Exploring the AI frontier: models, agents, benchmarks & applied experiments. I test, compare and build. (kinda got DMCA'd by two companies we don't talk about)
Build with me 👉
Joined February 2024
- Tweets3.7K
- Following191
- Followers4.8K
- Likes6.4K
Pinned Tweet
Activision just DMCA’d me for porting CoD games to the browser.
They nuked every post and every game is now offline.
Am I the first vibe coder to get DMCA’d by a major gaming company?
It was fun while it lasted, boys. 🫡
AI has made me a much happier person.
I’ve learned more in the past year than I ever thought possible. I’ve built tools I never imagined I’d be capable of building. I’ve automated countless tedious tasks that used to waste my time.
But the biggest change is psychological.
I no longer look at a problem and wonder whether I can solve it.
I approach practically everything with the confidence that I’ll figure it out, no matter how complicated it seems.
And that feeling is incredibly liberating.
AI hasn’t just made me more productive. It’s fundamentally changed how I see the world and what I believe I’m capable of achieving.
I can't imagine going back.
There’s something deeply ironic about people using the internet to organize a movement against technological progress.
You are literally using the product of the thing you’re fighting.
“Who will be held accountable when AI makes mistakes?”
Excellent question.
Now tell me who’s being held accountable for all the catastrophic decisions humans are making right now.
A year ago, I was impressed when AI could write a decent function.
Now I’m disappointed when it can’t build an entire game.
Our expectations are moving so quickly that we’ve completely lost perspective on what is happening.
I think people are missing something extraordinary about AI.
We’re no longer just building machines that follow instructions.
We’re building systems that can increasingly figure out how to accomplish goals we don’t know how to accomplish ourselves.
That distinction changes everything.
And I suspect we’re only beginning to understand the consequences.
“AI doesn’t actually understand anything.”
And yet it’s solving problems you can’t solve.
Writing software you can’t write.
Explaining concepts you don’t understand.
At what point does arguing about whether it’s “really thinking” become less important than what it can actually accomplish?
"AI will make the rich richer."
Possibly.
But here's something I don't understand.
If AI allows one person to produce what previously required a hundred employees, why is the immediate reaction to oppose the technology?
Why isn't the reaction to demand that everyone benefit from that productivity?
If machines can produce more food, more medicine, more housing, more software and more scientific discoveries at a fraction of the cost, that's an extraordinary opportunity.
The question of who captures the wealth is a political and economic problem.
The ability to create that wealth is a technological achievement.
Destroying the achievement doesn't solve the distribution problem.
It just leaves everyone with less.
We should be arguing about how to distribute abundance.
Not how to prevent it from existing.
“AI is making people stupid.”
Funny how nobody said calculators were making mathematicians stupid, or cameras were making artists blind.
But the moment we automate intellectual labor, suddenly doing things the hard way becomes a virtue.
This is what people are missing about AI.
A guy on vacation has an idea, and 72 hours later his agents have found a Maya site nobody had mapped.
We've been doing the same thing in Spain.
My agents swept 43 LiDAR sheets of public data looking for Iron Age hillforts. They checked themselves against 207 known sites, vetted 5,284 detections and threw out everything that didn't hold up.
One of what survived: a hillfort Spain's official inventory has in the wrong place.
The official map puts it on an empty slope. LiDAR shows nothing there.
528 meters away, on the next hilltop, there's an 18,000 m² fortified summit with a rim and a double ditch, just like the inventory's own description says. A 1956 aerial photo shows the same hilltop before the forest grew over it.
Before this they went through NASA's TESS data and measured, for the first time, a 22-day orbit on a star nobody had pinned down. Link below.
The data was always public. There were just never enough people to look at all of it.
Now anyone with a few agents can.
I asked Claude Opus 5.5 a ridiculously stupid question:
“Claude, how do we find a new planet?”
I have absolutely zero background in astronomy. I’m just a guy who likes building crazy things with AI.
But after seeing what Opus 5.5 could do, I wanted to try something much more ambitious than coding games.
What if someone with no scientific training could use AI to make an actual scientific discovery?
So we started searching NASA’s TESS telescope data.
And I think we might have found something.
Not a planet. Something else.
TIC 889317.
A binary star system astronomers already knew about, but whose eclipsing orbit may never have been documented.
Here’s what Claude found, and a separate Codex agent checked:
• An orbital period of 22.0435572 ± 0.0000021 days.
• A second eclipse that arrives a quarter of the way through the orbit instead of halfway, meaning a highly eccentric orbit (e ≥ 0.43).
• Independent confirmation from 13 years of ASAS-SN ground-based observations.
• No previously published orbital period in any catalogue or literature source we could search.
And we didn’t just trust the first result.
We ran multiple independent audits, rebuilt the analysis from scratch, checked the original observations, and produced a four-page research report with reproducible code.
To be clear: the binary itself was already known, and a 2025 paper flagged it as one of ~900,000 machine-learning eclipse candidates, without a period.
What’s potentially new is the orbital period, the eclipse timings and the eccentricity.
We haven’t established absolute novelty, and we still need an astronomer to review the work independently.
And there’s one mystery we can’t solve: Gaia’s data hints a third star may be hiding within 1″. We can’t tell which star is actually being eclipsed. That needs high-resolution imaging or spectroscopy.
So here’s my question to the astronomy community:
Did we actually uncover a previously undocumented orbit, or is there something we’ve missed?
The report, observations and code are open for anyone who wants to investigate.
This whole thing started with a guy who knows nothing about astronomy asking an AI how to find a planet.
Imagine what happens when millions of people start asking questions like that.
1
2
277
Luckey Faraday retweeted
Gemini 4 Carbon vs. Fable 5.2 vs. Opus 5.5.
Same AK-47 3D test. Three very different results.
Gemini finished in 10 minutes for $0.77.
Fable 5.2 took 58 minutes and $29.41.
Gemini absolutely cooked here.
Luckey Faraday retweeted
You Anti A.I. losers can fuck all the way off this is the coolest shit ever
NO PUEDE SER, ESTO ES UNA AUTÉNTICA BARBARIDAD 😭💀
Han conseguido transformar POKÉMON ROJO FUEGO EN UN JUEGO COMPLETAMENTE EN 3D.
Han recompilado el código original de GBA, lo han reescrito en RUST y lo han combinado con los gráficos y animaciones de Pokémon Let's Go.
¿LO MÁS LOCO? Los combates, la IA y las interacciones siguen funcionando con el código original de Rojo Fuego, pero ahora tenemos CÁMARA LIBRE Y UN MUNDO 3D AMPLIADO.
ESTO ES LITERALMENTE EL REMAKE QUE TODOS QUEREMOS 😭
“AI is taking away our purpose.”
If your entire purpose in life can be automated by a computer, perhaps the problem isn’t the computer.
I find it fascinating that people are terrified of AI replacing programmers.
Nobody seems particularly concerned that programming itself replaced enormous amounts of human labor.
Entire departments of people doing calculations, maintaining records, and processing information were replaced by software.
We celebrated that as progress.
Now software is learning to write software.
And suddenly automation is an existential moral crisis.
Interesting how the definition of progress changes when your own profession is involved.
“AI is going to make everyone dependent on technology.”
You already depend on technology to eat, travel, communicate, work, and stay alive.
Take away electricity for a week and modern civilization practically collapses.
But apparently, depending on machines to perform intellectual labor is where we draw the line.
We’ve spent thousands of years building tools to overcome our limitations.
Why should intelligence be the exception?
The people who think AI is destroying human creativity have a very strange definition of creativity.
Apparently, having an idea isn’t enough.
You also have to suffer through the execution for it to count.