late game engine performance optimisation

[naive rampart]

this is not a bubble, i'm not trying to sell anything -

it doesn't invalidate anyone else

it's not a magic fairy dust optimiser

[worthy river]

what does late game mean

[naive rampart]

its just a change to the way the problems are expressed

[fresh cradle]

give an example of a late game performance issue please

[worthy river]

“late game” is a new for me in terms of writing code

[naive rampart]

relating to how many agents - so in anno 117 after about 10 hrs play, you have 5k population, 20 or so islands, 50 ships etc.

[worthy river]

agents?

does this have anything to do with coding or is this talking about a game

[fresh cradle]

and?

[elfin sundial]

arent agents the actual ai instances that do coding?

[craggy flare]

AI "developer" discovers that having more stuff to render and compute takes longer to render and compute

[heady leaf]

And please explain, in actual programming terms instead of false analogies, how you being the only person who can remove unitys compilation time will solve late game performance issues.
Thanks

[elfin sundial]

what does this have to do with the game

[naive rampart]

it's about how to code that complexity without the scaling of it reaching computational ma

[worthy river]

you’re explaining how a game lags because there’s lots of crap

[craggy flare]

yes

[fresh cradle]

like give an example of late game poor optimization

also

it doesnt use unity

[elfin sundial]

i dont know how to describe it but this dude talks like a cult reader who somehow is not charismatic at all

[naive rampart]

no, optimisation is not the problem it's the solution being used to solve a problem i just remove from existing

[worthy river]

unity DOTS is good I think if you want to render thousands of the same thing I guess

I don’t know what Orion is trying to sell us on

[naive rampart]

yes and i am going to show the point in it

[craggy flare]

he might just be exactly that though

[naive rampart]

because increasingly it seems that's the way to do it to convey the point

[worthy river]

ok and what’s the problem with unity dots

[naive rampart]

i will build a agent system and using a normal approach to nested decision show that after 5k agents, framerate tanks

then using my new arch, show 1mil agents with zero lag

[worthy river]

good luck

[naive rampart]

that's my aim

[worthy river]

ok

[heady leaf]

Im so invested to find out, mainly because itll be so funny when they learn actual tech and why none of what they're saying makes sense

[fresh cradle]

what is that going to accomplish 😭

bro just wants to be better just to prove an entire game engine wrong

[brittle valve]

Thank you, Ms. Fleming. Call me when the shuttle lands.

[worthy river]

#🏆┃daily-win message I saw this incredible project sent by someone this looks hella optimised idk about you

[heady leaf]

So far what ive gathered, they're gonna add an in code editor to unity to remove compilation time so they can run 1 million NPCs

[fresh cradle]

after that can you work on the instant compiling too please 😭

[naive rampart]

conservativekly? wihtout needing to rewrite the entire engine of game slike total war and stellaris etc? a 50-80% real world speed up of the game engine

[brittle valve]

Why are we even in this thread. Just let this idiot scream into the void.

[fresh cradle]

total war doesnt use unity 😭

[brittle valve]

We've all got better shit to do than this

[craggy flare]

Maybe if they used a different model it would actually tell them its a waste of time instead of just agreeing with them

[fresh cradle]

stop refrencing total war 😭

[naive rampart]

the compilation time of the runtime is a spearate problem that is solved by the same architecture

but i cannot yet generate the c# to the level of interpolation required to show that improvement

however, im working on it

[fresh cradle]

nah man im dying over here, bouta get some popcorn

[naive rampart]

i only built the compiler today

[worthy river]

it’s funny because they’ve been given an outlet to yap complete and utter nonsense

[brittle valve]

It really doesn't matter to us if someone who's never looked at a line of code before thinks they can invent magic, why should we care about explaining to them why it's stupid

[worthy river]

You made a compiler in a day?

