Hacker News new | past | comments | ask | show | jobs | submit login
Knuth Airgaps and Knuth Buffers (taylor.town)
159 points by schmudde 10 months ago | hide | past | favorite | 50 comments



> I have been a happy man ever since January 1, 1990, when I no longer had an email address. I’d used email since about 1975, and it seems to me that 15 years of email is plenty for one lifetime.

> I have a wonderful secretary who looks at the incoming postal mail and separates out anything that she knows I’ve been looking forward to seeing urgently. Everything else goes into a buffer storage area, which I empty periodically.

Wow having someone do stuff for you is nice. Such deep insight. One day I wish I can afford to have someone do stuff for me so I too can experience this insight.


Secretaries were part of academia (something that came with his professor job). What is unique to Knuth is not the existence of a secretary, but his (attempted) refusal to receive emails except about bugs (but those he gets thousands of, as everyone in the world is encouraged to report bugs to him about every page he has ever written).

Anyway, Knuth's secretary (shared with several other faculty members) retired from Stanford several years ago (he dedicated one of his books to her), and although he hasn't updated the page, in fact he does all his own email now. (There is someone who comes in once a week to send his replies in the post — if you email Knuth about a bug you may get a reply a few months later from this person, asking for your postal address to mail you the reply: this will be a printout of your email, with Knuth's reply in pencil… and an enclosed cheque if your report/suggestion was accepted.)


Sometimes that someone can be a thing, not a worker.

I know that a couple of carefully crafted email filters increased my productivity almost twice and cut down unnecessary screen time almost completely. Knowing your vices can be more than half of the solution sometimes.


I managed to cut down the team lead's daily mail ingress from 3000 mails to sub 300 buy pushing the changes through the org to use a named distribution groups and other minor tweaks.

It's not easy to make the same on a personal account (at least without a personal domain name) but still can be done.


not to mention living in a time you didn't need an email to get a driver's license. heck in brazil, southafrica and some parts of india you cannot even get one without a freaking whatsapp (which you cannot get without a mobile phone account (which you cannot get without a bank account (which...)))

the other day i also had to take a selfie in a government owned app which required a phone with active google play store service, to allow me to see my own data, with no alternative method (being implemented so they say)


Ugh. Sorry to hear that, what a nightmare. First world countries tend to be better in this respect. I'm in the UK, and the worst that I've seen is that you need a phone to receive text messages for 2FA from the bank. What if you're abroad and can't receive those text messages??


I don't think an assistant is something you afford, so much as something your job requires.


In general, I think it's a good idea to be deliberate and judicious about your information diet.

In my experience, the more low-quality information you've got in your head, the fewer interesting ideas will emerge. Mainlining social media is very much like watering the fields with Brawndo the Thirst Mutilator.


I like the pond metaphor: if you want to see the bottom of the pond, it has to have been still enough for long enough for suspended sediment to have settled. And if there's enough wind for the surface to be choppy, you won't see past that, much less to any turbid layers.

> Email is a wonderful thing for people whose role in life is to be on top of things. But not for me; my role is to be on the bottom of things. — DEK


they say "you are what you eat" - physically and "you are what you perceive" - mentally..

so yeah, some (sometimes soft, sometimes drastic) hygiene is the way..


Pencil and paper is the best environment for working out anything at all difficult.

It would be nice if there were an environment to take a picture of written pseudo code and/or equations and spit out some matlab, numpy, or eigen code.


I'm partial to chalk, myself. The process of filling a board, erasing the parts that are no longer needed, then filling the board again distills your work down to the key ideas. And I just hate dry erase markers.


Yep. I know a problem is interesting if I reach for the pencil and paper. It's been on my desk today, which is good.

My PhD supervisor got me into writing with pencil and paper. I wrote near enough my entire thesis that way. It was only transcribed to computer for typesetting after it had been proofread at least once.


I like nothing (a walk) and a plaintext file when I need to remember many details.

Computers are distracting.


I don't know about paper, but having a large whiteboard in your room is amazing. A wacom tablet is great too.


