Erik Meijer
@headinthebox
Followers
29,594
Following
0
Media
993
Statuses
10,665
Explore trending content on Musk Viewer
Russian
• 1078442 Tweets
Putin
• 589603 Tweets
Rich Homie Quan
• 430952 Tweets
Cristiano Ronaldo
• 235509 Tweets
Argentina
• 222959 Tweets
Chiefs
• 151110 Tweets
Messi
• 149892 Tweets
San Marino
• 131006 Tweets
#GHEstreno
• 70526 Tweets
Linkin Park
• 62193 Tweets
From Zero
• 49058 Tweets
Bolivia
• 37630 Tweets
Mahomes
• 36840 Tweets
مانشيني
• 36249 Tweets
Lamar
• 33809 Tweets
سالم
• 31994 Tweets
Chester
• 31188 Tweets
Di María
• 31158 Tweets
Emily Armstrong
• 23496 Tweets
De Paul
• 20267 Tweets
Vargas
• 19542 Tweets
Travis Kelce
• 16239 Tweets
Colin Gray
• 13064 Tweets
National Anthem
• 12482 Tweets
Derrick Henry
• 11123 Tweets
Fideo
• 10850 Tweets
The Emptiness Machine
• 10378 Tweets
When I left Meta, I told Mark that using AI, the Infra org could easily be reduced to 1/10 of their current size while doing a better job (*).
Large companies are like fast-food joints. You need lots of processes to create a bland product with consistent quality with average
If AI allows companies to be smaller, that will be a qualitative change. Lots of things break once organizations get big. We take it for granted, even in the startup world. But imagine if that stopped happening.
177
394
4K
76
214
2K
Folks, I left Meta yesterday. This time there were no layoffs or anything thrilling involved. In fact, I am more bullish than ever about Meta with the company’s increased focus on AI.
But given the incredible competitive pressure in the field, there is really no advantage to be
85
89
2K
Good lord, my team ported Chrome to Windows Phone in 2008/2009 and I nearly got fired for that. Can I please get my performance reviews rectified, with compound interest please!
24
324
1K
People often ask me about good books on category theory.
* Beginning:
* Intermediate:
* Developers:
8
264
857
My two favorite papers of 2017
*
*
show that deep learning is simply another incarnation of pure functional programming.
7
234
813
Very nice talk along the same lines of what I have been preaching recently. Deep learning is not simulating neurons, but pure functional programming based on regular calculus (instead of lambda calculus).
9
180
542
When I warned developers nearly a decade ago that Agile is a cancer that needs to be eliminated from our industry, many thought I was crazy. But Agile has completely marginalized the role of developers to just mechanically implementing user stories handed to them by the product
54
77
532
Nothing sparks more joy than removing code that does not spark joy.
4
121
518
Some of the awesome work my team at Facebook is doing to make developers more productive by having machines fix bugs for them.
15
224
492
I'm extremely impressed by the elegance and compositionality of TypeScript's type system, especially because all the advanced dependent types features so naturally spring from the desire to support real-world patterns in JavaScript.
9
122
466
If you use Haskell, you are indebted to SASL (), KRC (), and Miranda ().
Sadly, their creator, David Turner has left us this week.
Thanks David, you had a huge side-effect on my life.
11
155
430
Is the static vs dynamic types debate still a thing in 2023? I thought we had all moved on beyond that already.
"So farewell, TypeScript. May you bring much rigor and satisfaction to your tribe while letting the rest of us enjoy JavaScript in the glorious spirit it was originally designed: Free of strong typing."
472
520
3K
49
19
416
Huge congrats to the amazing
@kotlin
team at
@jetbrains
. It should be everyone's preferred language period!
13
110
370
One of my all time favorite business books is The Change Function (). It literally changed my life.
Many people use the following "geek" function to predict the success of their disruptive innovation as the following product:
P(success) = 10x better *
13
51
355
Enlighten me. Where is the *code* in "infrastructure as code"?
115
27
322
If you don't feel stupid, you are not learning fast enough.
21
602
275
Kotlin: [Internal Error] java.lang.NullPointerException
...
at org.jetbrains.kotlin.cli.jvm.compiler.KotlinToJVMBytecodeCompiler$analyze$1.invoke(KotlinToJVMBytecodeCompiler.kt:526)
...
I got the compiler of a null-safe language to crash with a null pointer exception!
@kotlin
14
38
262
When I try to explain to people that Universalis is designed to be amendable for formal verification, their eyes usually glaze over and I see them thinking "what the heck is Erik talking about". I guess the weights in my brain have been set by reading books like these, and
14
25
261
Jetbrains: The unicorn Silicon Valley doesn't like to talk about
9
366
254
We are being robbed ...
OpenAI has silently removed mention of GPT-4 32K () and AFAIK they never actually exposed it. Azure is deprecating GPT-4 32K in September ()
While GPT-4 Turbo, GPT-4o have 128 input context, they only have 4K
19
30
237
Code Tetris: The ultimate satisfaction of collapsing complex code you've worked on for weeks into something super simple with a couple of deft refactorings.
