R. A. Dehi @radehi@qoto.org

Can you spot the circles?

It took me a long time. Once you see them, they'll seem obvious.

This is the Coffer Illusion, by Anthony Norcia

Apple has finally killed its ill-conceived plan to scan photos for CSAM. This is a direct result of work by experts and activists. Speaking up is important and sometimes we win.

https://www.wired.com/story/apple-photo-scanning-csam-communication-safety-messages/

Here's my new essay for Crooked Timber, "Your platform is not an ecosystem."

It's a big essay and work in progress from me, and part of a broader project with Robin Berjon who greatly improved this draft and provided the best line in the piece.

What if we didn't just glaze over when people call their closed technical architecture an "ecosystem"?

And how would the internet look if we saw it as a real but damaged ecosystem that we can yet repair?

https://crookedtimber.org/2022/12/08/your-platform-is-not-an-ecosystem/

Shared by my Daughter
"I need privacy, not because my actions are questionable, but because your judgement and intentions are"

In response to "if you have nothing to hide you have nothing to fear"

So often, I'm reminded about the broken state of IT today. Barriers to entry everywhere are ridiculously high.

Self-hosting anything is incredibly stressful, often requiring you to be skilled as a system administrator to the point where you could make a living off of it.

I know why, of course.

Capitalism rewards enclosure. Money flows to companies who want customers on their terms, so solutions are designed to service thousands of people and for professionals to operate.

We need #SmallTech.

Check this out -- major cities in North America replaced by major cities across the Atlantic at the same latitude
https://www.reddit.com/r/MapPorn/comments/d8tbs5/major_cities_and_towns_in_north_america_replaced/

We (meaning Daniel McNab in this case) just deleted the GLSL pipeline from piet-gpu, meaning that the WGSL version is the one being developed. That means the end of piet-gpu-hal.

I have very complicated feelings about this. It is the right decision, but there was also a lot special about piet-gpu-hal. The world needs a lightweight, portable runtime for GPU compute.

I have a blog outline on this (https://github.com/raphlinus/raphlinus.github.io/issues/86) but not sure when I'll write it. These days, I'm prioritizing coding.

Found a really nice #opamp to design my #UVC detector for my #bespoke #FPLC project. It has really low input bias current and offset #voltage. It's a bit slow with a slew rate of about 0.35v/us (would take about 14 #microseconds to rise from 0 to 5v) but, for my needs, this is perfect. Also found a better #UVC detector that has high sensitivity around 280nm. Time to draft a #transimpedence amplifier #circuit #PCB!

What happens when you use #autodiff and let your nonsmooth iterative algorithm goes to convergence?

With J. Bolte & E. Pauwels, we show that under a contraction assumption, the derivatives of the algorithm converge linearly!
Preprint: https://arxiv.org/abs/2206.00457

I will present this work this week at #NEURIPS2022

I can’t with these essays on leaving Twitter—the digital paradise—that fully paper over *years of harassment* organized on the platform (gamergate, yourslip, countless other examples). the company finally addressed harassment several years back (because advertisers were complaining) and progressively tweaked measures but it was by no means flawless by 2022.

#Markdown and #CommonMark are often used interchangeably. The latter refers to the formal specification published in 2014, resolving the syntax ambiguities of the original Markdown spec.
CommonMark and “#pandoc;s Markdown” differ in subtle ways, which is why pandoc has an extra CommonMark parser:

pandoc --from=commonmark

https://commonmark.org/

plan9port (Plan9 from User Space) running on NetBSD/i386.
Pretty sleek and very lightweight!

Two screenshots of my 1024x768 display, appended horizontally.
#plan9 #netbsd #desktop #screenshotsaturday

Perhaps at some point I'll write a thread on my deep concerns about our reliance on Google Scholar.

For now, though, why on earth does Google Scholar not let you sort your search results?

You have basically one choice: to see them in "relevance" order—and we're not even told the secret formula used to determine relevance.

You can also sort by date—for papers from the past year only.

It's really crazy that a mature tool supposedly designed to serve the community would be severely limited.

Fun fact: sharing this link on Mastodon caused my server to serve 112,772,802 bytes of data, in 430 requests, over the 60 seconds after I posted it (>7 r/s). Not because humans wanted them, but because of the LinkFetchWorker, which kicks off 1-60 seconds after Mastodon indexes a post (and possibly before it's ever seen by a human).

Every Mastodon instance fetches and stores their own local copy of my 750kb preview image.

(I was inspired by to look by @jwz's post: https://mastodon.social/@jwz/109411593248255294.)