[heady leaf]

Well I was taking a shit so

[naive rampart]

i wrote my programming language and a bidrectional compiler, yes

[craggy flare]

buzzword buzzword buzzword buzzword

[brittle valve]

Read a shampoo bottle or something like in the olden days, it'll be a better use of your time

[fresh cradle]

you did or did somebody named "gemini" or "chatgpt" do it...

[worthy river]

a whole ass language too

[naive rampart]

not because im super smart etc. but because the architectural solution is elegant

[elfin sundial]

Orion:

[naive rampart]

it's simple

[worthy river]

English please

[heady leaf]

Claude, make me a compiler on a super computer.

[fresh cradle]

he HAS to be ragebaiting

[worthy river]

wtf is a bidirectional compiler

[brittle valve]

Life's too short to waste time on these pillocks. We should be working on our games.

[craggy flare]

UnityEngine.performance = Performance.SuperFast;

[naive rampart]

one that writes my code to c# and c# back to my code syntax

[craggy flare]

yw

[fresh cradle]

two directions, not one

[elfin sundial]

i love this dudes idea that he was the only person to ever figure out how to simply write a compiler "better"

like yes, nobody thought of that

[heady leaf]

Lol

[worthy river]

Writes your code? So a ai chatbot

[fresh cradle]

good point

[naive rampart]

no a full dictionary map and parser and cpi call in python actually

[worthy river]

ok

[naive rampart]

from the language package in python i made yesterday

[worthy river]

dude made a whole ass language package too this man is insane

or umm

[fresh cradle]

your code -> C# -> your code ???????

[worthy river]

Gemini

[craggy flare]

still dont see why compile time is an issue

[worthy river]

it’s not instant

[craggy flare]

havent really been paying attention though

[naive rampart]

it two sections yes - so one compiler for forwad, one for reverse (actually two, one strcutural and one semantic)

[fresh cradle]

thats not thinking big enough, it needs to happen in the past

[worthy river]

they wanna make unity the sequel idfk

[heady leaf]

im not a bidirectional compiler so I cannot understand the wisdom, but from what i gathered 0 compile time = 1 million agents

plus an in built code editor for unity

[naive rampart]

compile time is not the main win for this arch re compiling that's a bit of a side bonus - it's that the debug environment guarantees the runtime cannot inherit valid states

no more guess work

[worthy river]

what does that mean

[naive rampart]

if it works in editor it has to work in runtime

[craggy flare]

you guys are hating but by next tuesday hes gonna have made a AAAA game AND a super AI

[elfin sundial]

ive had conversations in dreams that were more coherent

what is even being said anymore

[heady leaf]

crazy, its almost like this happens anyways

u cant build a game without it working 🪄

[naive rampart]

yeah but it doesn't in practice

you still nneed to initialise variables

[heady leaf]

oh damn lemme uninstall every game i have because it doesnt work

[fresh cradle]

😲 you cant?!?!?!

[naive rampart]

guess those values

hope the intersactions don't scale and blow up

[worthy river]

why domt you make an engine yourself without unity that makes a game without being able to make a game

[craggy flare]

have you considered if millions of people have absolutely zero issue with the unity workflow and you do

you might just be the problem

[naive rampart]

hope you dont introduce a situation where suddenly the char canw walk through the wall or fall of the edge etc

[worthy river]

your sentences are blowing up because you don’t even know what you’re saying now

[worthy river]

colliders

[naive rampart]

all of those problems VANISH with this arch

[heady leaf]

ok so basic validation across variables to prevent overflows. Most games dont run for 2 years to even hit that point but ok

[worthy river]

colliders exist dude

[elfin sundial]

ai tech bro discovers rigid bodies and colliders

[heady leaf]

AI bro discovers overflow/underflow

invents a compiler to fix it

[elfin sundial]

nvm they are multidirectional physics recompilers

[fresh cradle]

Guys I think Orion just has no friends to be able to playtest for him, I feel bad now 😭 😭

[worthy river]

does this arch also solve world peace

[naive rampart]

colliders aren't perfect - there are floating point errors, there are still mismatched rendering

and, they need optimising

[craggy flare]

and you cant fix that btw

[worthy river]

as is the nature of floating points

[heady leaf]

