Post

did a blog post of my Thoughts on the Actual State of Unreal in 2026, how it has come to this, etc etc. i've been an unreal user since 2008, and this is the worst the vibes have ever been. is there hope? perhaps www.joewintergreen.com/absolute-sta...

Good stuff! Some personal opinions: 1) IMO removing Blueprints isn't bad per se. There are great advantages to moving back to a pure text scripting language. But what sucks is that Verse is designed around crazy bullshit metaverse concepts which won't fit AAA gamedev or other customers (film, etc.)

Hard disagree, there are no advantages to getting rid of Blueprint. Adding a text language *as well* would be fine, lots of teams use the Angelscript branch for that

Of course there are no advantages to getting rid of it. But I'm sure that's what people also said when UE4 killed most of UE3's scripting tech. The difference is that UE4's Blueprints were arguably better, and so it turned out OK.

I don't think most people did say that. Unrealscript was very cool imo, I think it's sort of a shame Blueprint wasn't basically a visual layer on top of Unrealscript. That's where they shoulda gone instead of Verse

I suspect it's actually quite hard to make something that works simultaneously well in text and visual form. It probably requires being designed that way from scratch, and can't be "added on top". Happy to be proven wrong with examples though.

Sorry, but it isn't. I've seen multiple implementations of visual scripting that just transpile into something like Lua, and even aside from my vague references... Blueprint already gets compiled into what is a pretty bog-standard scripting language virtual machine.

Epic themselves have made a transpiler for it too! Blueprint nativization transpiled from the node-graph representation into C++. It sucked, because the method they used didn't get the perf benefits necessary to justify the maintenance cost, but it worked.

A friend of mine has also built a different transpiler for it that worked directly from the blueprint opcodes instead of the node graph, and my understanding is that worked even better. Regardless, it is not a particularly high technical hurdle.

The more likely reason might instead be that LLMs are much smarter (or less dumb, though I'd say the first one) at making code that is written text than making a flowchart that is literally an image. In no small part because text is more efficient than inventing pixels.

I forgot to mention the other piece disproving this claim: verse itself was (and maybe still is?) originally running on top of the blueprint virtual machine!