Vincent Bernardi

0 Followers
11 Following
85 Posts
@marcoarment my humble company also started with rented hardware and actually stayed with it. Up now to about 100-130 servers in various DCs. Never wanted to pay the cloud tax and never had too much reasons to regret that choice.
@siracusa @atpfm I can roleplay with the best of them. “Five years from now, SpaceX will have a market cap at least 5 times larger than Apple”
@yiningkarlli I agree with what you’re saying, but I think the original argument that agents could perform better for languages that have very specific compiler errors is valid.
@yiningkarlli replace the word model by agent and I believe the argument is quite correct. An agent will often write wrong code but provided with a better compiler error will correct itself better and faster.

Back in the early Unreal Engine 4.x days, Epic created a really neat plugin called PythonScriptPlugin, which allows you to use the UObject Reflection system to script the Editor via Python. On startup, it generates a nice unreal.py stubs file for your LSP.

Neat, right?

Only problem is that it wasn't designed to handle the staggering number of reflected classes & structs, and today, the generated stubs file for a blank Unreal 5.7 project is half a million (501,714, to be exact) lines long.

@stroughtonsmith my bet is talking is going to be so much faster than typing that you’ll mostly use a microphone and not a chatbox.
@stroughtonsmith I would be extremely interested in your feedback on Spark
@softmaus @neo @gernot I’m convinced “good enough” agents runnable locally are almost here.
@stroughtonsmith I always had a problem with the Fermi argument. There is nothing mathematical about writing a product of probabilities we have absolutely no way of computing and then say “but there’s so many stars it must clearly be true”.
@stroughtonsmith @takeoneapp probably partly great disruption, because a lot of things become reliant on the buggy behavior in 30 years