Stephen Wolfram
writings.stephenwolfram.com · Personal blog · on HN since 2007
Stephen Wolfram, founder of Wolfram Research, writes long essays on computation, physics, machine learning, ChatGPT, and his personal productivity setup.
What HN saysCommenters link his posts on combinators, Ada Lovelace, and productivity; one finds his machine learning articles excellent while noting his fixation on computational irreducibility.
608HN people shared it
349linked it in comments
465stories submitted
16Kstory points
3 months agolast shared
2007people per year2026
Most-shared posts
- 1What Is ChatGPT Doing and Why Does It Work? (2023)63 people · 171 points · first shared 2023 · 85 comments on HN
- 2Seeking the productive life: Some details of my personal infrastructure (2019)51 people · 494 points · first shared 2019 · 106 comments on HN
- 3Learning from 20 years of personal analytics36 people · 375 points · first shared 2012 · 59 comments on HN
- 4The Wolfram Physics Project28 people · 638 points · first shared 2020 · 323 comments on HN
- 5The Concept of the Ruliad16 people · 70 points · first shared 2021 · 19 comments on HN
- 6ChatGPT Gets Its “Wolfram Superpowers”14 people · 53 points · first shared 2023 · 6 comments on HN
- 7Wolfram Alpha as the Way to Bring Computational Knowledge Superpowers to ChatGPT11 people · 6 points · first shared 2023 · discussion on HN
- 8The Personal Analytics of My Life (2012)10 people · 1 points · first shared 2020 · discussion on HN
- 9Computational Law, Symbolic Discourse and the AI Constitution9 people · 24 points · first shared 2016 · 7 comments on HN
- 10What's Going on in Machine Learning? Some Minimal Models8 people · 239 points · first shared 2024 · 70 comments on HN
- 11Data Science of the Facebook World8 people · 211 points · first shared 2013 · 84 comments on HN
- 12Quick, How Might the Alien Spacecraft Work?8 people · 199 points · first shared 2016 · 59 comments on HN
- 13Combinators: A Centennial View (2020)8 people · 14 points · first shared 2020 · 2 comments on HN
- 14Untangling the Tale of Ada Lovelace (2015)8 people · 7 points · first shared 2020 · discussion on HN
- 15What Is Spacetime, Really?7 people · 208 points · first shared 2015 · 120 comments on HN
What HN people wrote when they linked it
fipar · 2026-05-26
Besides the productivity boost (and I know you already mention a boost to your well-being), this is one of the simplest yet effective things you can do to improve your cardiovascular health. I had a heart attack at 40 and 30 mins a day is the minimum recommended, so 60+ is great.
But back to your productivity angle: Stephen Wolfram wrote about the productivity benefits (for him) of walking while working: https://writings.stephenwolfram.com/2019/02/seeking-the-productive-lif…
0xCE0 · 2025-10-24
If someone wants to see SKs in their dreams, read following :)
https://writings.stephenwolfram.com/2020/12/combinators-a-centennial-v…
sw1sh · 2025-01-20
This is very cool! I'm wondering how does it compare to classical ATPs like Waldmeister or Vampire, can SupGen compete in CASC (https://tptp.org/CASC/)?
For example, can it proof this example from the recent Stephen Wolfram's post https://writings.stephenwolfram.com/2025/01/who-can-understand-the-pro… ? Which is basically about proving a•b=b•a from a single axiom ((a•b)•c)•(a•((a•c)•a))=c.
Is the source code available for this btw?
UltraSane · 2025-01-15
Wolfram really loves to talk about computational irreducibility.[1]
But I think his articles about Machine Learning are excellent. [2]
[1]https://www.google.com/search?client=firefox-b-1-d&q=%22computatio…
[2]https://writings.stephenwolfram.com/category/artificial-intelligence/
eesmith · 2024-12-16
Also relevant is "Untangling the Tale of Ada Lovelace" from December, 2015 at https://writings.stephenwolfram.com/2015/12/untangling-the-tale-of-ada… with 35 comments from the time at https://news.ycombinator.com/item?id=10709730 .
romwell · 2024-08-31
>The subject here and now is Wolfram and the irritation of many comes not from the work that Wolfram has achieved on his own but the utter lack of acknowledgement | recogonition of work done by others .. a substantial amount of work by others.
He wrote many essays on history of mathematics and science that IMO not just give credit where credit is due, but help immortalize these achievements.
We would be also be better off if the people who do interesting things that he wrote about bragged about them - so we'd get more exposure to the interesting things.