The possibility of having a personal secretary seems almost as outlandish as having a butler. No idea how much higher I'd need to go in my current company's org chart to get to that, but it'd be pretty damn high.


This in particular seems like a task an LLM might do a decent job at.

A prompt like

> You are esteemed computer scientist Donald Knuth's secretary. Your task is to determine which email is prioritized for immediate delivery, and which email is to be gathered and read at a later date. Label them "IMMEDIATE" or "LATER"


At my company, one must be a VP to have an admin, and even then, the admins are shared across the VPs. Only some of the C-suite have dedicated personal admins, in fact. It's a dying luxury.


Back when I was a summer intern, I shared a secretary.

At the time I had thought I was slumming it because I was in a cubicle, instead of (at my previous summer job) an office with a door.

O tempora o mores!


I hear you can hire a part time one pretty cheaply


Thanks for submitting, Mr. Schmudde!

I'm eagerly awaiting your next post on Beyond the Frame :)

[1] https://schmud.de


If you ever try to get a knuth check, you'll receive back a printed copy of your email, along with his hand written notes and a response. Really cool!


I once received such a letter, in which Knuth explained how my bug report was in fact mistaken. I wrote back to him (on the same printout paper) to thank him for his reply, and included a check for $2.56. He cashed it!


(-:


I got back a comic, and another time a T-Shirt (with the MMIX instruction set).


I subscribe to the idea of Knuth buffers when developing algorithms. They always first go on a piece of paper before they ever see a terminal. Slowing down the process and avoiding distractions helps the process tremendously.


paper or whiteboard for me too. I sometimes find paper too messy and too "constrained" I guess is the right word. A largish personal whiteboard in my cubicle was perfect for algorithm development or just putting down the flow of a app I was writing.


Fast or close to real-time capture can encourage sloppy thinking and carelessness, neither of which Knuth is known for.


Knuth is such a likeable man. I get the same feeling watching him talk as I get watching a very experienced woodworker using his hands to create a masterpiece.


I wonder if internet access and instant-entry is /necessarily/ a distraction or impediment to deep thought, or it could be stomached to the effect of great productivity with an especially sharpened mind.

It saddens me that the most accessible repositories of information are those that, allegedly, dumb me down.


One thing I've found comically underestimated is books. It's not whether the information is physical or electronic, but what actual corpus of information is available.

There is A LOT of information in printed books that is not on the Internet.

There was a project to put all books on the Internet -- Google Books -- but that famously got tied up in lawsuits.

As a result, if your information diet consists of the Internet and not books, you're missing out.

I occasionally write something "obvious" from a book on my blog, and people are like "wow how did you figure that out" ?

---

For what Knuth is doing, he certainly doesn't need to read much on the Internet. Most of it is in books, or at the Stanford library (or whichever library he goes to).

He's probably so busy with books that the Internet seems UNINTERESTING.

If you want access to newer publications, the Internet is more efficient, but those are also available to the library. (Sadly, Scihub is the best source for those without university access.)

So yeah I'd say 3 main repos of knowledge are: the open Internet, printed books, and Scihub, and many people today only use the first one.


I've been getting deep into computer graphics recently, and having a handful of in-depth books on the subject has been immensely helpful. I don't have to spend time scouring terrible google results for answers when I don't even know the question

As an aside, ThriftBooks has been amazing for increasing my collection! I;ve gotten so many books for cheap


Yeah exactly, on top of the library, I buy old used books online, and they're dirt cheap, and dense with knowledge

No ads lol!

It's honestly sad to me when I see people scrolling through terrible web pages with tiny morsels of information, which are often "interested" or wrong.


Internet provides first class access to third class information. Just a couple days ago I was seeking some information relating to my "Intro to DSP" course at uni. After an hour combing through a bunch of unclearly stated and poorly answered questions on various stackexchange subsites and SEO-optimized hellholes, I just libgen'd a book my professor's textbook cited and found my answer in a couple minutes.

The question was how the phases add or subtract when looking at a phase graph of a cosine wave modulated DFT transform, not exactly rocket science.

