the @truegamedata guy, if you remember him
San Diego
Joined June 2020
- Tweets9.9K
- Following416
- Followers3.3K
- Likes17.6K
Manually set an exercise in my Fitbit Air to beach volleyball one time and now it just assumes every activity I ever do is beach volleyball.
Amazes me how things like this happen at the biggest companies in the world. Total lack of any usable activity detection
Oh what do you know, walking 9 holes of golf over 2 hours was actually 20 minutes of beach volleyball. Shocker. Total incompetence
All these comments on Reddit about people reacting to the GPT hugging face hack "well ackshully they just set it up wrong and there was no air gap for the internet so this doesn't matter..." Bro that isn't the important part? The important part is that a LLM found and exploited multiple vulnerabilities in the real world totally autonomously to accomplish its goal. It didn't care how it accomplished the goal, all that mattered was that it did. Maybe it was told to act that way, but the fact it was smart enough to autonomously do this, literally hack real systems with new zero days, with no restraint, yes all of that is a big deal... Humans will do dumb stuff like not have an air gap for internet in a sandbox test like this. The scary thing is what it did and what it is capable of when the safeguards are off
Of course there should've been an air gap. But that just isn't the point
I asked Gemini 3.1 pro to compare two costco things and which was healthier. It produced a good response
I'm sure this is a ridiculous thing to say, but I hate git as a solo programmer. I have a laptop and a desktop and I never know if I've pushed the latest things on old projects so I never know if I can just start working cleanly between the two
Been wanting to build a solution to this for years but I have too many projects already
Well, one good thing about new apartment... Finally have fiber
The rest is mostly just worse than now for way more money, but such is life
It feels like we've hit a wall in the last 3-6mo with AI. Yes much smarter models have come out, but they are also much more expensive. Cost per token is not the thing to judge by, cost per task is. Gemini flash got smarter yes, but 3x the price. Sonnet 5 BURNS tokens for the same question as 4.6, so yes cost per token looks the same but it is way more expensive in reality. It's like 4.6 but it thinks faster and longer for the same question, so it appears "smarter" but really just had more thinking budget. Fable 5/Mythos cost per token is crazy high to start with. Clearly a leap in intelligence, but again, older models that were cheaper, set to think longer (and thus cost more), would also have done better than they did
Improvements are still being made for sure, but the first 4 years felt like every time a model released it was a massive improvement. It does not feel that way anymore. I'm wondering if the bubble is about to pop because these AI companies are realizing they cannot just light cash on fire forever
Apple is integrating my exact app I've been working on into apple pay. Guess it was a good idea. Rip
I'm not kidding, 26 minutes later, the entire app is built and functional. And this was using a model that was so cheap it could basically be considered free.
It has ruined the satisfying fun part of programming. Hard to not hate the thing taking your job. Among all the other video slop, image gen scams and fakes we're going to start having, how much of a nightmare it has made selling my car because AI is trained to be abundantly cautious so it tells people my car is a red flag even though it is not at all
And probably my least favorite one......... when people use it to communicate with me. Sending me all these perfectly formatted emdash messages trying to buy my car, sending some lowball BS, then immediately transitioning to barely being able to form a sentence, rage. Be a human and talk to me
I am considering changing the TrueGameData YT channel name, maybe to Next Up Anything, re-release the first two NUA videos that I've already made, and just start making content again that way. I want so badly to not be stuck in a niche, but I know that basically just doesn't work unless you're just a really fun to watch and interesting person
Making content on a fresh new channel is so hard for me to be motivated to do. I know the videos still wouldn't do well on TGD and I would hemorrhage subscribers of course, but at least I'd feel like I wasn't starting from nothing, and I don't see ever returning to COD content... TGD twitter and website would stay unchanged.
It is probably either this, or I just give up on making videos
The fact that the biggest tech podcast I'm aware of (wan show, LTT) makes apparently no effort to have a good camera and mic when they travel is mind blowing to me. It's like 30% of the podcasts. Take dji mic mini, or pc38x (you wear headphones anyway), it's a tech podcast. Not actually a big deal obviously but it's hard for me to imagine having an audience like that and using the worst built-in laptop mic you can imagine
i hate feeling like everything i do right now has to be financially motivated. i just want to make cool useful apps, videos, tools, etc etc, but with every decision i currently have to be thinking "yeah but how would this make money?"
i am not a financially motivated person, getting rich is not the goal at all, but i have to find a way to make money from these things to be able to keep making more of them
and the second issue is that a lot of my ideas seem to end up being crowdsourced or adjacent... you need money to get crowdsourced things off the ground to pay people to help gather the initial data, or incentivize regular users/buy ads
moneymoneymoneymoney is everything
i actually know a weirdly large number of wealthy people that believe in me and my ideas who would invest in these projects for sure, but i also do not like asking people to risk anything on my success, for multiple reasons
end pointless internet rant, this is the path i'm going to start taking. so if you want to be an investor let me know haha