When not programming or teaching CS @ucsd_cse, I go to the dog beach. https://xitter.cf/t.co/i6J3Qronk4 | https://xitter.cf/t.co/Omunb4WqwB | https://xitter.cf/t.co/iEAaulUul8

San Diego
Joined July 2010
Check out our Assistant Professor position @ucsd_cse! apol-recruit.ucsd.edu/JPF046… Since it is 2026, it is worth quoting explicitly from the ad that this not AI-specific: “The department is looking for exceptional candidates in all areas of Computer Science and Engineering.”
2
2
1
27
9,551
Joe Gibbs Politz retweeted
Replying to @lexi_lambda
A footnote from Stephen Dolan's Algebraic Subtyping, in reference to a paper by Redko: "This paper is out of print, in Russian, and quite difficult to get a hold of, so the interested reader is directed to a shorter proof by Conway [Con71, p. 105], which is merely out of print."
5
51
1,779
Joe Gibbs Politz retweeted
What happens in a major incident where the frontier models all become unreachable?
This quoted post is unavailable.
19
13
5
145
14,760
Joe Gibbs Politz retweeted
Remember SETI or Fold@Home? We need OpenSource@Home, where we use the tokens that would expire in everyone's subscriptions to contribute to the public good.
1
1
7
654
Joe Gibbs Politz retweeted
Replying to @garybernhardt
Please develop a deep understanding of the system without looking directly at it.
1
1
15
601
Is sharing stuff like this worth it anymore? IDK but this solved a real problem for me. Happy to provide the output of the few hundred k of frontier tokens, or even just the idea that if you have a Paper Pro you can make a better frontend for projecting. github.com/jpolitz/rmlive/bl…
10
363
Joe Gibbs Politz retweeted
Replying to @Thinkwert
"daddy, I had a nightmare" "about what?" "about school" "don't worry, you'll have those your whole life!"
1
20
283
Joe Gibbs Politz retweeted
Replying to @zack_overflow
I'm not even a coding agent hater it often is smarter and more knowledgeable than me I just wish that was the case 100% of the time instead of 99%
2
1
1
25
1,331
Joe Gibbs Politz retweeted
Replying to @hassanrahim
they're trying to cut open the bird to find the song
102
2,085
250
16,223
929,967
Joe Gibbs Politz retweeted
If you had a ticket to Grace Hopper and have not cancelled your plans to go to San Francisco, go to @github's LinkedIn right now if you'd like to attend GitHub Universe for free ❤️
5
33
172
12,127
Joe Gibbs Politz retweeted
3. The details of the JSArrayBuffer::views_or_detach_key_ technique were all developed by agents. Human guidance here was: "treat cage-resident state as attacker-controlled and look for assumptions made by non-V8 consumers that the V8 sandbox itself does not promise to uphold"
1
1
5
191
Joe Gibbs Politz retweeted
Replying to @engineering_bae
DrRacket is the IDE, Racket is the language! I'm forever thankful that my CS curriculum started with functional programming! It puts you in the right headspace (and forces you to learn the right mental models) to more quickly learn higher-level concepts. A great compromise for those overwhelmed by the parenthetical syntax is Pyret! pyret.org/index.html
1
1
1
26
1,936
Joe Gibbs Politz retweeted
Replying to @mraleph @realamlug
There is no magic pixie dust which just allows you to AOT compile JS without making sacrifices on something (perf, semantics, correctness).
1
2
15
2,427
Joe Gibbs Politz retweeted
We discovered that it's possible to create an AI-driven computer worm using an open-weight model that anyone can download. This work was conducted in a lab walled off from the outside world, & shared only after removing details that could aid bad actors. nytimes.com/2026/06/02/techn…
2
18
3
86
19,362
Joe Gibbs Politz retweeted
Replying to @joseph_h_garvin
Be careful asking models to generate UML, under Anthropic’s new AUP that’s technically considered cruelty to the AI
5
1
45
1,255
Ah yes agents with human use APIs, clearly that's next whether I like it or not. This is already a startup isn't it.
Replying to @clairlemon
Is this argument “LLMs can’t make a pizza appear at my door?” Because there is a fairly straightforward way for them to make a pizza appear at your door. Or is it “Well they can’t use humans; that’s cheating”, in which case cheating is delicious if slightly fattening.
1
126
Joe Gibbs Politz retweeted
Replying to @clairlemon
Is this argument “LLMs can’t make a pizza appear at my door?” Because there is a fairly straightforward way for them to make a pizza appear at your door. Or is it “Well they can’t use humans; that’s cheating”, in which case cheating is delicious if slightly fattening.
6
2
2
207
9,365
Joe Gibbs Politz retweeted
Replying to @ThePrimeagen
Other people have covered why this doesn't cover all irrationals, so I'll just instead share a cool math fact about all this: Mathematicians split the irrationals into two groups: the algebraics and the transcendentals. Algebraics are irrational numbers that are the solutions of polynomials (with integer coefficients), like `sqrt(2)` is the solution to `x² - 2 = 0` and `sqrt(2) + sqrt(3)` is the solution to `x⁴ - 10x² + 1= 0`. Every other irrational, like π or e or `ln(6)`, is transcendental. Now the cool part: algebraics are countable! To see this, you can encode any polynomial as an array of integer coefficients, starting with constant. So `x² - 2` becomes `[-2, 0, 1]`, and `x⁴ - 10x² + 1` becomes `[1, 0, -10, 0, 1]`. So "the set of all polynomials" is the same size as "the set of all arrays of integers". Using your shower approach, we can show the set of all arrays of integers are countable, meaning the set of all polynomials is countable. And then, because each polynomial has at most `len-1` roots, the set of all roots — aka the set of all algebraic numbers — is countable too. Incidentally, if you're a fan of gremlin languages, J has the `p.` operator, which automatically finds roots of polynomials. ``` p. _2 0 1 +-+----------------+ |1|1.41421 _1.41421| +-+----------------+ ```
2
3
50
1,712
Joe Gibbs Politz retweeted
"Inspiration is preparation plus panic" Kent Beck on creating XUnit, the first unit test framework - SPLASH/ISSTA keynote today
2
2
17
659