Many post about "reverse engineering" TVs but in reality it's just installing a custom Android TV OS shell, which is a use case supported by the manufacturer. This doesn't work on LG and Samsung TVs however, as they are more locked down and AI isn't good enough yet to "crack" that. That's what shows the difference between real intelligence vs something that can just automate the process. Ngl the latter is useful, BUT it won't magically crack things that are properly secured or haven't been done before in a similar way..

When someone, having access to best OpenAI models, still has to plug a USB stick into a *smart TV* to make it do nasty 🥸

Testing a much more efficient way to code with AI (IMHO) - I plan with its help (.md) - I write the code (!) - I ask AI to review the code Gives me much more agency (hah) and better results since I'm in control at all times, while getting defects spotted that I might have otherwise overlooked. It is also apparently much faster, in my hands 😅 Because I don't have to spend a lot of my time on: - reviewing the code (I already know it) - correcting what AI wrote - iterating with a 5-minute wait for every damn prompt 😂 - remembering what AI told me yesterday about the code - regaining my lost skills later on 🙈 It's a win-win IMHO, but not many ppl realize this "reverse" path can be more efficient and pragmatic, yet

Shoutout to cybersec community here on X - wholesome ppl who're fighting the good fight on AI hysteria and overhype 😍 It's become so hard to find balanced discussions on X, where everything is now either extremely for or against AI, and all discussions seemingly revolve around AI. If you dare to post something that's middle-ground, you get swarmed by insane folks/bots on either front 🙈 If it wasn't for such communities, I'd probably leave X already, because I'm not here for the loss of common sense..

Just troubleshot a VLAN issue with my local network, using my 🧠 I asked GLM 5.3 in parallel, which started spitting out a bunch of text wasting a ton of tokens. After submitting the prompt, I resisted temptation to wait for response and actually figured it out myself sooner than that (the issue turned out to be trivial). Meanwhile, AI response contained a cliche question, the kind that IT support not having a clue about my network would ask.. Feels good not to be relying on AI, which would likely waste both my time and $ going in circles 😅 And I learned something for next time, instead of offloading my memory to a clanker..

I'm a relative nobody compared to Zack, but his message needs repeating over and over to those not closely familiar with security. It really helps calm down the FUD spread by big AI labs, especially in ppl we care about.

Really good interview with actual authorities on financial management and fraud and AI, who have studied these things inside and out for many years, on why AI datacenter investments are just so deeply unsustainable. It's good to listen to the real experts in the feild, instead of random AI influencers and bros on X 😅 youtube.com/watch?v=g9KdLg9s…

My apologies for calling AI bros on X LARPers - it's offensive to the real LARPers 🤣 youtube.com/watch?v=0qD4rF3t…

Everyone is missing the point:

LLMs are very poor at production-grade software, because they over-complicate _everything_ 😑

Claims of death-of-coding-by-hand have been greatly exaggerated 😅

Fun fact: 80% of revenue of big AI companies is coming from 1% of its users 😅 There is simply NO way they can keep increasing prices and have any hope of return on their investment in the next few years.

One of my fave new YT channels critiquing senseless AI hype 😅 We need more such creators restoring balance and sanity to it all.. youtube.com/watch?v=4Ff0xc9M…

And then AI-obsessed folks wonder why regular ppl hate AI 🙈

I don't get Jev, why is making wrong decisions fast useful? 🤔

Coding with an LLM is productive sometimes, and UNproductive most other times 🫠

A very thought-provoking interview with an actual computer scientist @ProfCalNewport 👀 I prefer listening to such measured and level-headed experts. Instead, X timeline is filled with rando AI bros who (without any evidence) claim LLMs are the path to AGI and how you have a "skill issue" if you aren't able to use them reliably (plot twist: NO ONE can). Good thing to listen to if you're a smart person with critical thinking. youtube.com/watch?v=gN2FULNS…

Ladies and gentlemen, this is the AI we need to slow down

It's so funny to observe ppl freaking out because of some stupid AI chatbots 🤣 Looking back, we will wonder just how stupid we were to attribute them so much meaning and importance. We will also look back at AI-pilled techbros like we do now at Crypto bros 😅