It seems like the internet has dumbed down to the point where its front page is very surface level and always requires additional research assistance in the form of SearxNG, AI chats, or turning to less SEO prone engines like marginalia or even wiby to get good and honest results. I don't think adapting to human toxic environments like the current internet is a good model for the future, when we already have the tools to filter the wheat from the chaff.


Most deeply technical (and not computer-related) topics on Wikipedia have only one or two technical authors. They give you a view of the topic that is not particularly objective or complete.


My experience is that its mostly an impediment, see socializing here instead of thinking deeply.

But I've also used the internet to great effect when getting up to speed on a research topic because I had lots of access to high quality texts and tools (citation manager for tracking, spreadsheet for glossary of terms). Notably that didn't involve any communications, just searches.


I suspect Neal Stephenson has stopped composing on paper, given the precipitous decline in quality since Seveneves (or thereabouts; opinions may vary).


I think "Fall; Or, Dodge in Hell" was still pretty good. [0] I haven't been able to finish Termination Shock though. Too much "big rich guy will come save us" in the early parts for me.

[0] Though maybe that's just the hilarity of finally finding out who/what Enoch Root is after 4 books and many thousands of pages.


Yes, I thought 'Fall' was okay, and definitely better than the second half of Seveneves. I also liked the way the Enoch Root conundrum was solved, although I'm not sure 'hilarity' is the term I'd use.

His real turkey, to me, was The Rise and Fall of D.O.D.O [0], although he wasn't the sole author.

[0] https://en.wikipedia.org/wiki/The_Rise_and_Fall_of_D.O.D.O.


I think DODO really suffered with Neal's style of not explaining things to the reader that works well in things like Anathem but less well in a slightly madcap time travel story. I wonder if it would flow better on a second read where you had more info about what was going on, I remember spending a lot of time trying to figure out what was supposed to be happening and where the story was going and coming from.

For Seveneves the roughest part for me was actually the end of act 1 with all the mistakes you could see a mile off and act 2 had a lot of really fun stuff I won't mention for spoilery reasons but the adaptation of the whip and chain stuff that came up win act 1 was honestly a lot of fun for me.


> act 2 had a lot of really fun stuff I won't mention for spoilery reasons

Yes, there was some superb world building (notably the space habitats and transportation, as you suggest) in act 2 but (unfortunately) I felt that the plot didn't match it.


The idea that Knuth doesn't use email is basically just a practical joke that got out of hand. He actually received 31,997 and sent 19,910 emails between 1999 and 2019, and that's only including the emails from his work account at Stanford.

Sure, he uses email much less than you would expect from someone of has stature, but he still uses email about as much as the average person.


How do you know he received and wrote the mails.

Could have been his secretary in his assignment.

That's at least what his Stanford site suggests

https://www-cs-faculty.stanford.edu/~knuth/email.html


This appears to be a reference to https://epadd.stanford.edu/epadd/collection-detail?collectio... - a collection of "Don Knuth" emails.

The public versions are very redacted, but: 1) they seem to all be from his secretary, not him 2) most of the messages are not dictation or direct copies of messages Knuth wrote elsewhere.


His secretary may have typed them, but if they were mostly written by his secretary then Stanford wouldn't have paid to have them archived. Even though the messages are mostly redacted, actually going through and doing those redactions is still hundreds of hours of work.


"if they were mostly written by his secretary then Stanford wouldn't have paid to have them archived." - that is straight-up incorrect. Archiving emails is very cheap. And the redactions look to have been done programmatically.


So ePadd redacts everything except the named entities, but that still means going through each message by hand to ensure that the entities generated by the NLP software are correct. Plus the time spent fixing ePadd to make the import run correctly with his non-standard email client, the time spent negotiating permissions and restrictions related to the collection, etc.

C.f.: https://github.com/search?q=repo%3AePADD%2Fepadd++knuth&type...


I believe you but he does seem not to want to receive email because the page seems serious and it is still up.


He has a secretary to do that.


citation please?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: