Sam Bloomberg

xbloom.ioUnreal's World Partition: InternalsThis is a sort of stream-of-consciousness document explaining how world partition works under the hood. This is not a general overview of world partition, nor is it an explanation of how to use world...

24 Oct 2025

Wrote a long technical dive into some internals of Unreal's world partition because I've had to go quite deep into this stuff in the last couple of years. Basically none of the internals are documented, so maybe this will be helpful to someone. xbloom.io/2025/10/24/u...

youtu.beMansion blockout runthroughYouTube video by siliex

Quick (failed) playthrough of the new level blockout youtu.be/xOpQyiWdHJI

There's already a decent amount of variation run-to-run for this level, as the available entrances, enemy placements, and item placements are semi-random. Still figuring out the right balance to number of enemies in an area, but I'm getting there.

Also, the hints on the right aren't intended to stick around. They're a bit too heavy-handed, and I'd rather there be something in the level itself to hint at what entrances are available. But this is easiest to test things out for now.

youtu.beMansion blockout runthroughYouTube video by siliex

Quick (failed) playthrough of the new level blockout youtu.be/xOpQyiWdHJI

The objective of "Kill All Enemies" is placeholder - don't have others implemented yet. It makes sense for earlier small levels, but this level is a bit too big and sprawling. Planning on adding some kind of loot to steal and leave with, and a minimum value objective.

As is tradition, I must cover a new blockout in a spew of colors as part of a ritual called... setting up gameplay and audio volumes.

View full thread

The mansion has come along though! Putting a lot more effort into varied ways around the map - there's an always-unlocked door at the front that's guarded, and then various side/rear doors that are all locked with keys randomly selected to spawn and semi-random enemy placement.

And the interior is highly interconnected with multiple floors and hopefully lots of room for varied paths through the building. Haven't finished enemy placement yet - been focusing on getting the exterior right first.

As is tradition, I must cover a new blockout in a spew of colors as part of a ritual called... setting up gameplay and audio volumes.

garth marenghi's chart place

I wish it weren't necessary, but proper audio propagation is important for gameplay testing, and the systems I have for alerting enemies requires marking out rooms and outdoor spaces in a pretty airtight way. I have plans to require less of that eventually, but for now...

Sam Bloomberg reposted

nat is scared.

@natclayton.bsky.social

i think the games industry should have studios that exist

Sam Bloomberg reposted

@tef.bsky.social

web developer: oh hey you guys are back early infra: four's haunted web developer: what? infra: *loading a pistol and getting back on the computer* four's haunted

Sam Bloomberg reposted

Matthewせいじ

@matthewseiji.com

[imsim audio log voice] You nincompoops keep forgetting the key code to the storage room, so I’ve changed it to something you can’t forget: 4. That‘s right... 4! If you run out of ammo during tomorrow’s training exercise, you only have yourselves to blame! 4, got it??

Sam Bloomberg reposted

Ed Zitron

@edzitron.com

This means that the increased price of RAM and consumer electronics was a result of an artificial "supply chain crisis" created by Jensen Huang pushing NVIDIA's customers to take early delivery of GPUs. Jensen Huang is responsible for making life worse for everyone.

Ed Zitron@edzitron.com

NVIDIA seems to be channel stuffing, forcing Amazon and other customers to take early delivery. The argument that these things are "scarce" when you don't even have anywhere to plug them into is nonsense...unless NVIDIA is saying you NEED to take them now, or else. GPUs are stacked in warehouses.

Yet another blockout, trying to focus on giving more valid paths through a level (which I don't think I've focused on enough for previous blockouts) and increasing the complexity a bit to see how it plays. So... big mansion it is!

Sam Bloomberg reposted

Paul F. Tompkins

@pftompkins.bsky.social

Replying to @pftompkins.bsky.social

Now is the time to care about everyone, not “once we win” The time to care about everyone is always always always now

Basic art blockout for two new levels pretty much done. At this point I'm not really trying to finish any visuals - just get enough done to give an impression of what I want. This is a sort of beachfront infiltration level.

Butlerian Jihadist of House Dornick

And this is some canals/docks. Clearly more unfinished, but for my purposes that doesn't matter much at the moment.

Sam Bloomberg reposted

@shimminykricket.blacksky.app

Harvard just dropped a study AI productivity in software engineering and it found the average increase in productivity from using Claude code is.....zero and for exactly the reasons you'd think fion.ac/jellyfish.pdf

Just thinking about marathon a bunch recently, as the final cryo archive weekend of the season rolls is here... I really do think that in a slightly different situation the game would have been a much bigger success. Naysayers about the genre aside, I think Marathon was a victim of particularly bad

View full thread

be approached in a very different way than normal cryo - that is to say, I do think something is lost in vault breaker compared to normal cryo. It is still worth having, as more people should be able to experience cryo and accessibility is good, but I worry about future experiences being designed

purely around being the most accessible possible versions of themselves instead of what we got with the duality between regular cryo and vault breaker.

