
If everyone is thinking alike, then no one is thinking outside the chat box

I make AI agents think and act like humans

If everyone is thinking alike, then no one is thinking outside the chat box


oftentimes, you can guess which model people used only by their questions, comments and objections. they start to sound like Claude, they have a bit of a Gemini point of view, or a chatgpt way of questioning a problem. to me, the most dangerous thing about AI for humankind is not a rogue agent, but delegating thinking to LLMs. even if AI learns how to self-replicate, it doesn't matter; we are already giving up on one of the most fundamental things that makes us human. biology has a simple rule: if you don't use it, it will atrophy, because why waste energy on something you don't need or use at all? so as models become "smarter", humans will become even more stupid (surprisingly). but here's where the circle of life come into play: because humans will be even more stupid, models won't be able to learn anything new, simply because no one will be generating valuable critical thinking and new knowledge (new data). and we know AI is incapable of creating anything trully new by itself; it will stall at what we'll remember as the pinnacle of human intelligence. I'm sure monkeys and crows will teach this period at their schools, when they take over, calling it perhaps the era of the massive de-evolution. but if we want to avoid this trap, we need to sourround ourselves by artists, both in life and at work. they have saved humanity many times from the fallacies of reason. and dw, in the end, they will come once more to our rescue. halloween is ultimately a celebration and an ode to human expresion and creativity, something that no model can even come close to. so I want to celebrate it by acknowledging the artists and creative thinkers in my life, in my team and those who I cross paths with, even briefly. you keep me human and my brain spiking. ty! in the pic: left: a piece of art created by my teammate @Duart_Monteiro (you need to feel this artificial brain in your hands to understand). right: two pieces of art by el maestro @stavan Happy Halloween to all! Stay wild, stay creative, and never ever delegate thinking (even if it makes you feel smarter).

coding is solved. but engineering is not. don’t mistake the two.

Whatever people say, you have the freedom to do anything. The only thing stopping you are the consequences of doing it.

As a society, we've traded off taste for speed and efficiency
You can see the decline of society just by looking at what we used to build versus what we build now.

We made it to @TechCrunch's 2026 Startup Battlefield 200. @marketrix_ai will be at Moscone West, San Francisco, Oct 13-15. Here's what we're actually building.


Note to myself - In this day and age, clarity is everything. Use AI to make life less boring. But if you can't function without it, you are just "artificially intelligent".

When Copernicus proposed heliocentrism in 1543, it was actually less accurate than Ptolemy's geocentric model - a system refined over 1,400 years with epicycles precisely tuned to match observed planetary positions. It took another 70 years before Kepler, working from Tycho Brahe's unprecedentedly precise observations, replaced Copernicus’s circles with ellipses - finally making heliocentrism empirically superior. Terence Tao's point is that science needs a high temperature setting. If we only fund and follow what's most state of the art today, we kill the ideas that might need decades of work to surpass some overall plateau.

you can outsource your thinking but you cannot outsource your understanding

We just launched @marketrix_ai User Testing Agent on @ProductHunt! Most QA tools test what you tell them to test. But real users don't follow scripts. Marketrix flips this. Our AI agents explore your product the way real users do, uncovering bugs, edge cases, and UX gaps automatically. No scripts. No manual test cases. Just autonomous QA that actually thinks like your customers. We built this because after years of watching teams burn cycles writing and maintaining test scripts, we knew there had to be a better way. Instead of telling the tool what to test, you let the AI figure it out, just like a real user would. Here's what makes it different: - No scripts required - Works across staging and production - Generates test cases automatically - AI agents explore real user journeys - Continuous validation before release We're live on Product Hunt today and would love your support. If this resonates with you, head over and give us an upvote and drop a comment. Your support truly means the world to us as a small team building something we believe in. Product Hunt link : producthunt.com/products/mar… #producthunt #qa #softwaretesting #ai #devtools #startups #launchday

Interesting quote I saw today: “The flaw in connecting dots from past successes is that, it discounts why others who did similar things failed”

Someone just built a fully open source mocap system that works with cheap webcams.

TL;DR WebMCP turns web UIs into APIs. Soon, software will have both backend and WebMCP APIs. For developers - this means you'll have to build not just backend APIs, but also WebMCP APIs for UIs. It's a matter of time before both these APIs converge to just one MCP API. (5/5)

This means, web apps and web sites can now be designed in a way that makes sense not just for humans, but also for browser agents. When browser agents call these tools, they would work just right. Finally, they get some determinism. (4/5)