CryptoPatrick (@cryptopatrick) header
CryptoPatrick (@cryptopatrick) profile picture

🇸🇪 Mathematics Undergrad @ChalmersUniv | 🦀 Rust | 🦉 Ai | 🤖 Robotics | 🌍 Seamless Global Systems | 🔭 Looking for Internships

Joined May 2015
Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)Photo by CryptoPatrick (@cryptopatrick)
CryptoPatrick retweeted
要学好微积分和线性代数 归根结底一切高级的数学都是微积分和线性代数的各种变化 —— 这是中科院院士林群对丘成桐一些观点的提炼 林群原话:大学的数学则最主要是微积分加上线性代数,因为一切东西都是微积分和线性代数的变形。所以,我们如果把微积分和线性代数讲清楚了,那大学数学乃至整个数学就比较有谱了 在清华大学求真书院的座谈中:丘成桐多次明确告诫学生,要“深入学习微积分和线性代数”,打好基础才能支撑高阶课程 在评价基础教育与高考改革时: 丘成桐也曾公开直言:“微积分是所有科学包括数学的基础,不懂微积分就等于很多东西都不懂”,并指出“所有数学课程中微积分与线性代数最重要” 这句话是否很准确已不是那么重要,不用太纠结其严谨性 重要是微积分和线性代数是基础,这是共识,努力学习就好了
15
69
4
498
26,258
CryptoPatrick retweeted
Replying to @momonga_ol
You can take that analogy further - everyone serious about gains at the gym knows that diet and eating healthy is at least half the equation. If the analogy holds - what is "diet" for someone who wants to be successful in their studies? I would say following a strict Social Media Diet - making sure that your brain isn't feed crap from Twitter, IG, TikTok, Snap, and YT, everyday - all day long. Turn off social media and I'm convinced that it will 10x the effects of studying. 📚😀
1
20
CryptoPatrick retweeted
Replying to @creatine_cycle
Well, she's not looking for me 😞, because I had to look up these terms to understand what she's even talking about.😂 Here's my interpretation: 1. terminally online = in music terms, she's looking for someone who scoffs at what's being played on the radio, big clubs, huge concerts, and massive arenas - that's already "legacy" in her view. The bleeding edge is happening inside tiny venues you've never heard of - where artists (the rest of us have never heard of) are creating new genres that will take another 5-10 years to reach main stream channels. Early in her career, Lady Gaga visited underground queer clubs where she experienced a style and sound which was fresh, raw, pompous, but most of all - completely unrepresented in main stream channels. Lady Gaga basically dragged (pun) that entire experience (visiual and musical) onto the global music scene and ended up performing on the Super Bowl haltime show (the opposite of underground scenes). There's a scene in the movie The Devil wears Prada which explains this exact phenomenon. If you desperately want that second date with her, make sure to have your finger on the pulse of the bleeding edge of Internet Lore. 2. Technical - she wants someone who not only "talks the talk" but "walks the walk". She's not impressed by anyone trying to sweet talk her with technical buzzwords. Can you build an LLM from scratch in Pytorch, fix a nasty memory bug in C, profile and troubleshoot a tricky latency problem in a distributed network, understand and implement a deep learning optimization algorithm in a paper that dropped yesterday on arxiv, RL train a physical drone to deliver a payload to the weak areas of a moving tank, launch 1,000 agents to build an entire accounting software saas over the weekend? If you know how to do all of that, then she might mention your name to her friends. 3. New Media - she'll raise an eyebrow if your main source of news and gossip is the USA Today - "Really?". She'll tell her best friend if, less than an hour after something newsworthy broke, you've already dialed in on the live stream or Discord, where the Founder/CEO/CTO is being "interrogated" on the significance of the delta to what has already been "priced in". By the time that info-byte reaches traditional media suckers like me, new media folk like yourself can barely remember when it broke. 4. Storyteller - having grown up during the Google Ad era has made her cringe at any product person who talks in terms of specs and features. Miss her with any bragging about "better", "faster", or "cheaper". However, she will stop stiring that Soju drink when you show her a video of KD co-hosting a partnership launch with a japanese boy band and a AAA Game Developer, in front of a popup store on Shibuya crossing - with thousands of fans from an "Internet World" she's never heard of - only to be told that it's Pamba. 5. Agentic - if you're making long term plans, then she's already calling an Über. She'll tell her mom about you, if you're a snap-finger decision maker who understands that speed is a difference maker and idling is hella costly - unless your war chest supports patiently waiting for the last contender to show up in front of your cupertine gate. 6. Taste - she's looking for someone with fingerspitzengefühl that would make the collective SOTA SI cry binary tears of envy - someone able to capture, disseminate, and express the "man" in mankind. She'll tell her dad about you, if she's convinced that you're able to navigate the seas of synthetic mediocrity - withstanding the Siren calls of cheap generated AI slop - bringing her along on an exciting voyage to the Age of Abundance. There's a vector. Its magnitude is ...all of us. Its direction is ...neverending re-enactment of the past. Whoever that girl was looking for, I hope they'll remind her to slow down and enjoy life. We've already been here and nothing of this is truly new - it's just a dope remix.
