Mark T. Tomczak

https://mastodon.fixermark.com/@mark

Career software engineer living something approximating the dream he had as a kid.

homepage: https://fixermark.com
blog: https://blog.fixermark.com
GitHub: https://github.com/fixermark
BlueSky to Mastodon bridged account name: @mark.mastodon.fixermark.com.ap.brid.gy

Public Key

npub10zv2wul2hvcg6z9445zjdm5yldmyc6sqyr0u4xnnyygtfeym746s2aga0j

NIP-05 Address

[email protected]

Profile Code

nprofile1qqs83x98w04tkvydpz666pfxa6z0kajvdgqzph72nfejzy95ujdl2agpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dssjg6gp

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub10zv2wul2hvcg6z9445zjdm5yldmyc6sqyr0u4xnnyygtfeym746s2aga0j

Show more details

Last Notes

2026-07-13 16:37:40 UTC

- reply

What are 'asserts' in this context?

2026-07-08 17:48:15 UTC

- reply

I was unclear on whether they locked the doors or just lied to the kids (told them there was a maintenance issue and pulled over and then the kids voluntarily stayed in the car). I'm seeing both stories.
I cannot remember right now whether you *can* lock someone into a Waymo against their will; I think the door has unlock-on-pull handles on the inside. But I might be wrong!

2026-07-08 17:06:09 UTC

Ah, good. We're now in the "robotaxi will call the cops on you" phase of the experiment. This was always bound to happen.
https://www.wpxi.com/news/trending/waymo-car-delivers-misbehaving-teens-california-police/UCOGYIJRHJG2LMHYWJQ3PH7VEY
#robotaxi #waymo #av

2026-07-02 18:59:34 UTC

- reply

Uh... Mine has AI. 😕
I mean, I don't use them often, but the "find text in picture" and "translate from German" buttons are definitely doing *something* smarter than me.

2026-06-22 02:21:23 UTC

- reply

I've heard that too. And I'm not sure it's so much "hated each other" as "spent most of their lives living on top of each other." When you're fundamentally different personalities, that can be exhausting, especially if it's filtered through a professional relationship.
(There's a documentary on this, "The Boys," that I haven't gotten around to yet. Yes, same name as the superhero series, though the subtitle is "The Sherman Brothers Story").

2026-06-22 01:48:58 UTC

Currently digesting [this course](https://modernaicourse.org/ ) from Carnegie Mellon University on an introduction to modern AI.
I'm not terribly deep in yet, but I think I can give it a soft recommendation for anyone who has a bit of a programming background, a bit of a linear algebra background, and a bit of a calculus background and is interested in the modern state of the art regarding AI (in particular, LLMs, their capabilities and limitations).
And, hey, it's free.
#AI #LLM #CMU

2026-06-21 21:15:22 UTC

🎶 Miracles from polycules 🎶
(The Sherman Brothers would never) 😉

2026-06-15 18:08:41 UTC

- reply

The approach I've seen work best is AI as an automated reviewer in the pull request workflow. It can false positive and false negative, but in practice the false-pos rate I'm seeing is like 1-5% (and much of that corner case stuff, like "yes you're right but that is autogen code that cannot be fixed at this layer").
It needs human in the loop, but it doesn't get bored in the same way humans do.

2026-06-15 15:08:26 UTC

- reply

I hadn't considered AI for accessibility before... It may plug an unfortunate gap I've seen, which is that developers really, really, really don't want to support a11y in most firms and management doesn't care enough to force the issue.
But if an LLM can reduce the problem of identifying issues to one analysis pass... That's not a small thing. And I'm generally seeing LLMs are better at that than traditional heuristic approaches, because they end up trained on real-world examples of what goes wrong.

2026-06-08 19:25:20 UTC

- reply

Yep. And as a partial product of that process, to a first approximation this is fine. ;)
You want some of the best hackers you can create? Tell young people there's stuff on the Internet they aren't allowed to see. This was literally one of my initial motivations to learn HTML. 😛
(... Approximations 2 through n say **this is not fine.**)

2026-06-08 19:23:20 UTC

- reply

I'll have to check primary sources; I haven't read the specifics of the request from the PM yet.

2026-06-08 17:20:54 UTC

- reply

I believe a bloom filter (or trained recognizer) on the video layer will generally be hard to subvert; Lord knows they'll try. That gives a rough cut of "probably porn," and the whine screen can then be passed to an online detector for verification and then follow up (notify parents, ban account, whatever). It's impractical to use that to prevent the image from being seen though, unless we want the kid phone experience to be total ass.
It would also be very expensive battery wide, depending on how frequently it's scanning the whole screen. And "your phone can screenshot what you are up to at any time and send it to the government" feels like it should be pretty viscerally horrifying, even to Brits.
(I don't think this is a good problem to try to solve; my noodling is on whether it's actually unsolvable and under what constraints it becomes so).

2026-06-08 17:08:28 UTC

The UK is trying to ban nudes (store, transmit, display) on smartphones owned or operated by kids under 18.
Is young people seeing nudes bad? Separate problem, and could be its own thread, but for the sake of argument let's say "yes."
Is the problem tractable? I think the answer is actually "probably" for a partial solution. It'll be expensive (Apple and Google's money and everybody's electricity and, of course, privacy), but I'm playing Evil Oracle in my head and I'm not seeing a defeat strategy for "cliient-side hash check routed to server for detailed confirmation and revocation of account on offense."
... But blocking seeing them at all? Intractably expensive in energy and time for anything approaching a 100% case when bright and motivated young people are in the loop.

