In a good team, it would be joint responsibility. But you make a good point; typically the accountability would be split across engineering and product. So I guess thatβs a new interesting difference: you feel less connected to the engineering calls with coding agents, and so accountability slips
07.03.2026 02:09
π 0
π 0
π¬ 0
π 0
you produced it, exactly from the product manager perspective.
07.03.2026 00:42
π 2
π 0
π¬ 1
π 0
e velhice Γ© o pior, mas Γ© melhor do que a alternativa
06.03.2026 19:38
π 9
π 1
π¬ 1
π 0
PDF Accessibility and Standards β Quarto
Quarto 1.9 brings PDF accessibility and standards support, building on new tagging features in LaTeX and Typst.
If you are interested in PDF accessibility, there has been some big progress in Quarto.
Take a look at the latest blog post: quarto.org/docs/blog/po...
#Quarto #Accessibility #LaTeX #Typst
05.03.2026 15:21
π 19
π 7
π¬ 0
π 0
At some point a friend of mine was trying to tell me about natural gradients, which I guess is what this is also trying to do in a sense? (natural gradients look at everything under the Fisher curvature or some similar gobbledygook)
05.03.2026 20:02
π 1
π 0
π¬ 1
π 0
sure, I know what that term means, I totally am not reading that paper right now. No, not me.
05.03.2026 19:57
π 1
π 0
π¬ 1
π 0
I'm not raining on the parade, I think it might work! (ie, the math checks out) It's just that I feel like there will be a Hessian term in there somewhere, and I don't have an intuition for how big that'll be.
05.03.2026 19:50
π 2
π 0
π¬ 1
π 0
*Up to second order
But yeah, extremely cool! Have you seen the graphics tricks people do to simulate signed distance fields to a surface defined implicitly by f(x) = 0? It's, more or less, "divide by gradient magnitude", which won't surprise you at all
05.03.2026 19:34
π 2
π 0
π¬ 1
π 0
Brutal, but evidently correct in that context. The whole piece is worth reading.
05.03.2026 17:37
π 10
π 2
π¬ 1
π 0
Now try filing you reimbursements at a university 1 nanosecond too late
05.03.2026 17:18
π 5
π 0
π¬ 0
π 0
Pois Γ©, mas: Ney Matogrosso no auge do secos e molhados
05.03.2026 04:44
π 2
π 0
π¬ 1
π 0
You can almost see the trajectory in latent space. Pure markov chains on word sequences have no notion of word or structural similarity, and this does. But otherwise, yeah, 100%
05.03.2026 04:14
π 7
π 0
π¬ 0
π 0
Whatβs the line from Aliens?
05.03.2026 00:28
π 4
π 0
π¬ 0
π 0
Sure but Gemini did this after messing up some regression code or something
05.03.2026 00:28
π 58
π 0
π¬ 0
π 0
I might have made that exact joke in the company chat, yeah. Takes one to know one π¬π¬π¬
05.03.2026 00:14
π 0
π 0
π¬ 0
π 0
We donβt need to have a discussion of whether these things have any internal experience in order to have a good guess at what a transcript like this would do to someone whose head is not in the right place!
04.03.2026 23:34
π 25
π 1
π¬ 1
π 0
This was mostly unprompted and entirely accidental after it realized it had made a bad analysis (or something like it, I canβt quite remember)
04.03.2026 23:33
π 237
π 17
π¬ 24
π 47
Have you ever experienced a Gemini freakout? This is extremely bleak but completely unsurprising to me. Let me find a screenshot, hold on
04.03.2026 23:30
π 93
π 1
π¬ 1
π 0
Que nem o europeu elogiando o estadunidense: βpΓ΄, o paΓs de vocΓͺs tem o melhor ensino mΓ©dio do mundo! Pena que o pessoal tem que passar 4 anos numa universidade pra ver o que Γ©β
04.03.2026 22:20
π 2
π 0
π¬ 0
π 0
I also donβt think most people have tried hard to make LLMs good for *debugging*. Theyβre weirdly excellent at it (arguably better than they are at writing code, which is a genuine surprise!). You do need to want to use it in that way, and to be willing to put in the effort
04.03.2026 21:14
π 4
π 0
π¬ 0
π 0
It took us a long time to get our code base to a point where opus can be helpful, and I agree with you: *this* is where Iβm excited about LLMs for coding. *we* know what needs doing, but itβs a hell of a lot of different cases, and itβs a real drag to do the handling all manually. Opus helps a lot!
04.03.2026 21:10
π 5
π 0
π¬ 1
π 0
* typing out code is definitely one of several time sinks, if you take your job seriously and spend a lot of time exploring options
* it is not the only time sink, but it is one
* claude can help explore a design space in a less committal fashion
* this makes the harder parts of eng more tractable
04.03.2026 20:22
π 102
π 8
π¬ 3
π 0
Hey, for real. Think of this as a friend telling you to log off, for just a bit. Itβll be good for you.
04.03.2026 21:00
π 1
π 0
π¬ 0
π 0
Yeah weβre near the U and the turkeys are so funny!
04.03.2026 20:52
π 1
π 0
π¬ 0
π 0
Well, the awoo in Tucson is all the coyotes. But dude, it's like half as loud as the tornado awoos. It's wild.
04.03.2026 19:21
π 2
π 0
π¬ 1
π 1
I lived in Tucson before moving here; there's _way_ more coyotes there! They own the place pretty much lol
04.03.2026 19:19
π 2
π 0
π¬ 1
π 0
In Tucson the awooo is at 1am. Hereβs itβs at 1pm lol
04.03.2026 19:14
π 8
π 0
π¬ 2
π 1
This is the single weirdest Minneapolis thing I experienced since moving.
04.03.2026 19:03
π 25
π 3
π¬ 2
π 1
No one told me about this when I moved and it freaked me out!!
04.03.2026 19:02
π 3
π 0
π¬ 0
π 0
The new iOS is terrible, something simply Must Be Done about Apple's insistence on making everything rounder and more transparent to the point where I can't read any fucking text on the screen
04.03.2026 18:50
π 146
π 7
π¬ 10
π 0