@RisingSunInt_ Its obvious and really good, quality is through the roof.
What i would die to learn is what is their release and QA process, and project and team structure, which allows them to move so fast and so well
This game has a vibe thats very thin and hard to catch, its a fantasy that is not afraid of being a fantasy, game that is not afraid of being a game,
A living world with its own rules and (very important) limits,
Its close to LitRPG books, when everyone knows that its not real, but its massive, has tons of opportunities and secrets to explore while holding onto aestetics of it
@notsky That is very true, i have so many 50-100x perf improvements stories, sounds juicy on CV, buts its only made possible by crappy code and bad decisions done before
@thegenioo Certainly! Hereβs a casual human reply:
"Itβs not just annoying β itβs a testament to the digital landscape. Let's delve in! π"
Want me to make it sound less robotic?
@thdxr Yeah, brainwashed public lol.
Argon is very nice, but only as "mental" model - dont let it write the code, let it read, analyze, brainstorm and explain things to you and it will shine like a diamond
@YouShallNotGame I am making physics based game where everything is sliceable and destructable (even creatures)
btw spider walk fully on its own ML model xD
Ah, and also made Valheim mod to play as cool naked girl with butts and breasts physics)
I swear the public hate on Argon 4 is so forced.
And I found the perfect workflow with it, it crushes real work and explorations.
TLDR:
- main chat on Argon launches a new Opus 5.5 high thread with a detailed plan
- every 25 min a timer in that thread spawns a fresh Argon agent to check direction and progress
- if the work sucks or Opus gets stuck, Argon calls it out and steers it back on track
- main chat also has a 1hr timer to check on the worker thread from the outside
Keeps agents from going off the rails early or getting stuck nitpicking dead ends (you know when it starts going back and forth on some made up bottlenecks)
"as cheap as it is today, AI companies never turn a profit"
Because that part of the statement is not true, API prices are insane margins milkers, while subscription prices vs inference are pretty fair (aka "normal" profits)
So they will be VERY fine no matter how it goes
Lmao its never that simple, Astra is the best at explaining things in humanly digestable concise language, and digging through the flows reliably. If i want to understand the complex flow, and think through solutions, this is the only way to go.
Opus finally got good at coding, i like its style, but reasoning, digging and explanations are not the strongest by far. I simply cought it lying too confidently about things too many times (in the codebases which i know well)
Fable has better trust, but its sometimes so hard to read at times, its like opus^2, it literally wants me to "fuck off and merge"
So the reality is that if you do any real work, you need all of them ideally, that can maximize your understanding and minize model babysitting over stupid decisions and crappy code