2026-05-21 17:03:15 UTC

Olive Oyl: oh hey Popeye ur back early
Popeye: goon's haunted
Olive Oyl: what?
Popeye: *loading some spinach and getting back on the boat* goon's haunted

2026-05-17 14:14:39 UTC

I stumbled over a railing yesterday working on a project. And it wouldn't have been so bad except that it was one of those homeless hostile architecture railings so instead of being, you know, a regular railing for touching and keeping your balance, it was a railing that somebody had put a bunch of pointy bits on.
As a result, I couldn't grab it to break my fall and I face planted on asphalt.
My days of thinking anti-homeless architecture is an awful and human hostile idea are certainly coming to a middle.

2026-05-12 16:54:46 UTC

That feeling when the elevator arrives that you know maintenance was working on this morning and before you step onto it you look down the gap for light from the shaft to confirm you're not about to be involved in a monumental industrial accident.

2026-05-07 16:21:54 UTC

"That was a weird dream. I dreamed Nintendo decided to remake Star Fox 64 on the switch 2 for some reason. Not a re-release, a ground-up remake."
*Opens YouTube*
"OH NO I WASN'T DREAMING I WAS BROWSING AT 3:00 A.M. LAST NIIIIIIIIIIIGHT!!!"

2026-04-28 16:51:46 UTC

- reply

Whooooof. One of the best things about leaving the company is I don't have to leave over this.

2026-04-22 22:35:01 UTC

- reply

correct. What you're bumping up against is the general theory of relativity.
Special theory of relativity deals with the behavior of different inertial frames moving at different velocities relative to each other. But then general theory adds in what happens when you accelerate and shift from one of those frames to another frame.
I can't pretend to understand all the math, but it is general relativity that both explains why it is the people in the rocket, if they turn all the way around and come back to Earth, who are younger than the people who stayed on Earth... And why we sitting down here in our gravity well experience time a tiny, tiny fraction more slowly than the GPS satellites orbiting the planet do. Because a rocket ship accelerating and the constant acceleration of standing around in a gravity field are exactly the same as far as the theory goes.

2026-04-16 17:36:23 UTC

- reply

Do we know that?
TBH I can see a lot of potential advantages to throwing Copilot at a code base as hairy as a Ruby on Rails app, though I have not tried.

2026-04-09 22:12:33 UTC

Against my better judgment, I shall wade into the discourse on a topic I love.
I've seen enough people claim that modern Star Trek would be doing better if they didn't make gay characters the focus. And I think that argument is a big head-fake. Rather, I think the people making it have cottoned to what *is* happening in those shows, and it's much, much scarier for them than what they're describing.
I can name two gay couples, one bisexual character, and one almost-certainly-omnisexual (or maybe just, like, really hot, like, is it hot in here? Can I sit down? I should sit down) character in *Starfleet Academy.* I can also name one gay couple, two(?) nonbinary characters, and one gay character in *Star Trek: Discovery.*
I cannot name a single plot that was *about* any of that.
To be honest, I think *that's* what offended the offendable the most: it's *not* a focus. This era of Star Trek embraces and extends the Roddenberry-era notion that in the future, love would be more free, growing an assumption Roddenberry's era didn't necessarily embrace that love need not constrain across gender. If people of the same gender found mutual attraction, they just chased it. If opposing gender, they just chased it. What probably rankles the most for those opposed to that kind of world is that in this Trek, *it's background.* Boyfriends share quarters and bathroom sinks and nobody says anything about it. A character asks for their pronouns clarified and everyone just updates their mental rolodex. Jay-den has some feelings he hasn't quite placed yet for a War College cadet and he is willing to explore those feelings with the guy and *nobody stops him.*
And *that's* what's scary to those opposed to love being that free: this is not a Star Trek centering that conflict. This is a Star Trek depicting a future where their attitude on gender and sexual expression *already lost.* It's on the ash-can alongside capitalism, race, and in-species tribalism as "things humanity grew out of fighting about."
And, yeah, I can see how that kind of future would be terrifying for a certain kind of person.
And I'm sorry for them.
#StarTrek #StarfleetAcademy #Discovery #NuTrek

2026-04-01 15:55:25 UTC

I just saw "improving your emacs skills or learning new features of the libraries" referred to as "emaxxing" and I just can't.
#emacs

2026-03-19 18:25:52 UTC

- reply

I mean... It's pretty awful. An arbitrary 24-hour time-gate does nothing but make me wonder if you can work around it by turning off the network and manually setting the date forward...

2026-03-19 17:51:16 UTC

One of the more interesting aspects of the Ukraine war was the revelation of how badly compromised Russia's defense assets were. It was, honestly, a little startling to the public to learn that the feared Former Soviet State still had size on its side, but not a modern military; their tech was either out-dated or had been pillaged by corruption so badly that it couldn't be deployed as intelligence analysts had assumed it could be.
It is extremely fair to argue that Russia's greatest state-defense asset was perception and that the war in Ukraine damaged that and, in so doing, materially threatened the country's safety---that if they had simply never started a war, everyone would still perceive them as unassailable and incredibly dangerous to engage in combat and nobody would even *think* to try stochastic attacks, asymmetric drone warfare, or any other modern tactics under the *assumption* that such a grand superpower had a solution for all of that.
In short, all they had to do to keep everyone's perception of their strength was literally not start a war to test it.
Just a thing I'm thinking of right now for some reason.
https://www.cnn.com/2026/03/18/politics/us-ford-carrier-fire-iran-war

2026-02-24 16:31:15 UTC

- reply

I don't have a lot of patience for third-party voters (they fundamentally misunderstand the math of a system with hundreds of millions of inputs).
... but to be fair to them, it was a pretty awful move for the parties to both put up people well past retirement age and go "*Here's* who should be running the country."
Unfair to us the voters and unfair to the candidates; that job shaves years off your life. Obama didn't have grey hair before he was President!

