
SIGMOD Systems Award for Amazon Aurora · Creator of AWS SimSpace Weaver · Prev. Microsoft Research + Amazon Research · High-school dropout → MSCS + AI, @UW ·
us-west-2
Joined February 2007
- Tweets2.8K
- Following1K
- Followers4.1K
- Likes7.9K
Replying to @elliotarledge @infinity_ai_
I was just about to reach out to you with a job offer. Congrats on the new gig!
Replying to @eriknewsha
I have some leads. What’s the commission structure? What’s the valuation criteria?
Replying to @trq212
Time to join the rest of us plebs and leave the 1000 agent slop machine runs behind. You are free!
I’m looking forward to seeing what kind of artisan craftwork you cook up now.
Replying to @tenobrus
Honestly, the least hobbled model is still Sol 5.6. Use it on xhigh with fast mode and blows Opus 5.5 away for quick spec'ing and prototyping. There's no point to using 6 or 6.1 or Astra. None of them follow instructions, and they spew junk code and machinery everywhere.
I understand what he is doing. It's basic factoring and arithmetic. Have you ever heard of the game 24? Stop being dense and open Google or ask your AI.
Christ. You are the last moron I can abide today.
There is absolutely nothing spectacular about Scott's speed here. In fact his pauses at certain points are indicative of someone who is out of practice.
All you have to do is use the internet if you are confused about what where the bar is instead of running your uneducated mouth.
youtube.com/shorts/U6Dni3QvJ…
You are right. I can't multiply 2281x3 or 9x9x5x5. So incredibly difficult. I should probably stick to web and mobile development...
Or I could pull my head out of my behind and recall that I've been playing variants of these games my whole life, and so have many others.
I'm not claiming to be special, Samuel. I'm just claiming to not be an idiot. And I have plenty of my own gifts without needing to be jealous of another's.
If you applied a modicum of reading comprehension before getting b*hurt on the internet, you would see that I very clearly did NOT downplay his skill or call it unimpressive. Again, it is just not really a display of mathematical or intellectual sophistication, and if you asked him he would certainly say the same thing.
en.wikipedia.org/wiki/Mathem…
Replying to @tenobrus
We also lost access to that Fable and it is never coming back.
Do you want me to add two numbers for you or not?
And I'm not knocking math, friend or downplaying that Scott is a literal genius. But these are just mental arithmetic tricks you can learn out of a book.
I was offering discernment, not ego. People don't need to take it so personally. I'm sure there are all kinds of things that are easy for you to do as well.
Replying to @AlyssaJoyJaffee @anothercohen
Buy these books for your boys, Alyssa and they can do it too.
- Mental Calculators Handbook
- Dead Reckoning
- Secrets of Mental Math
Replying to @danushman
I've gotten a lot of these too. Perhaps it's time to make a quick 7-figure payday off of old codebases, spreadsheets, documents, and stick-figure drawings.
Replying to @Twittte88417841 @anothercohen
Sure. Give me two numbers between 1 and 10 to add together and I'll tell you the answer after I get out my calculator..
Replying to @anothercohen
The party tricks he is doing in this video?
I can already do them. And so can most card counters and poker players.
Multiplication and division of 4+ digit numbers, square and cube roots, prime factorization, etc? Happy to train up and if paid for my time. But these aren't secret techniques that require savant-level abilities.
Here's books people usually recommend:
- Mental Calculators Handbook
- Dead Reckoning
- Secrets of Mental Math
Replying to @anothercohen
An incredible number of people can do this kind of mental arithmetic and it is also incredibly easy to learn. This is not a knock on Scott at all or to say it is unimpressive. Just really isn't what I would call a hallmark of intelligence on its own.
Replying to @xlr8harder
We know that they have a realtime pipeline for anonymizing traces. We don't know a lot of about they mine these, but we know they go into the training data for pre-train and RL. And that requires them to be annotated, which means humans are looking at some of them, even if they are building automatic annotation.
Whether that means there's a NOC with big screens showing insightful prompts and responses flying by is another story.