Remote

Josh Simmons

@dotstdy@mastodon.social

mastodon 4.8.0-nightly.2026-10-06

living life a quarter cycle at a time

0 Followers

0 Following

50 Posts

Joined April 04, 2022

codez:

https://git.nega.tv/

website:

https://nega.tv/

email:

first name at website

Josh Simmons @dotstdy@mastodon.social

· 2mo ago

"Good tools are invisible" is something that I really don't agree with in general, but it's super funny to me when people use analogies to carpentry or construction to justify it, e.g. "builders don't have a fun hammer". Fuck off they don't! People in all fields have favourite tools, or tools they like to use even when it's not necessarily the best idea, tools which are fun or joyful. Tools they've customized into something they enjoy using.

30

2

5

0

Josh Simmons @dotstdy@mastodon.social

· 2mo ago

In the world of technology there are two separate but equally important groups. Those who believe you can solve social problems with technical changes, and those who believe you can solve social problems with new copyright license terms. These are their stories.

12

1

4

0

Josh Simmons @dotstdy@mastodon.social

· 2mo ago

After 20 years, Wayland not implementing global shortcuts finally pays off when Chrome tries to add a LinkedIn shortcut for their chatbot. Now that's forward thinking!

11

1

3

0

Josh Simmons @dotstdy@mastodon.social

· 3mo ago

Replying to

@0xabad1dea@infosec.exchange the amount of content on the internet that's generated from somebody feeding high profile posts into an llm is absurd at the moment, and it's worrying how bad people are at recognizing it.

15

0

1

0

Josh Simmons @dotstdy@mastodon.social

· 4mo ago

it seems the money hose has been turned off for gitlab, the executives dream of bringing it back via random firings and replacing people with subscription based chat bots. f to pay respects https://about.gitlab.com/blog/gitlab-act-2/ (the link is just here for completeness, you shouldn't click on it unless you feel like taking significant mental damage from CEO thought leadership this morning, there's absolutely nothing of value said)

GitLab Act 2

GitLab Act 2

A letter to our customers and our investors.

12

3

3

0

Josh Simmons @dotstdy@mastodon.social

· 1mo ago

Replying to

@dev@discuss.systems one can only assume this is the work of american trade war guerrillas

1

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

this vulnerability was i believe found by an llm (filtered by a human), then has a vibe coded root privilege exploit attached. it's mentioned as a "operator precedence bug" but I can't say I'd really categorize it that way. they messed up the write size calculation and it leads to a tremendous out-of-bounds write. you can fix it by changing precedence, but you can also fix it by changing a + to a -.

10

6

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

testing is only necessary for people who write bugs. the fools.

9

1

4

0

Josh Simmons @dotstdy@mastodon.social

· 2mo ago

Replying to

@fay59@tech.lgbt don't worry about the small issue where the firefox js engine is already 50% behind on those benchmarks compared to chrome either. im sure its fine

2

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Gotta say it's pretty embarrassing that real life people actually believe their llm has a thought process that it's regurgitating when asked to explain itself. Absolutely incredible lack of understanding.

7

4

3

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

I guess I can stop training for a sub 2h marathon now it's already been done

7

3

0

0

Josh Simmons @dotstdy@mastodon.social

· 3mo ago

Replying to

@joe@f.duriansoftware.com Entelligence = Tokens * C^2

2

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Lots of games have complex mission systems but idk why they don't just embed a web view and open a Jira account for each player. Then the player's quest progression can be properly managed in epics and tracked by production.

5

2

1

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

I really dislike microservices. Please don't decompose your application into a million tiny pieces for no reason.

5

1

2

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@danluu @whitequark there's a pretty high degree of introspection and tracking required to even make that napkin calculation, though. Most people (and I'm including myself here for the most part) don't keep track of the time they spent on things, instead preferring to guess based on how they felt during the process, and how different answers might reflect on their own ability (to both themselves and external observers). What you don't look at closely can't hurt you.

4

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 2mo ago

Replying to

@fay59@tech.lgbt on a 3 year old branch. i mean no offense to research for research' sake, but oof that's a long way from usable.

1

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 2mo ago

Replying to

@david_chisnall@infosec.exchange it's hard also just due to logistics. like the core foundation of the cloud is this idea of commodity compute, but that means in order to actually use a new sku it needs to roll out globally, or near to it. larger, more powerful individual instances would be super useful to some workloads, but cloud providers can't/don't want to give you those kinds of hardware allocations, and if they do it's regionally dependent anyway which makes it worthless.

1

2

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

And finally back to malmo! It's sunny and everything is starting to get green, so I've timed it perfectly.

3

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 3mo ago

Replying to

@joe@f.duriansoftware.com isn't that what gaben is hunting for with his deep sea submarine

1

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 3mo ago

Replying to

@lritter@mastodon.gamedev.place @dev@discuss.systems what makes you think he started with marbles :')

1

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@gfxstrand@social.treehouse.systems Yeah the caching is bonkers, I also suppose it's a lot about whether making the analysis more complex is profitable in real shaders. Like if there's no potentially aliasing stores in a shader then you could theoretically promote the bda loads to NoAlias even if they're not marked as such. But you could also just argue that whatever generates the spir-v could do this itself and save everyone the trouble. (glsl has a pass which does this, but spir-v is a bit more "you break, you buy")

2

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

In fact... It would be cool if the compiler could take an oci image *as* a sysroot...