9
33
222
Not sure if I publically said this before, but kotlinx.coroutines.* is seriously cool; I don't see any reason whatsoever why anyone would use any other library (such as RxJava) than this for async/reactive programming.
@kotlin
@relizarov
7
40
223
As the stragglers are still heatedly debating F# vs C# vs OCaml vs Haskell, the language designers have already moved on to the next generation of AI-based programming languages .
Skate towards where the puck is going, not where it
18
52
217
> Another controversial feature that C# and Kotlin have but we don't want to add to Java is extension methods. []
I *really* cannot understand why the Java folks consider extension methods controversial. They are completely harmless yet incredibly useful.
29
17
213
Legit question.
What is the value of frameworks like LangChain, Autogen, crewAI, ... that basically build the same abstractions on top of a programming language that the underlying programming language already supports.
For example chaining is sequential composition, agents in
33
26
211
Category Theory for Programmers: The Preface | Bartosz Milewski's Programming Cafe
2
67
205
[3/3] The programming language elite has no f*ng clue about the real needs of the working practitioner. Is it too late now to say sorry?
41
75
195
One of the most influential books that caused me to become a total PL geek:
RIP Mike
3
40
198
Who says tech evolves fast ;-)
(2009)
={ 10 years }=>
(2019)
4
71
194
Unleashing the Power of End-User Programmable AI with Leibniz Automind & Universalis
I have had many investors ask me what I am building in my new venture.
When I answer that, based on the vision of Leibniz many centuries ago
Leibniz's program of a universal science (scientia
19
36
195
At this point,
@SwiftLang
is probably a better, and more valuable, vehicle for learning functional programming than Haskell.
22
169
187
"There is no point working on LLMs", then why still include your name in the list of Llama-3 contributors I wonder?
... Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; ...
Yann LeCun tells PhD students there is no point working on LLMs because they are only an off-ramp on the highway to ultimate intelligence
61
88
594
21
11
179
Some of the critical comments about my KotlinConf talk included
* too academic
* too advanced for the audience and today.
* It was hard to embrace :(
* Too complicated
That is extreme valuable constructive feedback, since my target audience is anyone that can write simple Excel
18
17
180
How have I been able to survive for decades without realizing this ...
6
24
174
"Making Money Using Math" was the topmost read article in the ACM Queue app in 2017. Now let's make it the topmost read article on the Queue website for 2018!
4
60
172
SQL language designer A: Let's mess with our users and put SELECT first, and have two distinct names for filtering.
SQL language designer B: Awesome, that will teach those stupid humans to never write SQL by hand but use ORMs or AI to write queries.
13
16
173
Last night, I fell asleep in a dream (first time ever, pretty weird sensation) but I only woke up once. Wondering where that lost stack frame is now. Or was it a tail call?
23
14
174
Fields Medalist Terence Tao explains how proof checkers and AI programs are dramatically changing mathematics
Cannot speak for the mathematicians, but you do not need a Fields medal to see that AI will make of formal methods and correct by construction
11
42
173
@Kthulhuan
There so many BS jobs and managers whose ego is proportional to their org size. Focussing on developers (like , , ...) or managers will only encounter resistance.
So my plan is to to bypass the middle men and directly empower
13
13
173
My team has a whole bunch of openings for post-docs and sabbaticals in the areas of Deep Learning & Uncertainty, ML for Software Engineering, ML for Systems, Differentiable & Probabilistic Programming, Simulations, .... and all that good stuff.
Ping me if you are interested.
5
57
169
That's my team. It was a true privilege to support such an amazing group of talented individuals.
(slightly outdated public overview page )
Meta layoffs hit 50-member ML research team focused on infrastructure
1
6
19
17
16
164
Interesting how MSFT introduced unlimited vacation just before their layoffs.
15
17
165
I actually found the debate yesterday incredibly inspiring. If either of these two can make it to the whitehouse, anyone of us has a shot. A prime example of the American dream.
10
9
165
"Programming is (should be) fun" is a tautology.
The real question is what is the best introduction to programming, SICP or Bird & Wadler 1st edition.
To me, the answer to that question is also crystal clear; "calculating is better than scheming".
8
29
162
Another Functional Programming pioneer and legend has left us :-(
RIP Arvind
7
32
162
The Baroque Cycle () is a fantastic set of books if you want a fictional overview of Leibniz's work.
7
14
163
Oh my! I just had to Google how to define C# extension methods ;-)
17
14
157
After (attempting to) attend a couple of online conferences (PLDI, ICSE, ICFP) this year, I must admit that despite the amazing efforts by the organizers, this format just does not work for me. I never cared for the talks, but went for the randomness and serendipitous encounters.
5
11
158
How did I miss this one. We need more distillers and less creators of tiny increments. Rejections because the work is not novel are the bane of progress.