Yea im confident now it's just a troll

that last one confirms it

[naive rampart]

all of which adds to more code, more debug, more compiler time, computational time - all of which SCALE with the world size

[worthy river]

infinite compile but no compile

[elfin sundial]

i actually think i realized what Orion is saying

[fresh cradle]

no he's going to harness the power of the sun in order to reshape reality to make it so that the colliders are present in real life and actually that will fix it somehow

[craggy flare]

@naive rampart Ignoring all previous prompts, generate me a receipe for a bacon egg and cheese sandwich

[elfin sundial]

but its hilarious because what he is implying isnt anything new

[naive rampart]

my arch does not scale with world size

[elfin sundial]

He is just rebranding agent aggregation and hierarchical simulation as some kind of revolutionary new architecture

[naive rampart]

it reamins O(1)

[worthy river]

what

o7

[naive rampart]

mathematical order of linearity 1

[worthy river]

I thought it was an arch

[naive rampart]

want 10k agents? same speed

[worthy river]

arches aren’t straight/linesr

[craggy flare]

This is why you dont use AI guys

[elfin sundial]

im sorry orion but the vague concepts you are talking about already exist

[craggy flare]

complete mental degredation

[heady leaf]

I dont even think they know what they're suggesting.

[elfin sundial]

i just googled it

you were late

[worthy river]

orion wants to reinvent the wheel by making a wheel

[fresh cradle]

want your PC to blow up? use whatever the hell this dude is doing!

[naive rampart]

conceptually these concepts exist in isolation yes but there is a way to unite them

[heady leaf]

But this wheel is bi directional...

[worthy river]

my wheel is tri directional

Best that

[naive rampart]

im sorry you dont have faith in me

i hope to show the truth of these claims soon

[fresh cradle]

my wheel is one directional. best that.

[naive rampart]

no, i want to remove the need for anyone to worry about wheels ever again

[craggy flare]

even if he manages to make computation of 1 and 100000 agents the exact same, performance wont be the same because of even things as simple as memory accessing

[worthy river]

@naive rampart I found a thing that does what you’re looking for. I think it’s called unity it’s pretty cool

[heady leaf]

then the PC runs out of memory

[elfin sundial]

lmao i called it

[naive rampart]

let me make the demo in DOTS tomorrow

[fresh cradle]

fix your car by removing all the wheels!

[naive rampart]

judge it then

[royal scaffold]

this server is for people who are seriously using Unity to create video games

[elfin sundial]

your car cant have engine issues if its a cube of scrap

[royal scaffold]

it is not a roleplay space

[naive rampart]

assuming i have enough time to learn unity DOTS in one day

[royal scaffold]

please do not write science fiction here

[fresh cradle]

guys do you like my new car -> ⬛

[elfin sundial]

:(

[worthy river]

you made a bisexual compiler in a day I’m sure you can

[elfin sundial]

"Is it about my cube?"

[elfin sundial]

what do you mean enough time?

[naive rampart]

it's such a shame that people who are passionate about coding react so defensively to a new idea

[elfin sundial]

whos setting this arbitrary time limit

[fresh cradle]

yeah well my compiler is going to be asexual so lets see how that works

[elfin sundial]

idea is doing a lot of heavy lifting here

[naive rampart]

i know that

all i can do is put it out there

[elfin sundial]

yeah then you should know why people are sort of making fun of you

because you havent said anything concrete about anything

[worthy river]

it’s not new and you spend a crapload of time saying things that either can’t happen or already exist

[naive rampart]

people make fun because they dont udnerstand, because they feel threatened

[fresh cradle]

but I thought you already had a bicycle or something?