2

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@whitequark @danluu i think the fact that people are writing pretty complex documentation to try and get their llm to work better is pretty strong evidence that they could have always written it, and just didn't care or see the value. and no degree of "it's easy actually" will change that.

2

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@neilhenning I got my 4:00 5k the other week, so only 10k, hm, and marathon to go!

2

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@gfxstrand@social.treehouse.systems It looks at some decorations (like read only). But it doesn't care about the implicit restrict decoration which an external resource has according to spir-v. So it behaves as if you're in OpenCL memory model mode at all times. Likewise there's a question about materializing pointers in the shader, since iirc there's not really any option in shader languages to say "this pointer doesn't alias even though I just materialized it" (other than read only) (or to say the reverse)

1

5

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@gfxstrand@social.treehouse.systems on amd it means unless you mark a bda pointer as RO, accesses through it can't be scalarized :'( (this is not a tremendous problem, it is just what triggered me to walk through the whole pipeline...)

1

3

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@gfxstrand@social.treehouse.systems Yeah you can really see the GLSL baggage in there for sure. Of course I only meant for external accesses, it just means that you *really* need to mark any external (BDA) pointers with ReadOnly or the lowering to nir will assume that your pointer can alias anything.

1

7

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@gfxstrand@social.treehouse.systems "In general, the approach is to always assume two derefs may alias and try to prove otherwise. If the proof fails, we have to assume the worst." :'( it was somewhat surprising to me how conservative nir / shader frontends are on this front, given that spir-v gives carte blanche to assume memory declarations don't alias unless they're explicitly marked with Aliased (as long as you're not in OpenCL mode anyway).

1

9

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@jon_valdes@mastodon.gamedev.place this is what they call the normalization of deviance

1

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@rygorous @nothings i suspect as well the roi on all those things is entirely incalculable at the moment anyway due to the money warping breathless excitement of management, plus the obfuscated pricing. many places don't pay the licensing for a "normal" static analyzer due to the price, and it's unclear that the compute and whatnot for these tools will end up particularly more sustainable. (plus, plenty of places skimp out on the compute required for even an efficient build turnaround...)

1

3

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@irene this year is the first time I've had noticable allergies from The Plants and I gotta say, I do not care for it

1

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Has anybody written up the smartprop system in hammer2? (Or hammer2 technically in more detail) I couldn't really find much info beyond basics in a couple of YouTube cs2 mapping guides

1

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@whitequark @danluu boring is probably the wrong word yeah, given that so many don't even attempt it in the first place.

1

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@lindsey@recurse.social I would say the way this is expressed is usually structural, people change the way they organize their data so that coarse locking can function. Like using UUIDs which cannot conflict for new object names. Splitting data along boundaries defined by frequency of change. Using systems which aggregate at a high level, rather than using conflict heavy lists. Etc

1

2

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@neilhenning like even in the most basic sense, (incorrectly) assuming that humans can only post-rationalize their actions, the *inputs* to your rationalization aren't comparable to a strictly bounded history of generated tokens. So it's pretty deeply different ya kno.

1

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@neilhenning humans are not token predictors and claiming that they are would require significant evidence. I think a lot of people are really led astray by a mix of llm marketing, and pop scifi tropes about the mind.

1

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

used the jj operation log in anger for the first time today (i typed some stuff on the wrong repo oops) and it is *so nice* to have a vcs that gets the fundamental operations correct for the modern day. snapshot the working directory on every operation, undo and redo, plus an easy to use operation log with restore.

1

2

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@joe@f.duriansoftware.com this is a fantastic idea purely because it'd be so funny watching people flip out about bloat. Bonus points for calling the project systemd-mesad

1

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 2mo ago

Replying to

@wiredfire@mstdn.social @jonikorpi@mastodon.gamedev.place @djlink@mastodon.gamedev.place they want regulation, regulatory capture of the market is one of the goals.

0

0

1

0

Josh Simmons @dotstdy@mastodon.social

· 1w ago

Replying to

@azonenberg@ioc.exchange it wasn't quite two days but i did find out rather late that OpenOffice can't render video using hw acceleration and so turns them into a desynced slideshow on my laptop.

0

1

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

I would have probably done this already except that all the Rust options for a web server library kinda suck (the async sadness infects basically all of them, and they have gigantic dependency trees).

0

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@GDCPresoReviews@mastodon.gamedev.place I guess I should also add for completeness that loading the shaders when you load the asset also works fine if you plug it into your regular asset loading / streaming flow. That's basically how it works in snowdrop and there's no stuttering. It's just when you have parts of the PSO which are bound too late for the engine to hide it that problems occur.

0

2

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@lindsey@recurse.social in games we don't typically care much about offline editing, but many systems of data require coordinated changes, and dealing with large 'transactions' without constantly having issues is more or less the same problem, with similar solutions in practice (avoid merging at all costs)

0

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@fay59@tech.lgbt hah I pretty much agree, you trade improvements in some areas for regressions in others. It's just about choosing something that floats your boat.

0

0

0

0

Josh Simmons @dotstdy@mastodon.social

· 5mo ago

Replying to

@msinilo@mastodon.gamedev.place @GDCPresoReviews@mastodon.gamedev.place yeah it's going to depend how tight the margins on your streaming are for sure, and exactly how many permutations you're trying to compile all at once.

0

0

0

0