2026-02-17 18:12:52 UTC

@npub1fdr…lvhs I love when paths of desire end up on the walls instead of the ground.
https://mastodon.fixermark.com/system/media_attachments/files/116/087/317/064/944/636/original/c7eae2b02d1d79f3.png

2026-02-16 23:45:17 UTC

- reply

My goodness. It's always very, very weird to watch geological-time processes happen in one human lifetime. Things like that really feel (pardon the choice of words) bedrock-permanent until they just... Aren't.

2026-02-11 16:03:01 UTC

- reply

I like this one specifically because the Cloudflare gate is there to address the problem of "Too many visitors."

2026-01-30 16:07:56 UTC

- reply

Turns out AI is more an existential threat to traditional news than I had considered.
Ultimately, if the news professional cannot tell AI editing from an authentic article... What use do we, the audience, have for the professional? Because I can get doctored images of what's going on in Minneapolis from my uncle for free.

2026-01-26 20:05:36 UTC

- reply

> How do I vote to put better candidates on the ballot in the first place?
The primaries. Disquietingly few people *still,* after everything that has happened, actually show up to vote in the primaries, still.

2025-12-23 17:33:40 UTC

- reply

On the one hand: people *really* need to understand that Google isn't authoritative on much of anything, and a Google search is no substitute for a background check, *especially* if you're an authority figure.
On the other hand: I feel like Google AI straight-up claiming "This person is a child predator" and they're not should *really* not be protected by Section 230. it *feels* like if the AI bot isn't "editorial voice," *nothing* is.

2025-12-12 18:30:22 UTC

- reply

I gotta say, I got problems in my life, and I'm not trying to solve nearly enough of them with a million random digits with 100,000 normal deviates.

2025-12-12 18:22:40 UTC

For some reason, it makes my heart sing in a very weird way that RAND *still* provides *One Million Random Numbers with 100,000 Normal Deviates.* Online now!
https://www.rand.org/pubs/monograph_reports/MR1418.html
Do you have a problem that you need to solve? Have you tried one million random numbers with 100,000 normal deviates? You really can't go wrong with one million random numbers with 100,000 normal deviates.

2025-12-05 18:57:56 UTC

In the *Starfleet Technical Manual*, there is a throwaway detail of the LCARS user interface that when it is upgraded, users can still utilize the previous several versions. This is because their work is mission-critical and it is unacceptable to compromise the mission by forcing unfamiliar changes; eventual retraining is part of their job, but the flexibility allows them to take the retraining at their own pace.
I think about this from time-to-time and how this is *still* a sci-fi idea that is seen all too rarely in how we do actual computer UIs.
#StarTrek #LCARS #UI

2025-11-12 20:38:33 UTC

- reply

Of course, the list of things he has in common with Nazis is longer than one might like.
... having strong opinions on the race makeup of the UK, a country he does not live in. For example.

2025-11-01 16:16:26 UTC

- reply

https://mastodon.fixermark.com/system/media_attachments/files/115/475/333/805/575/839/original/7557b4b95ae9b090.png

2025-10-29 21:33:35 UTC

- reply

I didn't notice, but your comment made me Google it (LOL) and... This is now a rolling thing, apparently? I'm seeing they did this to other sections of the company 9 months ago and in June.
... oh man, if I'd just held on six years they could have decided they didn't need me instead of me deciding I didn't need them.
(lol).

2025-10-28 20:46:51 UTC

- reply

I think you're right. I still think it's the right direction to go because when a human kills a child we can't<li>pull their logs</li><li>simulate the event</li><li>understand precisely the sequence of decisions made by the person that resulted in a death</li><li>push an OTA update to every single driver on the road so it never happens again</li>
... but people like to *feel* far more in control than they de-facto *are*, and I think the feeling of "robot killed my kid" and "stone-cold sober driver who had a single moment of inattention killed my kid" (much less "drunk driver" or "driver having a heart attack") is viscerally different in a way that is likely counter-productive to maximizing vehicle safety.

2025-10-27 15:42:01 UTC

- reply

Man talking a lot of shit for someone referred to the House Ethics Panel.

2025-10-22 17:31:58 UTC

- reply

I really wish someone would call her bluff and just run that story. "Our reporter's mom chooses summit location for Trum-Putin summit."
Interview the mom, let the whole thing hang in the air.

2025-10-22 03:20:21 UTC

Gotta love watching old 1940s commercials.
This woman just spraying RAID directly on her tomato plants to kill the bugs.

2025-10-21 22:11:17 UTC

- reply

Cool. That has already fallen out of my head but I'm glad it's in yours. 😉
Seriously though: bash is easier to use when I remember that Perl was trying, among other things, to be a better shell-style scripting language. So syntax like "we loop on a read command, the command you think of as getting keyboard input but is also just the command to dice up stdin" is entirely sensible from that era and not weird nonsense that mixes statement and expression context, because those contexts aren't precisely a *thing* in Bash. And I ignore that the approach is entirely different from the approach of looping over space-separated values in a variable, which is how I'd accomplish this task if it was a one-argument list of values instead of a two-argument list of values.
... but alternatively, if I do it in emacs I can gut-check the incremental steps of stream modification as I go and that feels way safer. I'm glad shell is there if I want a terse way to describe how to repeat the process though!