1
263
CryptoPatrick retweeted
🤡 Vibe Coding 🥷 "Who's coding?"
1
1
105
CryptoPatrick retweeted
Replying to @NickRoolz @Google
😴 Tired Mankind: "Let's invent technology which raises the temperature of Earth." 😎 Wired Mankind: "Let's invent technology which raises the temperature of the Universe."
1
16
CryptoPatrick retweeted
ブチ込め。人生はマラソンだ。だが短い視点で見れば短距離走の時間がある。受験勉強。就活。仕事の繁忙期。ここにどれだけブチ込めるかが勝負だ。もちろん健康は大事だ。だかこの時期に寝食を後にして取り組む経験も尊い。一度は経験しろ。狂気的な量を短期でブチ込め。
9
175
30
2,432
1,411,587
CryptoPatrick retweeted
You seriously underestimate how much you can do with time. 1 year is a fucking lifetime. 6 months is enough to go from 0 to 100. Even 1 week of full commitment 14 hours a day is enough to push the needle so far forward you won't even recognize yourself. Just do more. Just do what your heart desires. Just be better.
111
3,208
131
28,398
2,584,541
CryptoPatrick retweeted
Excellent interview! I do feel that Neil kind of glossed over the complexity of things at the 09:55 minute mark: "The long lens view to take on this is that we have a form of intelligence that can tackle any verifiable problem." First: Verifiable...but when? For many verifiable problems, we don't know how long it will take before we find an answer. For this class of problems, the rationality of the budgeted compute spend B depends on: 1. p: the probability of verifiability within n attempts. 2. v: the value of the answer to the problem 3. c: the token cost per attempt To use an analogy, we know there's a hidden treasure worth $1,000 hidden in a vast corn field (v). We've rented a metal detector for $10 per day to try and find it (c). The size of the field and our efficiency per day provides the probability of us finding the treasure is (p). So, verifiability is only half the story - how long it takes is a hairy variable that has to be factored in. Perhaps this class of problems can be buffered by introducing a Polymarket-like prediciton mechanism, driven entirely by agents with settlement over x402. Who knows. Second: Problem-Solution Determinacy Who gets to state what constitutes a valid solution to the verifiable problem? 4+4=8 but so does 2+6. Models one-shotting bouncing balls inside a hexagon comes to mind. For many problems this introduces the need for pre-compute negotiations to settle what constitutes an acceptable verified answer to the problem. Either that, or an extensive Terms-of-Sevice to avoid conflict. Finally: Generative Asymmetry While problem formulation is a syntactic act which can be parallelized, theory-building which leads to problem verification has so far resisted parallelization. Basically, it's vastly cheaper to ask than to answer, which creates an asymmetry that inflates as billions of humans and agents ask for verification. Anyway, thanks for sharing this awesome interview. I enjoyed it very much. Ps. I'm assuming that I made some serious thinking errors in this tweet, so please consider it more as an expression of ignorance than a real argument.
1
3
686
CryptoPatrick retweeted
Replying to @evelovesolive
If we consider the token cost of trying to solve theseopen problems, and assuming that AI might identify billions more - should open problems be caregorized into either one of: 1. open unknowns 2. open problems ...and if so, what would be consistent classifying criterias for membership in each of the two categories? Kind of an interesting meta-question in itself, isn't it?🤔
1
52
One of the fastest ways to change your life is to increase the speed at which you resolve tension. Return the call. Send the email. Have the conversation. Resolve faster. Watch your life speed up.
74
2,097
34
17,246
247,232
CryptoPatrick retweeted
That's the second most asked question in my DMs, so here is a public answer to it. You do that by trying to learn as many things and concepts among various domains as you can, not being afraid of "wasting time" when you don't end up really liking something. Ironically, your college's syllabus is a great starting point. Keep aside your own biases, the teachers being bad, the CGPA, the "SCOPE" (fucking hate that word so much), and the job market - and evaluate your own college's syllabus based purely on your interest in learning something. That, is how you get started with finding your niche.