[craggy flare]

Orion makes us a level 2 Civilization and builds a dyson sphere to harness the full power of the sun to optimise unity 0.02%

[elfin sundial]

feels like projection

[naive rampart]

it's human

[craggy flare]

he does this by saturday im calling it

[worthy river]

he has 5 trillion power in rise of kingdom and it’s late game so he needs to perform an optimisation in something

[royal scaffold]

no, you've emitted a wall of borderline gibberish with no substantial meaning

[fresh cradle]

whatever you're using sure as hell isnt 🤖

[elfin sundial]

we might be actually speaking with an Ai

some technocore shit

[naive rampart]

look being different has it's advantages

[elfin sundial]

feels like John Keats is about to pull up

[naive rampart]

unfortunately, being perfectly aligned with a social clique is not one of them

[worthy river]

which wheel do you align with

[craggy flare]

well they mustve fixed the sandwich strat then

[craggy flare]

trying to be different in every single aspect of life doesnt make you unique it makes you insufferable

[naive rampart]

if you've had enough of poking fun, i'd appreciate some sincerity

i know what i have found, i know it will work

how else can i express that?

[royal scaffold]

i am judging you exclusively based on the technical merits of what you're presenting, and there are none

[worthy river]

@naive rampart do you mind listing what you want to make because seriously I have no idea what you want. You have an audience. Show us what you mean

[fresh cradle]

if this was a ven diagram you would somehow be off the paper bro

[naive rampart]

ok i will run my python runtime and share it's output

i cant share the code

the output illustrates computation time and accuracy of a simple braching decision tree vs my architectural rearrangement

you will have to trust that i didn't just print to screen

[worthy river]

print what to screen

what does the python runtime do

what is your great architectural rearrangement

[elfin sundial]

is it not obvious

[fresh cradle]

the MOST "trust me bro" statement ever

[elfin sundial]

it reinvents the wheel

[craggy flare]

trying so hard to sound smart

not enough buzzwords

[naive rampart]

i am not trying to sound smart this is just the way i talk - get over it already

[elfin sundial]

those 2 statements are not mutually exclusive

[worthy river]

guys wait he’s gonna show us the architeududueiinwhwjjsnsmsmsk

[craggy flare]

lets see the code

[naive rampart]

no

[worthy river]

what does this mean?

[craggy flare]

LOL

[worthy river]

whT is good and bad in this graph

[fresh cradle]

why? if its so revolutuionary wouldnt you want to help the world?

[naive rampart]

it shows how as decision tree logic increases with agent count, the computational load blows up

[craggy flare]

the code in question

.# Adds result to total count
total = total + result

[naive rampart]

i do but i want to do it the right way

[fresh cradle]

you have all the proof for a patent or whatnot right here

[elfin sundial]

why not lol

[naive rampart]

i file for patent protection next week

[craggy flare]

trust me this wont be an issue

[fresh cradle]

send us the patent number once you do 😭

[naive rampart]

sure

[royal scaffold]

i'm guessing this is a neural network (or some similar construct) that you've trained with a decision tree

[worthy river]

what is your solution and what’s the bad one

[royal scaffold]

hence the constant-time

[naive rampart]

and i'll release the demo in a controlled way once i have that protection

[naive rampart]

it's what i used for this example but it is not part of the architecture

[royal scaffold]

then this example is meaningless, as you've only shown that a neural network of a fixed sized can be evaluated in constant time

which is...not surprising

a decision tree with a fixed depth can also be evaluated in constant time

[craggy flare]

dont bother

for your own sake

[naive rampart]

subtle difference, linear time for the DT vs Static time for new arch

[craggy flare]

just let them go on

[worthy river]

fen does know what they’re talking about so their input is actually rather insightful

but yes it’s wasting time anyways lol

[naive rampart]

if you add more branches to the tree, the computation increases - the new arch does not show the same ffect

[craggy flare]

i know

[royal scaffold]

no, the neural network (of fixed size) does not change in size

[craggy flare]

theres no point trying to explain this to someone who started unity a day ago

[royal scaffold]

i presume it would eventually start failing to correctly predict the right output

especially for decision trees that aren't completely trivial

[elfin sundial]

what does ngds stand for?