2025-10-21 21:08:42 UTC

"Oh man, I have to run this shell script on every pair of values in this CSV file. I know how to loop over one argument, but how do you loop over two in Bash?"
**The regular-brained way**<li>Look it up</li><li>Write a bash script to do it</li><li>Test it on a small sample of the file because you definitely messed it up on the first try</li><li>Run on the rest of the file</li>
**The emacs-brained way**<li>Copy file to a buffer</li><li>Split out the comma-separated values into space-separated values</li><li>Prefix every line with <code>script-name</code></li><li>Save as a Bash script and mark executable</li>
I'm sure there's a very lovely syntax in Bash for iterating over two-argument lists and I won't be learning it today.
#emacs

2025-10-13 16:28:54 UTC

You see that show about the time traveling hare who hops into other hares' bodies throughout history to set some wrong right? *Quantum Lepus,* I think it was called.

2025-10-09 20:55:16 UTC

- reply

Look, it's well-known that brain is stored in the dick.
That's just science. ;)

2025-10-09 15:24:51 UTC

- reply

I have been told that kind of thing should be treated with suspicion because it originates as a 4chan meme.
... but I think it's also broken quarantine and mutated so wildly on the open Internet that it's probably quite divorced from its original intended meaning.

2025-10-08 19:10:07 UTC

Of everything Mariner ever said, "I am trapped inside of an egg, no, leave me alone I don't want to hatch" is probably most relatable.
#StarTrek #LowerDecks

2025-09-26 21:07:19 UTC

- reply

We do build highways. :)
You chose an interesting example for personal reasons because I grew up west of Richmond Virginia. So let me tell you about the beltway. Tuck in; you've activated my trap card and I'd rather spin yarns than finish this performance review. 😉
Richmond had a beltway on the books for decades that would circle around the city to the east and to the west. They built out the east one with relatively low political turmoil, if I recall correctly; east of Richmond is mostly swamp and an airport.
To the west is Midlothian and Chesterfield, the places that white flight went *to* back in the Sixties and Seventies. A group of, let's say, entrepreneurs took a gamble: they bought a bunch of land in the 'redzone' where the highway was slated to be built for cheap (because it was slated to have a highway built through it), built upscale suburban housing developments (I will stop just short of calling them "mini-mansions," but they were delightfully-apportioned three-story works abutting a golf course), and sold them to the folk fleeing school integration.
By the time the state got the money and political will together to execute on the highway, the builders had cashed out, taken the money, and run (fine people; some of them were the parents of my classmates in school!), and the neighborhood was sitting there, smack dab in the middle of a "PLANNED FOR EMINENT DOMAIN" sector.
... but here's the thing about planning to eminent domain a whole-ass neighborhood of some of the wealthiest taxpayers and campaign contributors in one of the wealthier central-VA counties. 😉
The whole thing threatened to turn into a decades-long legal quagmire. These folk were willing to support any politician who would scuttle the highway project, and if none were, they were willing to pay for lawyers to fight the issue forever.
You want to know what ended the fight?
J.B. Watkins was the owner of the Watkins Tree Farm just east of those housing developments. Watkins is an Old Virginia Family Name. For those fair readers who don't know why I capitalize those words: Virginia has a tradition of noblesse oblige dating back to its founding by second-born mid-tier nobility who were never going to get land or title in their homeland, so their options were join the clergy, die nobly in war, or go to the New World. Yes, it's a paper thing and yes, it didn't stop or even mostly fight slavery... But it still matters. The Watkins Family Is Compelled To Serve. Like the Lees. Or the Fairfaxes. Or the Hopkinses.
Watkins went to the General Assembly and put his entire tree farm on the chopping block if they could re-route the highway through it. And he could do that because he was one guy who owned a lot of land.
Sometimes we build highways through single philanthropy too. 😉
(... I'm not actually disagreeing with your example, I just think it's funny that of all the examples we could have discussed, you stumbled on the one where the story I most know about it is "that time we built a highway because one dude let it happen" 😄 )

2025-09-26 20:01:55 UTC

- reply

This may be the cynic in me talking, but I've had enough trouble getting a group of six people to decide where to eat that I'm highly skeptical of the masses endowing libraries.
I'm not saying it *couldn't* happen. I am saying there is a certain simplicity to the process of a captain of industry just deciding to devote a huge chunk of their fortune to philanthropy that you don't get out of a committee (for good or ill).
(And for the record, I'm still very much on team if-you-make-a-billion-dollars-your-marginal-tax-rate-should-go-to-100%, while at the same time not having *deep* faith that the money the government would get their hands on would go to a library and not, say, police drones).

2025-09-26 19:52:48 UTC

- reply

Actually no, I haven't. I wouldn't be *deeply* surprised, but that one's new to me.
The Homestead Strike is a story I'm familiar with, but I don't remember a "burning them alive" part of that story.

2025-09-26 19:16:37 UTC

- reply

I live in a town that is up to its eyeballs in libraries and museums named after the like four rich families who were here in the 1800s and 1900s. Libraries and museums that survived the subsequent economic collapse of the region because their endowments put them outside the regular need to turn a profit that imploded most of the small and medium sized businesses in the area between the '60s and the '90s.
If you're looking for someone who wasn't *born* a mark, you'll have to look elsewhere. 😉
(Lotta people justifiably against robber barons.
Not a lot of those people endowing libraries with all that idealism though.)

2025-09-25 21:03:50 UTC

- reply

I went down to the library a couple weeks back because it's right next to the barber shop and the barber blew through my haircut so fast I had a rare spare, unplanned hour.
You know what happened?
1) A conversation behind my head where someone was astonished you could take the books home and the card was free (seriously, if we tried to invent libraries today they'd make them illegal)
2) A local author had her book reveal and signing in the front of the library. I just wandered up and met a neighbor for the first time!
Damn, libraries are fucking sweet.