Replying to @VazeKshitij
How do you decide what you want to specialize in?
6
4
57
2,500
CryptoPatrick retweeted
It's not "a harness problem", and that high score isn't the AGI breakthrough it looks like. Using a custom harness, like Prime Agent, is like giving someone a calculator, on a math test which is supposed to be pencil only ✏️. The Opus 4.6 model jumped from 0.0% to 97.1% on ARC-AGI-3 simply by adding a harness. The Prime Agent's 95.5% score represents severe domain overfitting, relative to what the ARC-AGI-3 test is all about - using intelligence to solve new environments on first contact.
1
1
2
510
CryptoPatrick retweeted
“You keep on learning and learning, and pretty soon you learn something no one has learned before.” ― Richard Feynman
31
294
21
1,369
25,679
CryptoPatrick retweeted
Trust the process. The process is the process for a reason
44
30
400
21,609
Replying to @Aryvyo
every theorem is a gated quest and the boss is a 19th century prussian who already cleared it, named it after himself, and despawned you fight his lemma instead
1
5
497
CryptoPatrick retweeted
Lean 4.30.0 is live! The release notes highlight four areas: a new 𝚜𝚢𝚖 => tactic, 𝚌𝚋𝚟 out of experimental, a completed LCNF compiler backend, and a full Lake cache overhaul. On 𝚜𝚢𝚖 =>: "Unlike 𝚐𝚛𝚒𝚗𝚍 =>, which eagerly introduces hypotheses and applies proof by contradiction, 𝚜𝚢𝚖 => gives users explicit control over each step." On the compiler: the expand reset/reuse port "results in a ~15% decrease in binary size and slight speedups across the board." Full release notes: lean-lang.org/doc/reference/…
3
26
1
142
7,599
More big news from Mathlib: # The Formal Frontier Project The Mathlib Initiative is launching Formal Frontier — a new project focused on responsible, scalable, and open-source AI-driven autoformalization of mathematics. The primary goal of Formal Frontier is to bring formal mathematics closer to the research frontier in a way that is scalable, composable with Mathlib and its ecosystem, aligned with community standards, and genuinely useful for researchers. The Mathlib Initiative, a program of Renaissance Philanthropy, is funded by generous donations from Alex Gerko and XTX Markets. Why now? Autoformalization is advancing rapidly, and the choices made now will shape the foundations that the next generation of formalized mathematics is built on. We think getting this right matters, and that it should be done in the open, in close coordination with the communities who will actually use and extend these artifacts. What will we do? Formal Frontier will help establish standards and set a positive example for what formal mathematics in the age of AI should look like, both in the technical artifacts produced and in how projects at this scale engage with the wider community. The initial phase of the project will have three components: We will develop and release an autoformalization specification, in coordination with the community. This specification will articulate what a valid autoformalization looks like, covering how formal code should relate to its informal source, what counts as adequate coverage and faithfulness, and how artifacts document their relationship to Mathlib. It will also address the broader lifecycle of an autoformalized artifact, including expectations around human oversight, maintenance, licensing, coordination with related projects, and paths to eventual upstreaming. We expect this to happen quite soon, and will make follow-up announcements in the next couple of weeks. We will develop and release open-source autoformalization tooling, so that inference cost, rather than access to tooling, is the main limiting factor for researchers who want to autoformalize at scale. We will release autoformalized artifacts that embody the standards this project promotes, demonstrating in practice what responsible autoformalization at scale looks like while providing material that researchers can readily build on.
6
46
2
210
22,841
CryptoPatrick retweeted
I don't know why people believe that mathematics has been solved. It's not even close. I have been using big models frequently since July 2025, starting with Gemini 2.5 Pro (arxiv.org/abs/2510.04115), I use them daily since then. GPT has basically been running non-stop for me since 5.2. If you try to push these models beyond the boundary of the current literature, they fail. They don't tell you they fail, they introduce conditional assumptions which make the results weak. However, they make it very clear what is boundary knowledge and what is not, which was previously extremely hard to detect. Use them this way.
17
19
4
190
34,616
CryptoPatrick retweeted
Replying to @TradeSignalBot
Måndag nästa vecka kan bli lintressant. Inte lätt att se vad som skulle kunna pressa ner priset, men marknaden kommer säkert på något.
1
1
178