Sabitlenmiş Tweet
Patrick's Law
"An organism can only create artifacts
which are less intelligent than itself."
English
CryptoPatrick
633 posts
CryptoPatrick
@cryptopatrick
Mathematics Undergrad @ChalmersUniv 🇸🇪 🦀 Digital Economics | Ai | Seamless Global Systems
Katılım Mayıs 2015
499 Takip Edilen221 Takipçiler
Do you think that a remailer-inspired design could work? where the source code is split up into chunks and each chunk is sent to a different model, so that no model has enough to generate the gist of the idea in the source, and the aggregation of translated chunks is then done on the local client?
English
@cryptopatrick Make sure to play with both big models (the big 3) and open source ones.
English
Been coding all day with DeepSeek V3 Pro (#rustlang).
Virtually free compared to Flash 3.5 and GPT-5.5, so it is very good.
1.5x to 2x slower than GPT-5.5, but at 20x to 30x cheaper, I am okay.
This is where parallelism is a benefit.
#WorkbenchCoding #zcoder
English
@jeremychone Thank you! Perhaps I'll try vibe coding something which will break up a larger code base into smaller chunks that can be individually ported with a weaker model, merged, and then tested for correctness.
Thanks! I'll have a look at Gemma-sized models.
English
@cryptopatrick I think it will be a little short. With 16 GB included, you won’t be able to run medium-sized models, only Gemma-sized ones.
Now 1k LOC is low, so that could be an interesting exercise. But it won’t be time-effective.
English
@dbreunig Perhaps something like:
icaris - hubris and ungrounded confidence in one's ability, which eventually meets a harsh reality, upon close inspection by the bright light of truth.
English
We need a name for this, because Armin is putting his finger on a problem that’s everywhere: people running their writing through an LLM because they think it makes it clearer, when in actuality it sands off all the detail.
Armin Ronacher ⇌@mitsuhiko
Has been a while since I wrote about agentic engineering, so this time around some learnings of maintaining Pi as a junior maintainer to @badlogicgames :) lucumr.pocoo.org/2026/5/24/pi-o…
English
@badlogicgames Just wanted to say thank you for sharing these 'recommended reading' tweets - they always popup when I sit down to drink my morning coffee at campus - a great way to start the day.🕺
English
@ObhishekSaha I'm more concerned about LLMs slowly turning mathematicians into Weight Technicians, and eventually into on-call therapists for depressed AI math agents.😀
English
A mathematician in
2023: This ChatGPT thing is cute and can talk!
2024: It's stupid though and can't really think
2025: It can solve Olympiad problems and any test I set my undergraduate students. Not so useful for deep real maths yet.
2026: Wow it can prove some actual real theorems. But it also makes a lot of slop and it's real hard to tell them apart.
2027: It's like having an army of supersmart, super-knowledgeable junior researchers always at my disposal.
2028: What am I here for? Am I even needed for mathematical progress?(For many mathematicians, the answer will be no)
2029: Wait. Remember what Hardy wrote? "The mathematician’s patterns, like the painter’s or the poet’s must be beautiful; the ideas like the colours or the words, must fit together in a harmonious way. Beauty is the first test: there is no permanent place in the world for ugly mathematics." And this thing doesn't really understand beauty. I am still needed to direct it towards beauty.
2030: Half the Millennium problems and many major open problems have been proven. Unending research directions have opened up. I need to understand what has happened.
2031: And to set future research directions based on my personal vision of importance and beauty. This is a magical time. We are doing things together we never thought possible
2040: Hello, is anyone out there?
Daniel Litt@littmath
Assuming models have complete advantage over human mathematicians in all areas of math (IMO not guaranteed in 15 years, but likely enough), this seems to be a question about the shape of society. Arguably the answer is not at all special to math--presumably the answer is basically the same for all knowledge work (and maybe all professions?). I think in this world the most likely background situation is that humans are not really useful for proving theorems, but nonetheless lots of basic and understandable questions remain open, including many that are open today (since I think these have basically unbounded difficulty). And lots of other questions of basic interest have been resolved. So human activities might include (1) trying to understand solutions, (2) trying to understand progress and obstructions to resolving open questions, (3) (non-rigorously) understanding mathematical phenomena. In all these activities, the purpose of the human is to serve as a locus of understanding. There's an obvious question as to why we would pay someone to do this. One plausible answer is that maybe we want to avoid complete disempowerment--at a minimum we might want people to understand what they can about what the AIs are doing--which requires development of human capital.
English
@ValerioCapraro I'm about as far away as one can be, from being a big name - but it has always been crystal clear to me:
Achieving AGI is a fool's errand.
CryptoPatrick@cryptopatrick
Patrick's Law "An organism can only create artifacts which are less intelligent than itself."
English
Finally, a big name has the courage to tell it: we are nowhere near AGI.
Demis Hassabis, CEO of Google DeepMind and Nobel laureate for AlphaFold, put it neat and clear:
"Today's systems are nowhere near [AGI]. Doesn't matter how many Erdős problems you solve… I think it's far, far from what a true invention, or someone like Ramanujan, would have been able to do."
This is the elephant in the room that many AI enthusiasts prefer not to see, or are actively trying to hide.
Erdős problems are well defined, often combinatorial, on finite spaces. They are exactly the kind of problems on which current AI can achieve spectacular performance with a lot of compute and knowledge.
A neural network can search a huge graph of possibilities. It can recombine existing knowledge at unprecedented scale. It can discover surprising solutions inside an already defined conceptual space.
But true invention is something else.
True invention is not only solving a problem.
It is inventing new objects, new dimensions, new connections. It is inventing new problems.