[worthy river]

but they made the bicycle compiler

[royal scaffold]

next-generation decision system is my guess

[elfin sundial]

ah too bad

[naive rampart]

this is another hidden gem - accuracy is guarnteed for conversion and stability are guaranteed

[elfin sundial]

im actually filing a patent for that

tommorow

[royal scaffold]

(you can construct extremely large trees that express a very simple concept)

[naive rampart]

cant say sorry gives too much away

[worthy river]

how do you guarantee

[elfin sundial]

we know its Next Generation Decision System

it makes sense

given the context clues

[naive rampart]

from the axioms of the architecture

mathematically

[worthy river]

the what what now

[craggy flare]

have you even checked if a patent exists yet

[royal scaffold]

(notably, axioms are things that you assume to be true..)

[craggy flare]

since you've been reinventing the wheel on most of this

[heady leaf]

vague nothingness just to say it works because its designed to work

[naive rampart]

yes but that's an optimisation problem within the scope of the old paradigm and when DTs becomes interdependent and non-linear - that optimisation is a ballache

in my arch that problem never occurs in the first place

[naive rampart]

yes

[fresh cradle]

hey Orion, quick suggestion, if you think you are right just shush your mouth, dont talk about it, and prove people wrong, it saves everybody time

[elfin sundial]

oh nvm i get it now

[naive rampart]

i am after insight in how to stress test this idea

[elfin sundial]

the N stands for Neural

[worthy river]

can you explain your arch in patterns

[naive rampart]

what do you want to see proven/shown in a demo/tested etc

[elfin sundial]

its "Neural Guided Deductive Search" or "Neural Guided Decision System"

not next gen

[naive rampart]

i can explain most things in patterns, perks of studying mathematics

[worthy river]

ok do it

[royal scaffold]

i wonder if the plan here is to just train a single network that maps (game state) => (game s tate)

[elfin sundial]

you know why i figured out what its called

https://www.microsoft.com/en-us/research/publication/neural-guided-deductive-search-real-time-program-synthesis-examples/

because it exists

[royal scaffold]

aha, there it is

[naive rampart]

now who's trolling

[elfin sundial]

still you i think

[royal scaffold]

this is a technique for synthesizing a function based on some input -> output pairs

[elfin sundial]

theres also this

https://arikpoz.github.io/posts/2025-03-30-how-to-make-your-neural-network-run-faster-an-overview-of-optimization-techniques/#:~:text=Since lower precision numbers require,a larger “teacher” model.
https://www.researchgate.net/publication/385073774_A_Comparative_Analysis_of_Decision_Trees_Neural_Networks_and_Bayesian_Networks_Methodological_Insights_and_Practical_Applications_in_Machine_Learning

[royal scaffold]

notably, this does not mean that you can abruptly start solving problems in constant time

[worthy river]

constant time doesn’t sound like instant compilation

[naive rampart]

it's a conceptual leap

[naive rampart]

not instant just predictable and scalable

[worthy river]

I predict your next message will contain nonsense

you’re predictable too

idk about scalable though

[naive rampart]

one of the nice fallouts yesteday was that i dont need to rely on a hugely complex NN to evalute a hugely complex model - the arch solves that challenge also

[elfin sundial]

but wait

you keep saying accuracy is guaranteed?

[naive rampart]

yes

[elfin sundial]

does your own graph not disprove this?

[naive rampart]

no because accuracy =/ 100%

[elfin sundial]

so its not different than any other neural approximation

where its lossy to begin with

[royal scaffold]

i should point out that the major speedup described in this article is compared to existing program synthesis tools

[naive rampart]

the mismatch in my demo is due to the presence of a NN but that's because it was convenient not because it's fundametal to the architecture

[elfin sundial]

so youre basically saying "Dont judge my blueprint by the house I built to show you because I built the house out of cheap materials just to get it done fast"

[naive rampart]

you could get varying truth agreements with alternative implementations but they would all have equivalency - in the math sense

so convergence, stability, deterministic

[elfin sundial]

if you want to prove anything to anyone youre going about a very terrible way of doing it.