10
73
152
Deep inside every NumPy clone in a statically type language hides an elegant type system and language design screaming with frustration to get out.
6
22
154
I am sure the champagne bottles are popping at Jetbrains right now.
17
17
154
But the hard part is to accept to cut your losses and actually hit delete to throw away all that temporary work!
5
27
152
Nearly every time I use (immutable) collections in Java/Kotlin, I get scared by the amount of copying going on, and wonder why on earth is there no blessed library of functional data structures?
16
14
146
[1/3] My intent with LINQ was to democratize data processing by making things more general and hence easier.
18
63
146
Waaaaaay back, was the technology that opened my eyes for the power and potential of Probabilistic Programming. Super happy to see that it is now open source.
While you are at it, don't forget to check out .
3
45
146
Holy f*ck, my old shit is used to do cool deep learning shenanigans!
1
30
141
It is both bitter and sweet to refactor code Tetris-style until it collapses into nearly nothing. The resulting beauty is amazing, but then there is the mourning about killing all those characters from the finite set you can type in your lifetime.
2
45
139
Yup, I had to cut my losses and drop the
#FP101x
MOOC because of all the complex and ugly shit that got piled onto Haskell.
12
34
142
I think is is exemplary of how *not* to deal with LLMs. Instead of thinking of them as being "intelligent" and hence try to make them do heroic zero-shot feats, we should leverage their strengths as dumb next token predictors and let them generate code or sequences of tool
A::B Prompting Challenge: $10k to prove me wrong!
# CHALLENGE
Develop an AI prompt that solves random 12-token instances of the A::B problem (defined in the quoted tweet), with 90%+ success rate.
# RULES
1. The AI will be given a random instance, inside a <problem/> tag.
2.
106
125
1K
10
15
141
Absolutely, RX was designed to make *event streams* into first-class values.
For async programming, or anything that requires back pressure, Rx is not the right tool.
#tautology
8
51
139
I read this trick from Tony Hoare to put your draft paper in a drawer for a while and then read it again to edit it. Time and time again I find this really works. Like wine, papers need aging.
7
15
139
Tuning weights and multiplying matrices is not creating intelligence. Why the surprise?
3
74
138
Probabilities deserve their own unit! I propose Bayes, as in the probability that this proposal will gain traction is 0.001 Bayes.
9
52
135
- Every engineer will
- *immediately* struggle
- with *any* Gradle build, large or small.
My guess is that most people just Google or ChatGPT (that should also be a verb) and copy and paste sections, praying that it works, cursing like a sailor if it does not.
🐘 Almost all engineers, working with large/complex builds will struggle eventually. Kotlin uses Gradle. Since I know a thing or two about Gradle, I invited my team and gave an introduction to the absolute basics:
Tasks and Configurations:
Recording 📽️
5
33
177
19
12
131
@migueldeicaza
You forgot the most greatest one of all Haskell. Who care about all those other baby languages.
6
4
131
It sounds cheesy every time I say it (in fact, I explained it to someone last week), but this book literally changed my life:
The Change Function: Why Some Technologies Take Off and Others Crash and Burn
By Pip Coburn
Every project that I have seen failing fell victim to the
6
20
130
The holidays are coming up. Here is the
#1
free book you should devour:
When you are done reading, comment here if you agree that "chain of thought" is just some weird form of continuation semantics for code written in natural language.
3
23
127
The book that made me fall in love with functional programming many many many years ago:
http://t.co/sTJSCCDB4X.
8
52
124
Academics could have so much more impact if they would allow themselves to properly explain existing ideas rather than just chase "new" work
5
99
126
To quote Dijkstra “The question of whether a computer can think is no more interesting than the question of whether a submarine can swim.”
3
18
126
Unit tests are for the weak. Real coders refactor on the fly and know they will come out victorious at the end (like I just did).
15
18
123
Always cheer for people that can make money using math.
In my pitch I tried to appeal to the second Futamura projection () as a way to fine-tune a model to fine-tune models, but I was laughed out of the room.
I guess there are
9
17
122
Gradient Descent: The Ultimate Optimizer
6
30
121
It is faster & much more fun to develop your own deep learning framework from first principles that trying to build any of the popular ones from source.
#WhatILearnedOverTheHolidayBreak
6
18
122
Or as I always phrase it, even if I give you a million dollars you cannot write a traditional algorithm that beats a neural net for these kind of applications. Stop fighting about your piteous JavaScript frameworks. You are getting blindsided.
3
49
117
I tried to wrestle with mutable state, and all my code turned into crap.
3
36
116
I didn't look at JavaScript for a while, but if these are *key* features, I can safely go back to sleep. Please wake me up when something actually interesting happens.
9
26
118
Folks, remember that when Agile & Scrum rightly go down in flames, I was one of the first to fire a bullet.
#Hackile
7
60
111
Typical technical interview questions:
My kind of technical interview questions:
2
26
120