Ignoring bragging is easy. Finding out what Ramanujan's thought process was when he was coming up with his brilliant mathematics, in the absence of his own detailed notes, is pretty damn hard.
[1] https://writings.stephenwolfram.com/category/historical-perspectives/
john-tells-all · 2024-08-25
I agree visualizing neural networks is challenging. The flexibility and consequences of network connections is also confusing. If you decide to go down that route again, the best work I've seen by far is the recent one by Stephen Wolfram: https://writings.stephenwolfram.com/2024/08/whats-really-going-on-in-m…
deng · 2024-08-23
Say what you will about Wolfram: he's a brilliant writer and teacher. The way he's able to simplify complex topics without dumbing them down is remarkable. His visualizations are not only extremely helpful but usually also beautiful, and if you happen to have Mathematica on hand, you can easily reproduce what he's doing. Anytime someone asks me for a quick introduction to LLMs, I always point them to this article of his, which I still think is one of best and most understandable introductions to the topic:
https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
al_borland · 2024-08-17
I feel the same way. I find tracking exhausting, because if I’m going to do it I want it to be correct, which takes a lot of effort. In some cases it’s not really possible to be correct, which causes me a great deal of stress if/when I try.
That said, some people are into it. This post from Stephen Wolfram is something I think about often. It’s cool to see, but not something I could ever realistically do. Kind of like the people who take a picture of themselves every day for 20 years.
https://writings.stephenwolfram.com/2012/03/the-personal-analytics-of-…
ypeterholmes · 2024-08-01
I was basing my number on Wolfram's work here: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
"In human brains there are about 100 billion neurons (nerve cells), each capable of producing an electrical pulse up to perhaps a thousand times a second."
So maybe that's wrong. Thanks for the links. Even so if we look out 10, 100, 200 years from now, that level of complexity will be greatly surpassed by AI.
amanzi · 2024-05-06
Stephen Wolfram has a great post on this: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
It's really long! But worth the read.
koonsolo · 2023-10-25
For those interested, you can walk and work at the same time! This article was my inspiration and it works great! https://writings.stephenwolfram.com/2019/02/seeking-the-productive-lif…
simon_acca · 2023-09-09
As you'd would expect, Wolfram has a post about his home setup, including a custom built desk [0]. Also as one would expect, the post is as lengthy as it is insightful
0: https://writings.stephenwolfram.com/2019/02/seeking-the-productive-lif…
f0e4c2f7 · 2023-06-09
At a high level, Stephen Wolfram's "What is ChatGPT" video and accompanying blog post are a great introduction and may give you some ideas about what topics to learn from there.
https://www.youtube.com/live/flXrLGPY3SU
https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
rocmcd · 2023-05-31
I'd recommend starting here:
https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
It's pretty lengthy but doesn't require a PhD to understand. If you can get to the end of it you'll have a much better understanding of what's going on.
vharuck · 2023-05-05
As opwieurposiu said, embeddings are high-dimensional vectors. Often, they're created by classic math techniques (e.g. principal component analysis), or they are extracted from a model that proved useful for something else.
For example, a neural net model accepts a massive number of input values that directly map to the input. So those initial values don't add any info. But a layer further inside the model, with fewer values and probably close to the end, is smaller and should reflect what the model's learned. Like a lot of deep learning, three values work but don't give much insight.
If I'm wrong, I hope somebody more knowledge corrects me. I got my understanding from basic into tutorials and Wolfram's essay on ChatGPT: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
jamez · 2023-04-14
To be clear - the point is not to display what can or cannot be connected. The emphasis on this project is about the long and tortuous path chosen.
I take your point about the limits of knowledge graphs written manually vs LLMs. IMHO it's not either/or. We need both curation and statistical approaches, and when they are merged they give the best results. Just ask Wolfram: https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-s…
Edit: fixed link to Stephen Wolfram's blog.
Izkata · 2023-04-06
> It looks great on the surface but it isn't anything like the kind of sci-fi ai I think we're aiming for. Not in anything but a superficial way at least.
And then there's ChatGPT querying WolframAlpha:
https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-s…
Version467 · 2023-04-03
The recent post from Stephen Wolfram[1] is pretty good as an introduction, but I haven't seen any super comprehensive material that tries do dissect all the interesting behaviour we see in the really big llms. For that just reading the relevant papers themselves has been pretty fruitful for me. Some of them are actually very well written, even if you aren't used to reading scientific papers.