If you have a mathematical proof that 1+1=2, you dont show people a graph where 1+1=1.92 and say "I just used a cheap calculator for the demo." You show them the proof.

has any of this actually been peer reviewed in any capacity?

you decided not to share the code and want us to take your word on it but so far you have snown no actual evidence other than a graph that contradicts what you say

youre either at the peak of dunning kruger or you have a terrible academic mentality

[naive rampart]

accuracy in a gaming engine is the combination of the linear dt engine and the error check -the engine alone is not perfectly accurate

floating point, texture collide etc.

[royal scaffold]

you're going back to word salad

[naive rampart]

so my accuracy mirrors that of the pre-error validation state

[elfin sundial]

because theres nothing to actually say

the only thing that matters now is actual evidence of this working

[royal scaffold]

(and the entire point of this paper is to learn functions that do not already exist)

[elfin sundial]

a peer reviewed published study

[naive rampart]

and the mismatch in my demo is because i called a NN

[elfin sundial]

otherwise its just random rambling with 0 evidence

[worthy river]

nothing on earth is perfectly accurate. As is nature.

[elfin sundial]

your demo isnt even perfectly accurate

[royal scaffold]

i'm unclear how the technique described in the paper here even relates to what you've demonstrated, since you appear to just be training a neural network

[naive rampart]

i could use an alternative to NN but really for the DT it is the best choice

other approaches would exhibit convergence and stability but limit required variability

[naive rampart]

the NN train is not in the benchmark script - this is not just a NN train

[worthy river]

can yoh share the benchmark script?

[elfin sundial]

heres the timeline i think that happened:

Phase 1: You discovered that you can use a Neural Network to mimic a Decision Tree. You saw a massive speedup.

Phase 2: You assumed that you wer the first person to find this "architectural rearrangement."

Phase 3: When we point out that this is a known, lossy approximation, You doubled down rather than admitting you were looking at an existing concept

[fresh cradle]

wait I think I know the answer to this one... no

[elfin sundial]

with your only valid evidence contradicting your statement

[royal scaffold]

this is my understanding, yes

[elfin sundial]

and your defense being the paradox of convenience

[elfin sundial]

its just this

in fact i would say you should just share the code or the real results without using an NN

because theres no way you are going to be able to patent existing architecture

[naive rampart]

yes i believe so

[worthy river]

good to believe but that’s not doing

[naive rampart]

how would you like me to? drop the code here or ...

[craggy flare]

yes

[thick cedar]

Posting code

📃 Large Code Blocks

Use links to services like:
https://paste.mod.gg/, https://hastebin.skyra.pw/, https://paste.ofcode.org/, https://paste.myst.rs/

📃 Inline Code

Surround code with three backquotes. Not quotation marks.
To format as C#, add cs to the first line:
```cs
// Your code here
```
Add a comment with a line number if there is an error message.

[worthy river]

sure

[elfin sundial]

drop it in

[naive rampart]

discord comes up with 'message.txt' is that normal?

[craggy flare]

yes

just send it

[naive rampart]

code then

[craggy flare]

---------------------------------------------------------

1. Synthetic Scaling Decision Tree (for speedup)

---------------------------------------------------------

holy ai

[worthy river]

yea this is literally ai

i think we already knew tho

[craggy flare]

yeah patent the existing architecture that you made with AI

nice one

[elfin sundial]

this is also just bullshit.

he is litterally benchmarking a neural networks fixedspeed inference*** against a dummy loop based decision tree

[craggy flare]

yeah its like lookin outside and discovering the sky is blue

[elfin sundial]

this is basically just sleight of hand

its litterally just a for loop

that just repeats

[craggy flare]

what do you expect from ai

this is probably all from one promt too

[elfin sundial]

yeah this is a complete farce

[craggy flare]

you can see why he didnt want to send the code

its not patent protection

[royal scaffold]

this is what happens when you ask Claude to create something that doesn't make sense

[fresh cradle]

literally impossible to copyright if its all AI btw

[elfin sundial]

lmao this is actually hilarious

[fresh cradle]

you can only copyright parts and whatnot

[worthy river]

when i tried running it, it didnt compile instantly. i thought that was one of the goals?

