Sign in with Twitter

Username:

Occasional OCaml programmer

126 Following   4,036 Followers   3,360 Tweets

Joined Twitter 6/21/09


@yminsky I tend to say most programming languages cannot add but only multiply
Retweeted by Yaron MinskyIndeed. The nearly complete lack of proper sum types in mainstream languages is a deep mystery. https://t.co/MpTWYLr9Xa
4/28
2017
Yesterday I added a second bool for a new state instead of properly refactoring it into a three value enum, and it bugged me all night.
Retweeted by Yaron Minsky
4/26
2017
@ezyang Or at least, that's what I'm hoping. I'm curious what other peoples' experience has been. @ezyang Writing is a real skill, and there are things that a good writer could add that few programmers can. @ezyang Also, writing various kinda of overviews and summaries. @ezyang But building focused tutorials, and going over existing documentation for fundamental pieces and improving it. @ezyang This is not just getting someone to relieve devs of the need to document their libraries. @ezyang For sure. Indexing is definitely important. But I'm really thinking about how you generate the writing itself. @shajra I'm confused: what does this have to do with documentation?But finding the right person seems like a tall order.I'm thinking about how we can improve the experience for developers at Jane Street, and better docs and teaching materials would surely helpAnyone have experience working with technical writers who work on documentation for developers. Any success/failure stories? @leonidasfromxiv Yeah, work on a refresh got stalled, but we're back on it now.
4/25
2017
.@let_def These numbers seem a little hard to interpret.Not as good as @let_def, who added 12,171 lines and removed 1,290,419.Looking at github, I notice that I added 1,791,601 lines and removed 1,745,585 lines in Real World OCaml. Editing at work.#ocaml A reminder to avoid depopts in your opam packages https://t.co/st7v2vRyY3
Retweeted by Yaron MinskyOne of the most rationally useful things you can cultivate is finding pleasure in the discovery of your own ignorance
Retweeted by Yaron Minsky
4/24
2017
Well that's the most troubling paragraph I've read in a while https://t.co/3O8h0z6TrJ
Retweeted by Yaron Minsky
4/22
2017
@cbarchak @paulg I agree on that. Not sure what to make of those stats. The definition of "exposure" is very weak. @cbarchak @paulg Very few learn when they're 10 or 11, which was more common when I was growing up. @cbarchak @paulg I bet most started with AP CS in high school. So, not in college, but not 10 years earlier either. @paulg Giving kids a chance to just smash around is a positive thing too. But principled presentation of programmin… https://t.co/TmRPqhHrBj @paulg I think that's a false dichotomy. https://t.co/d9x7mvVVeH does a good job of teaching in a principled way th… https://t.co/XKjGGBKQ7V @cbarchak @paulg Today's push to teach CS is largely vocational. I prefer Papert's view from the 70's/80's which wa… https://t.co/LYOGyuVUxe @cbarchak @paulg And that I think is because CS is now viewed as a safe and high-return career choice, so people pi… https://t.co/0rOvUBy0nY @cbarchak @paulg I also suspect that now CS attracts a much broader audience, including people who didn't fall in love with it as children. @cbarchak @paulg I think there are a few effects. One is that there was indeed a decline in teaching CS. @paulg @cbarchak I teach my kids to code early, but not because I think it's strictly necessary for them to be effe… https://t.co/VMLDsgwFvz @paulg @cbarchak I wish it were different, but at the same time, I've met many exceptional programmers who programm… https://t.co/F7cUGiwiWQ @paulg @cbarchak The objection is factual, not political. Few CS undergrads have deep experience with programming before school.
4/21
2017
@genTauro42 @scratch Tooling for Racket is great! Lovely IDE. I highly recommend.This is not the performance you were looking for. A new post on our blog from the inimitable Andy Myers: https://t.co/hNAjImqZY1
4/20
2017
And it turns out, the basic ideas are comparatively easy to come up with, but figuring out how to make it work in practice takes more time.I'm excited about this one because it covers a lot of hard-won lessons about the pragmatics of various forms of code-as-config.Another JS Tech Talk: Caveat Configurator, or, how to replace configs with code, and why you might not want to. https://t.co/YsoTuRwvwY
4/19
2017
Root system daemons [...] should be written in typesafe language like OCaml and Rust https://t.co/dYVOWqBda1 #linuxkit #MirageOS