I can recommend the Sparks of AGI paper[2] and the toolformer paper[3].
Obviously there's much more out there, those three things are a pretty good read.
[1]: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
[2]: https://arxiv.org/abs/2303.12712
[3]: https://arxiv.org/abs/2302.04761
kzrdude · 2023-03-23
Just take a peek at the other thread about https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-s… and look at the "wrong Mercury" example. I think it's a great example of using an external resource in a flexible way.
guns · 2023-03-17
This is what Stephen Wolfram concludes in a recent article about ChatGPT:
> The specific engineering of ChatGPT has made it quite compelling. But ultimately (at least until it can use outside tools) ChatGPT is “merely” pulling out some “coherent thread of text” from the “statistics of conventional wisdom” that it’s accumulated. But it’s amazing how human-like the results are. And as I’ve discussed, this suggests something that’s at least scientifically very important: that human language (and the patterns of thinking behind it) are somehow simpler and more “law like” in their structure than we thought.
https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
ec109685 · 2023-03-16
This is a truly great writeup: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
anonymouskimmer · 2023-03-15
I don't see that this follows. It doesn't seem materially different than knowing that U always follows Q, and that J is always followed by a vowel in "legal" English language words.
https://content.wolfram.com/uploads/sites/43/2023/02/sw021423img21.png from https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
I imagine it's technically possible to do this in a piecewise manner that doesn't "understand" the larger board. This could theoretically be done with number lines, and not a geometry (i.e. the 8x8 grid and current state of each square mentioned in the comment you replied to). It could also be done in a piecewise manner with three ternary numbers (e.g. 1,0,-1) for each 3 square sets.
I guess this is a kind of geometric representation on the order of Shannon's Theseus.
a_c · 2023-03-13
To me, ML is tending towards religion. Religion in the way of blackbox processing. Seems no one can reason what happens within the network, the number of parameters, the number of layer, the choice of temperature, etc. Or it is just my ignorance. While the enlightenment nudges mankind toward science, could it be that the scientific breakthrough of ML is to knock us back into pre-enlightened age?
Per stephen wolfram [1]
> It’s worth emphasizing that there’s no “theory” being used here; it’s just a matter of what’s been found to work in practice.
https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
zuppy · 2023-03-06
from my understanding, anybody please correct me if i'm wrong, ChatGPT can not really invent anything, it can just generate text based on probabilities obtained from the mountain of source documents used for training it. it does not think in the same way we do, it is just amazing at writing coherent phrases (and very simple code).
there's a quite long article from Stephen Wolfram about how it works and this is why I belive it can't do that:
https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
ChatGTP · 2023-03-06
Kind of, existing things only with 0.8 points of randomness injected for "creativity", I'm overly simplifying of course but it is kind of like that.
I really recommend trying to get through this if you can: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-…
It raises some questions for me about how far this current model can actually scale / evolve without getting a bit too far out to be useful for human use. But it can (sometimes) connect things in ways that people have likely never seen before.
Maybe if you just clicked the button all day, you'd wind up with some really impressive new games people have never seen before.
curiousgiorge · 2023-02-15
Can someone help me understand how (or why) Large Language Models like ChatGPT and Bing/Sydney follow directives at all - or even answer questions for that matter. The recent ChatGPT explainer by Wolfram (https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-… said that it tries to provide a '“reasonable continuation” of whatever text it’s got so far'. How does the LLM "remember" past interactions in the chat session when the neural net is not being trained? This is a bit of hand wavy voodoo that hasn't been explained.
bumbledraven · 2022-10-13
"I do basically no videoconferencing. Screensharing is great, and critical. But typically I find video distracting." – Stephen Wolfram (https://writings.stephenwolfram.com/2019/02/seeking-the-productive-lif…
m12k · 2022-04-07
> I looked pretty hard a couple of years ago and couldn't find any explanation of the most familiar QM phenomena like the double slit experiment
I was following links from the OP, and while reading his Wolfram Physics one year update[1] I came across this:
https://www.wolframphysics.org/bulletins/2020/08/a-short-note-on-the-d…
[1] https://writings.stephenwolfram.com/2021/04/the-wolfram-physics-projec…
auggierose · 2022-02-07
Interesting read. Why I am not sure about "Ruliads" :-), it led me to a previous post of his [0], where he talks about how he found the minimal axiom system for classical propositional logic.
[0] https://writings.stephenwolfram.com/2018/11/logic-explainability-and-t…