2025-09-25 18:56:09 UTC

- reply

I think there's a *lot* to this.
I've been at a place in my life where legacy starts to matter a lot. And I look at a lot of the techbros and the ways they're unspooling and I say to myself "These are men (and yeah, mostly men) who have realized they will not live forever, no matter what. This news is hurting them more than some because they bought into Kurzweil's stories at a young age and thought they actually *might...* And the need to conceive of a world without them is *terrifying* them into doing *really stupid shit.*"
Bill Gates, for all his faults, seems to have made the transition into the second stage of his life by dedicating himself to eradicating disease, so at least he has that going for him. Elon Musk is 54 years old; Gates had started the Bill and Melinda Gates foundation a *decade* before where he is now, and they didn't have as much wealth as he does.

2025-09-25 17:50:02 UTC

- reply

As an (I guess by modern standards) old techbro, this shit from these weirdos is disappointing.
https://mastodon.fixermark.com/system/media_attachments/files/115/266/191/199/701/875/original/6c317f668618f5ca.jpg

2025-09-25 17:28:15 UTC

- reply

I can't help but think what this sort of thing looked like in the past, when it really *was* a meeting of a bunch of unelected kings, half of whom had inbred conditions and half of whom (not the other half, overlapping set) had syphilis.
Like... It's sobering to think that this modern arrangement is *better* than what came before, even if it's not good enough.

2025-09-17 15:45:07 UTC

- reply

I'm not sure I want to be in the business of having to stack-rank people who built a concentration camp and people who authorized Americans to torture other human beings.
Like... Worst Olympics ever, yeah?

2025-09-12 15:46:39 UTC

- reply

Normally I'd kind of ignore that sort of reference, but given the current administration... Weird to be talking about Valhalla, yeah? Kirk wasn't a warrior, he did not die in glorious battle... So what does Patel think Valhalla is, exactly?

2025-09-11 20:13:53 UTC

- reply

True. I'm not sure yet whether it's noteworthy because "maybe the people who can access that data don't want to find him" or because "it turns out it's way, way easy to dodge all that stuff: leave your phone at home before you go commit an assassination today."

2025-09-10 14:47:31 UTC

- reply

My stock knowledge is tiny, but I'm beginning to think one can perhaps draw conclusions based on what *isn't* growing and *should*.
I'm watching the stocks on renewable energy stall out and backslide in a situation where I'd expect they should be continuing to demonstrate value, since they're grounded in technology that is working and under current deployment. That feels like signal that wherever the money *is* is a hype bubble if the money *isn't* where progress is actually happening right now.

2025-09-09 15:43:18 UTC

- reply

Yep. It's just straight-up smokescreen to even debate it.
It ultimately doesn't matter if it's Trump's signature because what he *isn't* claiming is that the note isn't *from him.* Nobody cares if he auto-penned it or Xerox'd it on there.
It was from him. It wasn't from anyone else.

2025-08-27 20:17:05 UTC

- reply

Uhmmmm..... 😞
Sadly, I think the record shows Facebook somewhat directly helped kill *several* children.
https://www.pbs.org/newshour/world/amnesty-report-finds-facebook-amplified-hate-ahead-of-rohingya-massacre-in-myanmar
(ETA: Ah, I see your response on the other subthread; sorry to repeat signal. Sometimes Mastodon's nature causes race conditions.
As opposed to Facebook's nature which causes I'LL SEE MYSELF OUT! 😉 )

2025-08-27 16:40:50 UTC

- reply

okay, so *slightly* more impressive. I guess. ;)

2025-08-27 15:48:02 UTC

- reply

Definitely. Popping stack to the current tragic story: if *Rockstar* was liable for putting a sex mini-game on a disc that couldn't be accessed in the game they wrote at all because there were no codepaths to it... OpenAI should probably be liable for their machine providing novel directions to end one's life and a justification for doing so.
(Incidentally, I caught a bit of *Forbidden Planet* recently and I had to smile at the part where the Great Machine was apparently awe-inspiring because it was 20 miles on a side. "Oh, neat, so a few miles past a Large Hadron Collider." 😉 )

2025-08-27 14:28:44 UTC

- reply

There is a sci-fi game called *Lancer* that includes the concept of "NHPs," "non-human-persons." These act like AI, but they are really something far larger and more dangerous: they're intelligences existing in a wider-than-three-dimensional cosmos that can be coaxed into existing in our reality by constraining their perception to match human expectations (my favorite description by a player: "you build a box that does enough hard math fast enough and a ghost shows up to haunt it"). This setting has a concept of "shackling," which is the tech that holds these NHPs in human-shaped subjectivity; damage to an NHP's core or intentional meddling can weaken or damage the shackles, and the results are disastrous (in the setting's case, physical manifestations of disaster, since these things are basically Lovecraftian monsters that someone painted a smiley face on).
... it's a poor analogy, but it comes to mind often as I read things like OpenAI saying, in essence, "we trained our storytelling machine on *all human stories,* but we know people don't want us to tell *every possible story* because, yeah, some of those stories are damn depressing and disquieting if not actual cognitohazards. But don't worry! We *added a layer* to remind it to *not tell the bad stories,* so it's all good, okay?"
All the state of the art of porn filters slapped atop *the entire Internet...* only it's attempts to guardrail people against suicide enhancers, triggers, and instructions.
... Jesus. There is a nonzero possibility that the Internet was a mistake.

