. The further down we go. Honestly, I just feel like we're all going to have to be  super vigilant and just be babysitters at some extent. Hello guys, welcome to the channel. So,  it's been a while you've seen me on camera and today is going to be different. It's going to  be different from what you've been seeing on the channel for some time now. Now, we are  going to begin a podcast series where each week we try to talk about some um activities  happening in the tech industry and I have a friend that is going to be on the podcast  as well. And with this weekly podcast, what we are going to do is to have some form  of breakdown on topics and ideas happening in the tech industry. And with that, I'm going  to welcome Manny into the channel. [DRUM ROLL] What's good y'all? It's your boy Manny. And uh  like Danzzy just said, I'm going to be here a lot more. So, a couple weeks ago, it was I reached  out to Danzzy because I've been watching a lot of his content from way back. And if you follow him,  if you know his work, you know he does a lot of great stuff from how to work in Windows, how to  manage your system, and how to do all that. And I was like, "Hey, how do I contribute to this  community?" And we all learn and grow together.

[00:01:30]So if you've looked at the news, if you've if  you have a phone and you get the internet, you could tell it's a lot happening in AI, in cyber  security, in the stock market, in fintech, in the game industry, it's a lot happening. GTA 6 is  coming out this year later. So, it's a lot we tech heads want to learn and grow and share together.  So, we want to make this community one that's always up to date and caring with all new things  happening. And so, I'm going to be here to co-host with Danzzy and overall we just cover some other  stuff together. So, it's going to be some great times, exciting times, and we're going to learn  and have so much fun together. So, I'm looking forward to just seeing y'all down in the chat and  seeing what y'all want to see us talk about in future videos. So definitely have it here and then  let's just stay together and hear your opinions.

[00:02:17]So today we actually wanted to talk about  something that Danzzy and I interestingly enough we're talking about on the phone and um I'm pretty  sure a lot of us have heard about it, Danzzy um Tokenmaxxing and I feel like we're getting to a  point where AI is getting widespread and people want to just show how super user they are and  how much they're doing with AI. So, I think this phenomenon is getting bigger and bigger and today  we just wanted to cover a little bit more about it and see what y'all think on Tokenmaxxing. So,  with all new technologies um with the internet, it was like everyone was always rushing to have  a www...com and I feel like with AI companies want to push people to use it even if they're not  really being productive. So I feel like for a lot of companies like Facebook, Microsoft, Meta,  you can literally see week in and week out, it's a lot of layoffs. And a lot of these are  based on who's using AI the most. Not even who's being more productive. It's just how much of AI  do you use, are you doing good stuff with it? They don't care about that. And there's like metrics  where people use to track who's, you know, being more effective in using AI. So it's like people,  you know, always find ways to game the system.

[00:03:33]um you can literally go claude code your usage. It  has a little percentage score and it's pretty easy to ramp it up without being productive. So I feel  like with the topic of AI and just the more we get into it, it's going to be more of just going  forward where people try to game the system. So we kind of found this newsletter which we thought  was a little bit interesting. So The Pulse: 'Tokenmaxxing' as a weird new trend. So in here  they're just saying employees at meta platforms who want to show off their AI super user chops are  competing on an internal leader board for status as a session immortal or even better token legend.  Yeah, I mean with Tokenmaxxing it's an interesting way of calling it and um yeah, there was a video  where I saw that um the company what they were doing was that they were comparing the kind of the  lines of code that you pushing onto GitHub and so I was thinking who is going to review all these  pull requests and then push them into production or into the main branch. It's I don't know if  you've seen this. Um so Amazon CEO I keep blanking on his name but um apparently recently due to like  they had a lot of issues with Amazon Kira. I don't know if you heard it but um they had some product  issues with um I think it was CloudFront cuz one of their junior or mid-level engineers used Kiro  to push their software and it wasn't the greatest software. It had a lot of bugs in it and it led  to one of their recent uh vulnerabilities and day off. So it was like now they have a new policy  where before code can get merged and you know get pushed it has to be reviewed by senior devs  and I just feel like now we're hiring less people and we're forcing all these people to use AI and  having senior devs do it is literally going to be glorified babysitting where instead of having more  time to do work senior devs now have to just like see hey how's this junior dev doing their work  is their code great and I just feel like I don't know where we're going with this, but I feel like  there's got to be a point where we're like, "Hey, how do we limit how much code we're getting out  there?" It's a difference between shipping code and shipping bad code. And I think right now  we're doing the latter not the former. Yeah.

[00:05:54]And you even notice that um of late Amazon cannot  boast of having four nines. It keeps on reducing week on week on week on week. Well, this week  I haven't checked to see where they are with their availability of their services, but I mean  they talking about using AI in everything and all of a sudden you see that their services are  becoming bad and they now have to go back to resolve it and stuff. I mean it seems uh we need  to take maybe take a step back on the AI and try to find a way of um integrating it with how the  humans will use it so that there's no any form of um conflict right there. Yeah. And I think that's  a good point you mentioned because I feel like the biggest culprit of just getting work out there  whether it's good or bad would be anthropic and just going here if we want to look at their  downtime you could see this is pretty bad. They have under four nines of reliability which means  they're down like a lot. Wow. And just look at that. Look at um claude code down here. That's  that's a lot. It's been down literally minimum once every week. We could we could actually say  it more because I see a lot of reds and a lot of oranges here. So, I just think it's not a good  sign. And they're the biggest culprits cuz one, they push out a lot of code. And I didn't  know recently they had their whole claude code base get leaked. And that's just like  one of few instances where moving fast isn't really the goal now because I just feel code  is easy. the technical work is getting easier, but having a good mindset where you know what's  important, what's necessary, and what's not, that's going to be even more important the  more we go down this AI line and just, hey, how much of AI are you using? How much power user  are you? And just, you know, the further down we go. Honestly, I just feel like we're all gonna  have to be super vigilant and just be babysitters at some extent. Just just look at that. That's a  pretty bad figure in here. Claude AI has been down a lot. Claude console. Interestingly, Claude for  government, I mean, it should be better because, you know, you're serving the US government and  some other big governmental organization, so it better be good. So, they're going to be good, but  for the regular user, good luck with you. And if you want to keep learning more about where we're  going with AI, we're going to send a link up here so you can click on it and go see and learn more  about it. We'll catch y'all later. And we're out