@theo not sure what u gain from downfalls of open weight models... Did u get paid to make this post?
Anybody that loves tech, loves open weight models. I guess u like more money than tech.
@Alex_tra_memory LLMs can't really write the individual wgts, instead they take a classical algorithm and translate it into wgts. That's what happened in MIT game experiments
@inherent_labs Super cool! @inherent_labs I have some questions from a newbie:
- why replication is non-verifiable?
- why Faraday uses codex as a tool? For what it uses and shouldn't then we compare Faraday with Codex, rather than Qwen3.6