2025-08-22 15:15:22 UTC

- reply

Not *everything* else, but GPUs are basically big hyper-parallel processing units. Here's a partial list of the kind of workloads they're better than CPUs for:<li>physical simulation</li><li>image and video processing</li><li>cryptography</li><li>data analytics</li><li>medical imaging analysis</li><li>nuclear reactor simulation</li>
These aren't datacenters that are fungible to, say, storing cat photos (although they *are* fungible to finding cats in every photo if someone wanted to do that for some reason). But I don't think it's likely that they'll be junked, because the world isn't likely to run out of simulation problems any time soon, even when we tire of simulating sticking words together in a believable sequence.

2025-08-22 14:30:45 UTC

- reply

This I haven't heard. Did they put dedicated machinery in there that can *only* be used for LLM training?
Generally, the same math used to train LLMs can be used to solve other problems; I wouldn't be surprised if part of their computation of cost / benefit here was "Even if LLMs don't pan out, oh no, we filled a room with fast linear algebra solvers, guess we'll just have to... Model nuclear physics for the next generation of reactors or fold proteins or something."

2025-08-21 15:19:29 UTC

- reply

It's another manifestation of Onebox results being hilariously off sometimes, but worse because the AI speaks with an authoritative tone and (much as I'd like to pretend it doesn't) that matters.
All of these features suffer the same issue: they're just the "I'm Feeling Lucky" button in a fancy dress. They're replacing getting data from multiple sources and making the human do the brainwork to correlate it and synthesize an answer with letting the algorithm try its hand at it.
(Now that I've typed that, I think that's probably the best lens to think about these features through. "Do you trust Google AI results? Well, would you trust the 'I'm feeling lucky' button gave you the right answer every time with only one press?")

2025-08-21 15:13:26 UTC

- reply

... the funny thing is, I know enough about how the search core is built to make an educated guess at how this happened.
Search core farms the query out to (possibly dozens of) completely independent software engines to answer the query. Those engines have their own backing datastores and they don't synchronize to each other; it's necessary they don't to make search fast. So it's entirely possible that the Top Stories core is pulling data that is more recent while the AI Overview core has built its information off of an older search result cache (and possibly hot-cached this answer before the NYT confirmed the death when the news broke locally / on social media). Google, as an institution, leans on the tradeoff "eventual consistency is good enough," so this kind of bug is seen as a predictable side-effect of the chosen tradeoffs.
Does that make us feel better?
No, nor should it. *Get your shit together, Google.*
#Google #search #AI #fail

2025-08-21 15:10:27 UTC

*sigh* Get your shit together, Google.
#google #ai #fail
https://mastodon.fixermark.com/system/media_attachments/files/115/067/385/542/986/340/original/d0469aa9c9cf31f6.png

2025-08-19 21:01:32 UTC

- reply

"he will not give up UNTIL you are [expletive] [expletive]" really feels like it could go right at the top of your résumé, honestly.

2025-08-19 18:41:22 UTC

Guy at a party in 1925: "You know what's the hottest thing? When a girl wears a long skirt and like there's a draft from below and it blows the skirt up and you can see her legs because the skirt isn't covering them."
Guy's friends: "Yeah whatever Billy, nobody cares about your fetish."

2025-08-13 18:05:38 UTC

- reply

and you're s thousand percent right. It's deeply baked into Google's DNA and self image that they can be impartial vendors of information, and they are deeply wrong about that in principle.
You can't satisfy the mission statement and be impartial. "Information" is a value judgement.

2025-08-13 16:57:01 UTC

- reply

It's been a feature request for ages. I strongly suspect we're seeing it land now because it was legitimately that hard to implement it atop the existing search infrastructure.
You're right about the risk of filter bubbles, but the door swings both ways; this is the feature people have wanted to be able to say "I want to look up this JavaScript command, *and I want no fucking links from w3schools thankyouverymuch*" since w3schools was called expertsexchange.
(In terms of the filter bubble risk: oh yeah, that's a good question. *Sure would be cool if Google had some kind of 'ombudsman' position to weigh in on these kinds of questions, wouldn't it?*)

2025-08-12 17:44:17 UTC

- reply

I feel the same way. More than any other company, I definitely felt like Google understood that if its privacy story breaks down, the party ends.
Where they tended to fall on their face was a patriarchal assumption that their privacy model was always the best one ("What do you *mean* it's a problem populating your Buzz friend list from your most-often-contacted Gmail addresses?"), but at least there *was* a model and it was taken seriously; people who broke policy actually got fired.

2025-08-12 17:36:36 UTC

- reply

As much as breaking up Google is *probably* *some kind* of good idea... Yeah, this is the issue, isn't it?
We don't trust Google with our data... *And we trust everyone else much, much less.*

2025-08-12 15:11:19 UTC

Interesting culture collision I've observed in my employer:<li>Automotive manufacturing (at least around here) has a "checklist culture": you have a process, so you make a checklist, so you tick off every box in the checklist after you've done it.</li><li>They added a checklist for the manual steps of code verification (the stuff the linter and typechecker can't automate) to the PR template</li><li>But programmers hate toil</li><li>So every programmer I know has some variant of a self-written macro that does something like "I paste in the checkbox list text and it returns me the text with all the boxes checked.</li>
This of course defeats the purpose of the checklist, but what did anyone expect?
(This is one place I deviate: FIRST Robotics taught me not to automate a checklist, so I actually still tick all the boxes. But I can't in good conscience pretend I'm always meditating on the meaning of each box rosary-style as I check it...)