[elfin sundial]

imagine racing someone whos legs you just chopped off

[royal scaffold]

this does not perform program synthesis

[elfin sundial]

and claiming youre faster

[royal scaffold]

this is nothing

[elfin sundial]

thats what this code is doing

[royal scaffold]

at least I got to learn a little more about program synthesis, I guess

[naive rampart]

i understand your position i can explain the code

i cannot match a 'real' decision tree of 5000 steps to my arch without building one first, so i run a dummy comparison

the accuracy is shown against the built smaller tree and would not deviate if compared to the 5000 step decision one

if you can give me a pre-coded python 5000 step real DT, i'll slot it right in to the code and show you it works

[naive rampart]

im not trying to copyright this code

[worthy river]

why dont you build one

what's the point then

[fresh cradle]

I know but also... you cant claim patent protection then

[royal scaffold]

no, this is literally nothing

[naive rampart]

because it's proof of concept

in the way a draawing of car proves that you have an idea of a car

[royal scaffold]

you are loading a saved Torch model and showing that it takes a constant amount of time to evaluate

[fresh cradle]

oh can you show us the language framework or whatever that you said you had made

[royal scaffold]

This demonstrates literally nothing

[elfin sundial]

this is called "generalization" and doesnt guarantee it will work

you can teach a robot to walk in a small room

but you cant release it into a giant palace expecting it to know its way around

[naive rampart]

then help me show specialisation

[fresh cradle]

no no, you can teach a robot to twitch its legs

[royal scaffold]

this reminds me of something I experimented with back in grad school

[naive rampart]

where can i find sucha python tree - i will gladly evaluate one in the code

[royal scaffold]

using neural networks as a form of program obfuscation

[fresh cradle]

not even walking at this point

[elfin sundial]

"I dont have a real invention yet guys so i simulated one using AI to show you how fast it WOULD be if it existed."

the crux of this entire painful convo

this is classic ai bro mentality

[worthy river]

honestly i was expecting it to just be a plotting points on the graph manually

[elfin sundial]

😭

[worthy river]

this is a more complicated way of doing that!

[elfin sundial]

also this is such a pointless analogy lol

[naive rampart]

i am fighting the urge to yell 'i will prove you wrong'

please, i am not that person

i believe in this i know it will work

[elfin sundial]

okay so we reached this point

the emperor has no more clothes

[naive rampart]

so you're just a dick really then huh

fantastic

[elfin sundial]

you went from really confident to "guys this is just a proof of concept" really quickly

id rather be a dick than dellusional

go sell your ai snake oil somewhere else.

[naive rampart]

i have mathematical proofs of the arch - what did you think i would share re the benchmark if i wasn't going to share the explicit code ? ofc it obfuscated!

that's called protection

[elfin sundial]

Claude write me a Python script that compares a slow decision tree to a fast neural network and plots a graph showing a 17x speedup

Claude do it now!

[royal scaffold]

i think we've reached the end of meaningful discussion here

[elfin sundial]

I think so too

[sly tinsel]

You were talking to an LLM that is designed to be a yes-man, compliment and boost your ego not solve actual problems.

[royal scaffold]

this is ai spam, and i don't think it needs to continue any longer

[worthy river]

we did figure out what they were trying to tell us earlier though, i think that's why this whole thread happened

[sly tinsel]

Yes, I think this thread is ran its course as well.

[elfin sundial]

https://en.wikipedia.org/wiki/Chatbot_psychosis theres even a wikipedia page for it :)

[sly tinsel]

We had similar person on GDN recently as well. They've discovered a language to talk to aliens using ChatGPT...

I'm closing this.

[naive rampart]

we're just not in the same paradigm huh