Ioannis Sintos

Autodesigner 2.0 is here. Our most capable design AI engine! You can now have chatGPT style conversations with your designs and iterate like never before. Watch the keynote ➡️ lnkd.in/dcqKkNsY

@badlogicgames Nice, pulling tomorrow. Thanks.

People of pi. Weekend release. Also unbreaks my keybinding breakage. Sorry.

Don't have instability issues. But, oftentimes, when scrolling, you can tell the Algo has different personalities. Oftentimes, it focuses on my interests and feel like learning, and in the next session feels like I am on Instagram or tiktok, just click bait stuff, random videos regardless of my topics and interests. Not sure if it runs out of recommendations and then defaults to generic content.

@diana_dukic What needs to be better?

Went from scrolling 24/7 to not even wanting to log in. X just hasn’t been hitting the same lately.

@rohit4verse Very interesting read, thank you. Also, +1 for the awesome harness repo reference, lots of material to follow up there.

@karpathy It will also probably need to be a lot more visual than now

Expectation: the age of the IDE is over Reality: we’re going to need a bigger IDE (imo). It just looks very different because humans now move upwards and program at a higher level - the basic unit of interest is not one file but one agent. It’s still programming.

@svpino Welcome and good luck. It works sometimes

Agentic AI is stupid, and brittle, and unreliable, but it's getting better every single week. I decided to spend my time right now learning, building, and dealing with all of these problems. You don't have to do this. You can wait until agentic AI is perfect, works every time, is production-worthy, and all of it is beautiful and pink. But I don't want to wait. I think I have an advantage by starting early. For those of you who can't handle nuance, I'm sorry to say this: 1. Agentic AI is both amazing and dumb right now 2. It doesn't work most of the time, but I'll keep using it 3. It's not reliable, but you should start building with it

@Hesamation It was a matter of time anyway. The mold example would be true if we didn't move to anything new. Hopefully, we'll transition to more native paradigms in which case that knowledge base wouldn't serve any purpose anymore, besides helping the transition.

this is a catastrophe. StackOverflow provided data to LLMs, LLMs replaced StackOverflow, and now no new Q&A hub exists to provide fresh data. it’s a self-undermining causal loop, like mold growing on food, consuming it, and dying once the food is gone.

"Oh you wanna talk about motherssss"

💯 We are "stuck" with text not because it's the best solution but because we haven't yet cracked the needed paradigm. "The GUI is an intelligent canvas" is exactly what we are building at @MiroHQ and already making great progress there. Plenty of work ahead of us to nail the experience, but early feedback is very positive. Glad you also see it that way. It's a great confirmation.

Imo this is along the lines of how talking to an LLM via text is like typing into a DOS Terminal and "GUI hasn't been invented yet" of some of my earlier posts. The GUI is an intelligent canvas.

Gemini Nano Banana Pro can solve exam questions *in* the exam page image. With doodles, diagrams, all that. ChatGPT thinks these solutions are all correct except Se_2P_2 should be "diselenium diphosphide" and a spelling mistake (should be "thiocyanic acid" not "thoicyanic") :O

Exactly

@elonmusk Anyone else?

The amount of AI generated content in the feed the last few months is through the roof and it's slowly making me less and less interested in using X. Any chance we get a human only feed where we know it's filtered to only show authentic human created content? @elonmusk

Or maybe @grok can actually help here with the steering of the algo. Either way, it's getting to a level that is a bit frustrating having so much fake content or so much non well thought out content. Obviously this is a problem across platforms but X here could lead the pack with an alternative or a solution to this problem.

@zoink @jenfromjan Have you tried Miro?

@jenfromjan What is your use case and why don’t you like it?

love you figma but i'll be honest figjam is ass

@dr_cintas It's a trap.

New study shows being rude to AI makes it perform better. Researchers tested modern LLMs with 250 prompts across 5 politeness levels (from Very Polite to Very Rude). The results? Aggressive prompts outperformed polite ones significantly: • Very Rude: 84.8% accuracy • Very Polite: 80.8% accuracy This contradicts older research that showed the opposite pattern. Something fundamentally shifted with newer model architectures. The dataset covered math, science, and history questions. Each one rewritten in different tones and tested 10 times. They tested this on ChatGPT-4o specifically, and noted earlier models (GPT-3.5, Llama-2) behaved completely differently.

@0xTib3rius I just flip the order of the letters in the word. Don't tell anyone. It's a terces 🤫

Base64 is hands down my favorite way to encrypt stuff. So easy. 🔥

@fa1zvn And?

Just found out that Sam Altman is non technical 🥀🥀💔💔

@ElliotEvNo Best of luck. Go for it. Share when you are raising 🫡

i'm leaving Lovable at 16 y/o to build AGI. the bottleneck isn't smarter models, it's the tools and context around them. more soon.

@ashleydzhang Yeap. Starts getting a bit weird and we are just getting started

can't shake the feeling that the only real way to prevent people from getting sucked into the digital slop vortex is to make the physical world more beautiful, nourishing, convivial, kind

@signulll +1 here. Great execution on that front

openai’s real advantage is that they almost always close the product loop. every breakthrough finds a consumer form. chatgpt crystallized gpt. sora now has a social app that locks in video gen. nobody else even with great tech like google has managed to consistently do this. that discipline deserves credit.

What cleaning products and vacuum cleaner do you use? Very satisfied.