2025-08-04 16:51:38 UTC

- reply

You're not wrong about the mind-altering gases. ;)
More specifically / less tongue-in-cheek: I think that's where I keep getting hung up. Mass-crawl to populate a search engine? Generally fine, just don't crawl so hard you push my machine over. Mass-crawl to populate an answer machine? Get out of here. This is where I see the zeitgeist converging (including the fascinating apparent contradictions such as Stable Diffusion, which works by taking the identification engine *that artists who knew about it were generally fine with because it made it easier to figure out if a random image online was probably theirs* and turned that same dataset into a tool for generating novel images by reversing the nature of the question).
... so yeah, at least for many people, I have to conclude they're putting things online for fame / reference / I-want-it-to-be-seen-here reasons (even in the absence of remuneration reasons). And that's a blind spot for me because I... don't? I put things online because they're not online, and they're easier to find there. My primary reader of my blog is *me,* years from now when I'm trying to set up something similar and can't find my notes; I'm basically using Google as a free indexer for my own data. So whether their Ai tells me my own notes mixed with other people's (as long as the information is correct) or links me to my page alongside other people's isn't relevant to me; they're equivalent. I'd still be populating the blog because otherwise the AI doesn't have the answers I'll need.
There's a larger conversation to be had here. In the sci-fi community, I've gotten in arguments with people over whether even in Fully Automated Luxury Gay Space Communism, people would still do things for fame purposes. And fame / respect / recognition has its own capitalism-style antipatterns because attention is a finite resource; you can only cram so many citations on a work before it's more citation than content.
Like I said topside, none of this is "We should all think like me;" if anything, it's me debugging my thought process in public, and I appreciate your patience with me.

2025-08-04 15:54:57 UTC

- reply

I would prefer citation but, if I'm being honest with myself, I accept answers from Stack Overflow, Reddit, and other places all the time and trust the wisdom of the crowds (and Google's PageRank) to surface the likely-to-be-right stuff. Almost nobody on e.g. StackOverflow references a primary source for their suggestions; they just slap down some code that is "it works because it works, try it and see" and it gets upvoted if it does.
I don't think I see the moral issue (I'm probably missing experience or a proper ethical framework to hang it on) in feeding people un-cited info in a buyer-beware fashion. There may be a *utilitarian* issue; if there's no way for AI to close the feedback loop on whether answers are actually correct, it's a useless tool for the purpose of acquiring answers. But that's not a moral issue, that's an issue in the same bucket as "Google beat out Yahoo because it was using better signal to figure out if it was delivering search results people wanted and Yahoo was floundering with trying to categorize data and squash keyword-optimizers." Yahoo wasn't *evil* for trying a worse signal-acquisition algorithm and failing to deliver the goods.
(All this to say, I think I'd almost certainly rank an AI Q&A tool that answered with "Here's my answer and here's my sources so you can check my work" higher than one that just spits out a string and leaves it up to me to figure it out. I'd much rather use the first than the second!).

2025-08-04 15:31:54 UTC

(Personal observation, not a "way things should be" thought):
Surprised to learn of myself that of all the reasons AI crawling of the web should be concerning, "It stops redirecting traffic to the host sites" isn't one of them. To me, anyway.
The "old web" that I hear people sometimes pine for was a place that people were putting stuff online with no expectation of financial remuneration. People put stuff online *so it was online.* The idea that it could end up re-indexed somewhere else was... Yes please we'd like that info easier to find actually?
There are other issues with AI answers (personally: I want the back-link to the source because I shouldn't trust random ideas spewed at me by an AI agent any more than I trust random ideas from an anonymous Redditor, and pedigree and source citation helps there). To the extent that they divorce site maintainers from their ad revenue... That's not an AI problem, that's a "There are no stable models for capturing revenue off of infinitely-copyable information in capitalism" problem. There's not as much moral daylight between AI scrapers bypassing robots.txt and IP blockers and lying about their user-agent... and Aaron Swartz smuggling a server into a secure network to scrape a paid academic archive... as I think some people want to believe.
#ai #copyright #capitalism

2025-07-20 02:16:38 UTC

- reply

I learned that recently! That the reason they're 128.2 is because 128 was the first way we identified class B networks before masks became a standard and 2 because they were the second one!

2025-07-19 20:20:18 UTC

- reply

Carnegie Mellon, sorry. Old university campus. The big main open views ("the cut") are still there, but a lot of the new architecture is broken up by planter boxes and small trees and feels a lot more natural, especially given the overall urban environment.

2025-07-19 15:34:58 UTC

- reply

The old campus is a lot greener than it was when I was there. They've embraced integrated garden architecture. Fewer broad straight empty paths leading from one giant building to another giant building.

2025-07-19 15:28:14 UTC

The future looks different then it thought it would.
Sometimes, in good ways.

2025-07-18 20:49:14 UTC

Hearing people noting that Stephen Colbert getting the axe for criticizing the current administration echoes the Smothers Brothers, and I think that's true, but I think it's true in some very interesting ways.
Among the things the Brothers put on the show in the end (so sayeth Wikipedia) was Joan Baez paying tribute to her (then) husband David Harris who was going to jail for refusing military service. It is, and I believe was at the time, a crime to counsel people to refuse the draft. So CBS putting that piece on the air could have been interpreted as such an act, which the whole network would be liable for.
... and history remembers that instead they chose to can one of the most successful variety shows of all time. I'm not sure quite what the lesson is here (other than the obvious "this is what it looks like to be on the wrong side of history" one). Probably a lot of lessons crammed into a little space.