Retweeted by Yaron MinskyI also like the tagline on the back. https://t.co/fTyYj5Lhtk @mensch04 CS51 course staff fleece. The class is taught in OCaml.Hey look, I'm now an honorary member of the cs51 choose staff... https://t.co/2po1jHKF6j(at Harvard, to be clear.)Just arrived in Boston! Looking forward to reprising my Effective ML talk at cs51, and talking about our code review system, Iron, at 3pm.Definitely on the list of things I'd rather wasn't vulnerable to a buffer overflow attack. https://t.co/JSRk1wbDumI'm on the CUFP (Commercial Users of Functional Programming) 2017 program committee & our CFP is open https://t.co/0lboy9NS1c Submit a talk!
Retweeted by Yaron Minsky
4/18
2017
@glaebhoerl @pressron @samth @matthewhammer @PrincetonWalker Very much so. @d_christiansen But I agree that this doesn't really seem that likely to work... @d_christiansen The idea is, with enough data, you can see whether Java->Python rewrites tend to fare better than Java->Python rewrites.Also, probably not enough cross-language rewrites out there.Sadly, huge selection effects abound there.Random idea: identify cross-language rewrites, use some quality measure to see effect of language choice.So, what would be a good way of extracting useful information about language quality from the natural experiments out there? @pressron @samth @matthewhammer @PrincetonWalker @danluu's post on this subject highlights how hard it is to genera… https://t.co/6tuFDQjGsY @pressron @samth @matthewhammer @PrincetonWalker But I'm not excited about grabbing random numbers from it and calling it evidence. @pressron @samth @matthewhammer @PrincetonWalker I love telling stories, and I do think Jane Street's success story is informative. @pressron @samth @matthewhammer @PrincetonWalker Or, to repeat an old joke, the plural of "anecdote" is not "data". @pressron @samth @matthewhammer @PrincetonWalker Meta-analysis of data whose sample bias you don't know doesn't necessarily improve matters. @pressron @samth @matthewhammer @PrincetonWalker But that's not impressive, or even interesting. @pressron @samth @matthewhammer @PrincetonWalker But...but...with different programmers! I'm sure I can find progra… https://t.co/beOBq7rTTO @pressron @samth @matthewhammer @PrincetonWalker In the end, our experience is compelling good evidence that using… https://t.co/maAy87o041 @pressron @samth @matthewhammer @PrincetonWalker But those numbers would only serve to muddy the waters. Precisely… https://t.co/MJrnKoUDPi @pressron @samth @matthewhammer @PrincetonWalker I can come up with a long list of numbers describing how new, post… https://t.co/BBl7KjZc3i @pressron @samth @matthewhammer @PrincetonWalker But again, useless: we hire a lot of people who really like FP, so… https://t.co/ma7GzDF0LB @pressron @samth @matthewhammer @PrincetonWalker I can also say some words about what learning time is now. Kids ou… https://t.co/vG4J8g06s9 @pressron @samth @matthewhammer @PrincetonWalker But it's all useless. New team was composed of different people, a… https://t.co/wNDCgdLas4 @pressron @samth @matthewhammer @PrincetonWalker Also, defect rate was lower (previous attempt abandoned because of bugs.) @pressron @samth @matthewhammer @PrincetonWalker I know rough numbers: team size was similar, time to completion ab… https://t.co/uUHqD9dGj7 @pressron @samth @matthewhammer @PrincetonWalker Not terribly quantitative, though. Not sure what numbers would be helpful... @pressron @samth @matthewhammer @PrincetonWalker And spoken about them: https://t.co/ZVOP7wDjPU @pressron @samth @matthewhammer @PrincetonWalker Well, we've reported anecdotes: https://t.co/L6TY1Mdl24 @pressron @samth @matthewhammer @PrincetonWalker But it's hard to find scale and time without large numbers of conf… https://t.co/5yk9CQqicd @pressron @samth @matthewhammer @PrincetonWalker My view is that you need real scale and time to see the (quite lar… https://t.co/pIUYfTqsDK @pressron @samth @matthewhammer @PrincetonWalker But, of course, you don't believe that weak evidence, and you shou… https://t.co/zVc7WYxWy7 @pressron @samth @matthewhammer @PrincetonWalker What do you mean? My experience shows "weak evidence" of a massive… https://t.co/hwGhxEaRLn
4/15
2017
@samth @matthewhammer @PrincetonWalker Also, as a personal matter, I'd rather not be in the group that's randomized… https://t.co/JujbkBtL0e @samth @matthewhammer @PrincetonWalker But man is that expensive. @samth @matthewhammer @PrincetonWalker But get a few teams at reasonable scale adopt a technology and give them 18… https://t.co/DHvSycvw72 @samth @matthewhammer @PrincetonWalker The thing that's hard to experiment on in the lab is scale. I think you can… https://t.co/C0jM10iiZt @samth @matthewhammer @PrincetonWalker I'd love to see large-scale, believable research. But it would require a mas… https://t.co/rl9xj11Jc1
4/14
2017

0