anony-123

What are the benefits of skills at all. Does it increase the capability of the LLM?

So, what I am thinking is that, the AI generated numbers or tried to find numbers "a", "b" and "c" to check if aⁿ + bⁿ = cⁿ

Can not we do it by code?

NOOOOOOOOOOO, Chinees models are better than US models. They are cheap and open source. Which is the true blessing for humans. Otherwise the US would be the dictators for AI.

OPUS 5 is piece of trash and I don't think they would want to build the Opus 5 better than Fable, because fable 5 take more tokens and have 50% limit or runs on credits.

Why this comment is being flagged?

Docs looks good, but the landing page is having gradients, pills, badges which looks vibe coded AI slop :)

Why the website is looking AI slop?

Brilliant!!!

Yeah, I am also interested in learning how LLMs works. So planning to spend some time on building a toy LLM to understand the principles.

Do you have any resource in your had which you could suggest?

Oh yeah, context is the thing. But I believe you need to know what's going-on on architectural level.

For example If I ask LLM to build an analytics system where, there is soooo much data to ingest and LLM uses postgresql for the DB but we know ClickHouse is the great fit here, given the analytics queries we need. So, then you are right LLM can definitly figure out ClickHouse or something better, If we provide full context of what we want, but getting to know the full system and feeding LLM with the right context is something where we need understanding of the tech-stack atleast in general (might not be specific to the underline language).

I don't believe LLM can ever build a system from end-to-end reliably.

Code generated with AI has no worth (i think), even if its a core piece of any system.

I think the architecture around systems is un-replaceable like what I do now, is I create specs and then asks the agents to cook exactly as the spec.

What's your AI workflow btw?

The team of AI bots :)

Is your mercor.com site works really?

So, does it mean Antigravity is better than Claude code with opus model? Given this benchmark. I once tried Antigravity and it was just disappointing.