Mike Samuel 🟣

Mike Samuel 🟣

@mvsamuel.bsky.social

Programming languages person focused on software systems problems. Previously, first frontend engineer on Google Calendar, and was a security engineer who worked on the industrial-strength Mad Libs undergirding Gmail. Pro-trans-rights is pro-family.

916 followers947 following2.8K posts

For a project collaboration website, I want to avoid hover menus for entity references (e.g. project and people's names). More-info affordances should work with mouse, touchscreen, and screen-reader. Maybe instead of hovers, just a vertical oval that's reminiscent of the kebab menu.

Re his big idea slide, this is really the only viable approach. Treat the model as a big GIGO box; it is a pattern reproducer, so conformance with "the system must have property X" requirements has to come from elsewhere. For Odersky, the "medium" is the programming language and toolchain.

Hot take: serving a Markdown version of a website for agents and an HTML version for humans is the wrong pivot point. Better to serve a JSON version for programmatic users (API calls and Markdown) and an HTML version for browsers is the better cleave point.

Hot take: serving a Markdown version of a website for agents and an HTML version for humans is the wrong pivot point. Better to serve a JSON version for programmatic users (API calls and Markdown) and an HTML version for browsers is the better cleave point.

A structured form like JSON is a good basis for a web SDK, and it's lower token count; you can remove indentation from JSON but not from nested Markdown lists. And if you want any hope of dealing with prompt injection problems you need something like JSON string boundaries to hang provenance on.

Also, if you structure your HTML generation as a renderer over the JSON, then hashing your JSON gives you a good key for cache-control headers, and you have a great basis for an almost-static-site-generator approach to server design. Everything you need for it to work well with a CDN in front of it.

Dr. Hasan seems lovely, but the problem in Gaza is not that Israelis and Palestinians learn to hate one another. Even if it were, surveillance tech isn't helpful. The problem is the war crimes: Israel dropping bombs on civilians. Future thought crimes oughtn't distract us from ongoing war crimes.

A U.S. charity in Gaza is worried that teachers at its schools might stray from a curriculum it says promotes peace and tolerance, so it plans to record classroom conversations and use artificial intelligence to flag speech deemed “hateful” or “inciting.”

I'm guessing if you're raising funds for an education-focused NGO serving Palestinian children, you're going to get a lot of "what if they learn to hate" questions and fundraisers need to have answers even to naïve donor questions. I just hope that they can provide schooling without serveilling.

The more I see, the more I have to move Sen. Ossoff up in my rankings. And Rahm Emanuel down.

Mike Samuel 🟣@mvsamuel.bsky.social

AOC > Harris > Moore (love for a Marylander) > Pritzker > ...lit. everyone else in the party... > Kelly > Newsom > Emanuel

Mike Samuel 🟣 reposted

I am very pleased to share the program for LIVE'26: liveprog.org We have seven exciting presentations as well as a featured keynote by Brian Hempel! LIVE this year will take place remotely, so come join us over video-chat next Saturday (Oct 17) :)

Perl: There's more than one way to do it. Python (myth): There should be one -- and preferably only one -- obvious way to do it. Python (reality): There are multiple competing efforts to standardize 'the one' obvious way to do it, none of which will ever reach critical mass.

This is a legit great talk: accessible, engaging, builds intuitions. Bryan doesn't phrase it this way, but I love his counter-arg to "AI can solve that": There's intrinsic complexity in engineering when solving one problem interferes with solving others; sometimes you have to get your hands dirty.

Bryan Cantrill@bcantrill.bsky.social

I am surely not alone in having family over the weekend hitting me up asking "should I be worried about extinction?" As a response, I have found that my anti-Doomer talk from @monktoberfest.com 2023 (three very long years ago!) remains current and accessible: www.youtube.com/watch?v=bQfJ...

Path planning algorithms are not a hot research topic, they said.

John Berry@aniccia.bsky.social

Coco seems to prefer ~random-ordered roboclots. "A sidewalk deployment error caused several Coco robots to reroute simultaneously as they were leaving their station at the start of their daily routes, resulting in a temporary cluster" Coco, per @abc7chicago.bsky.social OP: ig.beccarunschicago

@kotlinlang.org's DSLs are just unreasonably nice. I've been using one to build ASTs in a compiler. class TreeFarm uses class Planting to allow specifying AST nodes as block structures, and there's some automagic that propagates node position metadata from injected AST nodes to synthesized nodes.

Mike Samuel 🟣 reposted

Replying to @welltypedwit.ch

so instead of bashing fil-C because it could take away attention from rust, what i would want to see would be people starting to think about how to combine the best of both worlds.

Mike Samuel 🟣 reposted

Replying to @mekka.mekka-tech.com

Member of Parliament Glenda Jackson said this within 48 hrs after Margaret Thatcher died. Elizabeth Warren couldn't get away with this type of ether. For US people: Listen to the whole thing, including the attempted rebuttal afterwards, and what the Speaker of the House responded to him with.

m.youtube.comGlenda Jackson launches tirade against Thatcher in tribute debateYouTube video by Barnet Bugle

Project 2025 is one more step in a RW project that's been baking since the Civil War, not a recent reaction to black people and women pushing for basic rights. There's no perfect time to push for needed change, nor an obligation to wait, and suggesting otherwise is coded language for "uppity."

Elad Nehorai@eladn.bsky.social

Many people, subconsciously or just as often consciously, blame BLM for how we got here. They see the reaction that happened to the movement, as well as to MeToo, and blamed the people calling for justice over the white backlash that always follows those calls. This is a pattern in US history.

The shielding weight / reliability tradeoffs have been known about for a long time. Here's a table detailing expected effects of radiation on processor's available at the time from a 1998 Robotics Institute thesis I was a research assistant for. publications.ri.cmu.edu/storage/publ...

Martin Kleppmann@martin.kleppmann.com

Analysis of the physics and economics of datacenters in space spectrum.ieee.org/orbital-data...

As a mathematician, let me just say the term should be "Christian-Judeo." The part you're subtracting out goes *after* the minus sign. ---- (Both christianity and judaism are fine topics in comparative religion but neither should be elevated above other viewpoints in state-approved curricula.)

Texas Tribune@texastribune.org

Rabbis and Jewish leaders criticized the biblical passages chosen by the State Board of Education as heavy on Christianity and dismissive of Judaism, reducing the term Judeo-Christian to “a fig leaf at inclusion.”

If you want to allow Wayback, their spider name is ia_archiver which you can use in robots.txt and page meta elements. Unscrupulous actors do spoof that, but they can already ignore a disallow * in robots.txt.

Josh Shepperd@joshshepperd.bsky.social

AI has started blocking the Wayback Machine from chronicling the web. Considering that nearly 40% of websites disappear after 10 years, this will enforce a giant loss in collective memory.

Cute. Not sure what to make of this. Mistral seems to think I do formal verification; important but not my lane. Deepseek seems to think I worked on V8, but I think all I contributed was answering Robert Griesemer's monotonically exasperated "do JS programmers actually do this?" questions.

Bethany Brookshire@beebrookshire.bsky.social

I saw this site which measures if you are "in the weights" (via @shiplives.bsky.social). At first I thought it measured how much of your work was used to train something, but that's not the case. In this case, it's how much an AI "knows" about you. I scored very high intheweights.com

To better understand our problems, what if we built a giant problem accelerator? Then we could smash our problems together at near light speed and analyze the resulting spray of micro-problems.