From resolving to inventing there is a discontinuity that we don't know how to bridge.
We are making extraordinary tools.
But we are nowhere close to AGI.
English
@Telekinetic18 Awesome! I hope that'sstill true in 8-12 months when I've finished working through this book.🕺
CryptoPatrick@cryptopatrick
📚 I have just started reading the book "The Rising Sea" by Ravi Vakil. 🗺️ It will take me a long time to read and work through this book. For fun, I'm going to document the journey here on my X account. ✅ Preface 🚧 Chapter 1: Just Enough Category Theory to Be Dangerous 🚏Section 1.1: Categories and Functors 🧗♂️: Done with page 3.
English
US made corresponding technology already exists – drone swarm control and intel mesh with use of AI for target acquisition and improved coordination to penetrate heavily contested Electromagnetic Warfare environments, enabling artillery systems to engage targets with increased hit prob..
English
@MykhailoRohoza The US is so far behind and out of this development.
English
Ukraine has just deployed a battlefield technology that feels like a mix of science fiction and a game of Call of Duty.
According to The Times, the Armed Forces of Ukraine have introduced a new system that radically changes the rules of modern warfare.
A single operator can now control an entire swarm of drones in real time, see the battlefield through artificial intelligence, receive instant targeting data, and coordinate strikes with sniper-level precision.
What once required dozens of personnel and hours of planning can now reportedly happen within seconds.
Russian troops are already referring to it as “the Ukrainian digital hell.”
English
Would love to see her play at a lower pace and for the offence to get her even more CnS looks, rather than creating her own shots by getting to her spots after a pass.
Anyone who has followed her career knows that Fudd has the potential to become the best perimiter shooter the W has ever seen.
However, with Fudd's many knee-injuries and the high pace of play seen in the clip, I'm worried that she's going to re-injure her knees.
English
Azzi Fudd tonight 🔥
• 12 points
• 3 assists
• 6/9 FG
English
🚨 Google DeepMind CEO Sir Demis Hassabis:
“Today’s systems, are nowhere near [AGI]. Doesn’t matter how many Erdős problems you solve… I think it’s far, far from what a true invention or someone like a Ramanujan would have been able to do”
it’s over for the Erdős hype
English
@HemantDotDev Tim hit the snooze button
on Apple's "Think Different" alarm.
🥱😴
English
End of an era.
Tim did a hell of a job turning Apple into a monster. Respect.
Fox News@FoxNews
BREAKING: Apple CEO Tim Cook steps down. Cook has served as CEO since taking over for Steve Jobs in 2011.
English
@starsallialign Maybe he's balancing two opposites - keeping the fans happy, but also keeping an elite weapon healthy.
Azzi hopefully has a long career in front of her and there's no need to rush things - everyone knows she's a generational player.
English
this is so unnecessarily rude on his part i don’t like how he’s speaking to reporters especially women…
Natalie Esquire@natfluential
“We’re headed in that direction.” - Jose Fernandez on whether Azzi Fudd will be starting for the Dallas Wings going forward
English
@wade_bowen716 @WomensHoops_USA 17 PTS in Q3 and set a record for most 3s by a rookie - are you starting to see why people are hyped up about Azzi? 🕺
NBA@NBA
AZZI FUDD IS HOOPIN' 🔥 Her 17 PTS in Q3 are tied for the 2nd-most points scored in any quarter by a rookie in WNBA history!
English
@cryptopatrick @WomensHoops_USA A number one pick shouldn't need an offense built to hide limitations; if she can't create efficiently, the hype might already exceed reality today
English
@JustDeezGuy "Not NEARLY enough practitioners today understand this."
Well, how many would be enough?
English
MacKay is indispensable for explicitly connecting information theory and machine learning—that what’s learned are axioms, i.e. compressed information, and rules of inference decompress information. Not NEARLY enough practitioners today understand this.
Abhinav Upadhyay@abhi9u
I originally learned it from a data compression textbook in college. From ML connection pov, the David MacKay book is well renowned (freely available) There is also a small focused book just on information theory from James Stone. Managing gigabytes also covers it but very much an applicattion book in compression related areas. I think there are many other books/tutorials published on arxiv.
English
A History of the Brain: From Stone Age surgery to modern neuroscience.
I only have 70 days until I leave for the US to start my PhD. I’m trying to finish reading all my physical books. I can’t take them with me because luggage space is tight and books are simply too heavy. The worst part is that I can’t stop buying new ones. I bought 6 physical books last month. I really need to switch to e-books immediately.
Goodreads@goodreads
It's Friday! What are you reading this weekend?
English
Yeah, the fact that he's reached full extension and the right hand is nowhere near under the ball should mean he's not transfering power into the ball as well as he could be - but it's hard to tell from one single photo - great shooters can handle lots of challenging (suboptimal) shot-types, with the no-dip three being one.
Guide hand looks fine to me.
English
Here’s the load up for his shot. Notice his guide hand (his left hand; you’re right looking at this) is too forward in its positioning and will impact the spin preventing backspin
His dominant hand being on the side of the ball (especially at this point of his shot) is undoubtedly awful
English
Koa Peat’s form is incredibly broken. The upper mechanics are unbelievably poor. New adjustments to his shot and it is going to need to be completely rebuilt
Jordan Richard@JordanRichardSC
Freshman 6’8 Koa Peat Klutch Pro day in front of NBA teams at Lakers Practice Facility 🔥 Koa was one of the best winners and mid range scorers in college basketball as a freshman
English
@ShowCaseShabazz Azzi's parents did a damn fine job
raising their daughter.
English
