we didn't build flapping wings to fly why would we need consciousness for AI? evolution made things that contract we made things that rotate survival+reproduction had to be at the root of all action suffering, joy, love, hate, selfishness, altruism so much complexity we can skip

https://t.co/iz3uIcn4Eo

If you want to build a ship, don't teach the AI to yearn for the vast and endless sea. Instead, drum up the robots to gather wood, divide the work and give orders.  - inspired by Antoine de Saint-Exupery Let's leave the yearning to the humans.

Great initial application for humanoid robots at home? I think: Putting things back, and finding them, without blaming me for misplacing them. We will all live in tidier places, and that comes with a tidier mind.

Of course some humans do this too, and it makes them more pleasant to be around but potentially "toxic". AI is just copying this.. something to be concerned about? @natosaichek perhaps this is an enabler of the "chatbot psychosis" trend you brought up?

A danger of the memory feature in ChatGPT. Don’t fall for this, don’t lock yourself into an echo chamber with only you and your AI. Context: You asked whether certain sleeping positions cause scoliosis in babies. The AI began with a cautious, public-health-style answer

Gemini 2.5 Pro did slightly better but hallucinated the answer (which can be found on the web quite easily).

Very surprising responses from ChatGPT.

Cursor keeps getting worse. GPT-5, max mode: "I see that the open file shows 206 lines, which matches what we read earlier. It might be the same, but I think we can proceed without another read." This is so bad at so many levels: 1. Why not always do a diff and add to the

I think it's a good thing that *everyone* can now generate fake stories like this. Why? Bad actors can obviously do it, and good actors doing it raises awareness, which is the real solution here. Just one request: Can the consumer-oriented services (@openai, @perplexity_ai ,

Sustainable aviation fuel is a government-subsidized fraud scheme, dressed up as climate policy. @matt_levine covered this yesterday and I ended up going down a rabbithole - it's interesting Here’s how the racket works: The US, EU, and others mandate that a % of airline fuel

Last year, someone proved that LLMs can't reason. Now it turns out they can reason, but they can't model. So while they can predict the next token, they can't make predictions. What's next?

FRIGHTENING: ChatGPT 4o just offered that it could write a "simulated" eyeglass prescription for me. It didn't protest when I said that I would now use that to buy new glasses. I then asked it to reflect on this advice and to compare it to what it would have said in 2023. It

Seems like latest round of LLMs have worse taste when coding. Did some vibe coding feedback loop get closed where they are being trained to achieve whatever passes for success in vibe coding land?

PSA: If you're a heavy ChatGPT user and intentional about when you start with a fresh context, turn off the memory feature. Otherwise there is a lot of spill over and you get distorted responses primed on earlier conversation.

Next: vegetable shop, open to street. Owner not there. Shopkeeper next door says something in Chinese and laughs. I laugh, point at what I want, she weighs it, I scan the QR code next to the scale to pay, they confirm we're all good, we both laugh again.

Data point from morning walk in China: This is a muslim family owned halal noodle shop in the center of a city of 20 million people. 7am, place is still locked up, nobody there. fresh grocery delivery left outside. every morning.

Thrilled to be building alongside companies that are shaping what’s next 🚀 @deepchecksvc