2025-07-18 00:57:29 UTC

Programming for emacs is interesting because the way you isolate state is very buffer-oriented.
I keep falling into the trap of "Well, I need to have a listener wake up later to handle the results of running this process once it's done. So I'll create a closure that~" no. The process has a process buffer. That buffer has local variables. That's your execution state. You need to know something later? Cram it in a buffer-local variable. Register a sentinel function that will get the process and the sentinel event and have that function get everything it needs from the process's buffer contents and local variables.
It's a lot like programming older-style event-driven JavaScript (before we became afraid of the JS event model and wrapped it behind three frameworks and a local database). Or like programming a videogame with agents and behaviors.
(Every time I sit down to write something big, I have to relearn this because all other programming I do does not work this way).
#emacs

2025-07-17 15:31:04 UTC

- reply

Authentication and online identity are too fundamentally tied to email to make this a practical suggestion for most users.
If I switch my email providers right now, I literally cannot pay my mortgage.

2025-07-16 14:38:00 UTC

- reply

Interesting. My understanding is she wasn't convicted on anything requiring "Epstein files" to exist. So this is basically blackmail in public, right? She's claiming she has knowledge of the files without claiming she has knowledge of the files and seeing if the President will blink?

2025-07-12 15:14:55 UTC

- reply

Is this Buc-ees?

2025-07-11 20:11:00 UTC

- reply

I mean, I don't know what to say, other than "As one of the people who's full-time job was to expose all that compute power to regular human beings to do whatever they wanted to do with it, agree to disagree." Without cloud compute, we'd be... Doing the same thing, worse and more expensively. Probably chewing up about as many resources too, possibly more (you'd be surprised how much electricity an idling workstation under a desk eats just keeping a screen on that nobody's looking at).

2025-07-11 19:33:16 UTC

- reply

Real hard for me to swallow the notion that the world would be better if Google never existed. Like, I'm old enough to remember both the era before search engines and the era before Google.
It wasn't better. In a lot of ways.
I'm curious how far you want to expand that notion. Is it Google specifically, or search companies in general, or software engineering companies in general, or the Internet... Where did we go wrong in your eye?

2025-07-11 18:07:25 UTC

- reply

Oh yeah, and then it got bigger still!
I think I commented to someone the other day that the actual protection G has against source code exfiltration is that what the culprit would have is... Source code that runs on their infrastructure. It's like making off with DNA without the cellular machinery to synthesize from it... Good luck re-implementing Chubby, Borg, BigTable, datacenters with PCRs, and the rest from scratch, nerds! You thought you were getting a technical boost and what you got yourself is *a box of hell!* 😉

2025-07-11 17:52:31 UTC

- reply

Remember in *The Hitchhiker's Guide* how the Golgafrincham sent their "useless" third of the population away, like telephone sanitizers, and then everyone on the planet died of a disease contracted from a dirty telephone?
I thought Douglas Adams' story was absurd when I was younger.
The older I get, the more levels I perceive him to have been writing on.

2025-07-11 15:27:26 UTC

- reply

In that case I agree, but in the same sense that I'm turning in the work of Eugen Rochko while writing this post using the software he created and maintains.

2025-07-11 15:12:45 UTC

- reply

I don't know this is a conversation I particularly want to have in a public channel, given the overall temperature of the room, but that's not how I would describe most of the use of it I've seen in practical application. This may just be different experiences.

2025-07-09 17:25:59 UTC

- reply

As a side-note: there's a bit more interesting history to the specific number. 435 had been the French standard prior, and British piano tuners had adopted a 439 standard in 1899. Sir James Swinburne, an electrical engineer, advocated for 440 because 439 was prime; 440 would be easier to divide and make the math simpler, which mattered to him because he was working on the burgeoning field of electronic synthesis.

2025-07-09 14:57:10 UTC

Did you know there's an ANSI standard that the A note above middle C is 440Hz?
And that the reason for that standard was to curtail "pitch creep?" This is the phenomenon where over time, orchestras would slowly drift pitch upwards. Tuning *slightly* above the note makes everything sound brighter, and audiences like that, so it kept happening. It happened so much that some sheet music from the 1600s has to be transposed to be properly played as it was originally intended to sound.
I don't know why I'm thinking about that today.
... on a completely unrelated note, Jira changed all the colors for status buttons! *Summer festive!*
#atlassian
https://mastodon.fixermark.com/system/media_attachments/files/114/823/852/548/272/420/original/1142581c1159440d.png
https://mastodon.fixermark.com/system/media_attachments/files/114/823/853/726/915/032/original/577de35365d86422.png

2025-07-09 00:52:53 UTC

When you're ripping your collection of *Firefly* DVDs to get them off the shelf to make room (and make it easier to watch them without flipping discs).
#firefly
https://mastodon.fixermark.com/system/media_attachments/files/114/820/534/471/773/434/original/0ebdeb02551691db.png

2025-07-06 19:10:00 UTC

Character idea: a killbot with its murder motivation extracted but its tactical strategic attack algorithms intact, along with the ones intended to hunt targets and use psychological warfare.
They are a highly successful psychologist.
"You mentioned feeling inadequate at work. Based on our past discussion profile, were I trying to mentally disarm you to inflict maximum harm, I would say 'Your parents never loved you and you know it.'" If you are all right with it, I would like to help you unpack why my protocols are identifying that as a torment vector and see whether that relates to your professional feelings."

2025-07-05 14:34:35 UTC

- reply

Guy's got a hell of a mural on the side of a building in Philadelphia.

2025-07-03 23:06:21 UTC

Mildly impressed at how Google can still throw a wrench in my whole-ass day even though I'm not there anymore.
FAANG's got a long reach.
#Google