Anyway... marathon is legitimately good if you can stand extraction shooters. It's unfortunate that some of the best level design in a PvP shooter is so inaccessible, but I also worry what would have been lost if it hadn't been that way initially.

Vault breaker showed that cryo can be made accessible to a wider audience and is, I think, a really really fun mode that does a great job teaching the map and potentially provides a template for how to make more difficult PvP experiences accessible via PvE, but it is also very different and must

manages to create near constant close-quarters battles (something that's relatively uncommon on other maps) without feeling suffocating - something Outpost's pinwheel gives a taste of, but pinwheel doesn't have nearly the complexity in escape and flanking routes that cryo does once you learn the map

and the way that cryo manages to create an entire progression loop unto itself with security cards/batteries/coolant/keys/credentials. Outpost has the cards but not quite as much of the strategy - it's a good introduction to what cryo eventually asks you to do, but it doesn't have the depth.

I also find it a shame that we'll probably never get another map like cryo archive - it is, legitimately, one of the best experiences Bungie has put together, up there with some of the best raids of Destiny. Despite how frustrating it can be, and despite some tweaks I'd like to see them make

(particularly in how high-level players can get into other players' spawn areas much too fast these days), the highs of cryo are really just unmatched. The sheer amount of stuff in the map is incredible, the mystery when it initially released, the interconnected design of the map, the fact that it

Anyway... I do hope they manage to turn things around. I think they would have needed to pivot to more PvE experiences regardless (as Arc is doing the same), and Vault Breaker already showed just how good even "simple" versions of that can be great. Marathon's gameplay works amazingly well for

both PvE and PvP. But I also worry that we'll lose part of what makes Marathon special, even if the wider gaming audience doesn't see what that is. I'm not saying we will, but any major pivot in focus inevitably brings about that worry.

less accessible and therefore even if you could objectively say it is better in general (which is not an argument I'm making) you couldn't say that for a decent portion of Arc's audience as they'd simply not want to try it given the seemingly more "hardcore" tilt of the game.

I think some of the hardcore-ness is a bit overblown, but Marathon very purposefully styles itself as being in a very hostile world where everyone and everything is out to get you - it's part of the tone of the game, and not having that would lose something.

the basic game is that good (outpost + cryo archive both being *incredible* maps - easily some of the best multiplayer level design out there - helps quite a bit, too). But I think it suffered by being too close to Arc Raiders - in a world where Arc either didn't exist or even if it had been

separated from Arc's release by a longer period of time it would have done much better. Arc's massive success meant that Marathon was never going to pull people away from it without being at least (pulling numbers out of my ass) 30% better, and while I love the game the reality is that it is simply

timing more than anything. Maybe it's just the way I play these types of games nowadays but for once it feels like it's a game that launched with the right amount of content for what it is - it's held my attention for the better part of two seasons without adding all that much new stuff just because

Working on yet another blockout...

View full thread

The animation is in world space so adjacent brushes will line up their animations *if* adjacent edges share vertices, so I have to slice up brushes a bit weirdly to transition from higher detail sections to lower detail.

My previous blockout does the same, though it's not as obvious since it's to transition between brushes that don't otherwise line up as opposed to going from a high detail section to low detail.

Working on yet another blockout...

The water is subdivided like this because it gets animated in the vertex shader, so I need to have additional verticies for that animation to work with.

Finally finished a bunch of more underlying technical work, so starting on a new blockout. Wanted to make something more open-ended and with canals/waterways for this one.

youtu.beLevel 3 Art PassYouTube video by siliex

Gameplay volumes make this map truly incomprehensible.

for comparison...

Basic art pass for this level mostly done. With the new light-based stealth mechanic. youtu.be/BHfxbrXHv4s

21 Sept

And now... Thief-like light detection. Wasn't originally planning on implementing this but it solves some situations where I felt I shouldn't be visible but enemies could spot me anyway, without having to adjust level geo or add manual visibility blocking volumes.

The AI uses the luminance value to attenuate the distance at which and time it takes to detect the player (distance/time percentages in the debug window).

21 Sept

I still need to go and optimize this properly - right now the texture readback and resize is done on the CPU, ideally it'd be via a compute shader, but that can come later.

Implementation is simple - a pyramid shape is at the player's position, rendered as lighting-only with a camera + 32x32 subviewport on a separate render layer. This is then resampled to 1x1 and the luminance value (0-1) read out (Inst. Luminance in the debug window).

21 Sept

Luminance is averaged over a few seconds, which solves gameplay issues (small buffer of time to react to lighting changes), UI issues (eye icon animates smoothly while still representing exact game state), and fixes the value oscillating rapidly in some areas.

20 Sept

Starting on another quick art pass for a third level.

And a bit more...

New ranged enemy! And some other miscellaneous polish - dead bodies (which enemies can see and investigate) weapon switcher HUD elements, and some new audio implementation.