@Kavya_learns Because Claude/Codex subsidizes inference cost. This means a 200 bucks subscription can get you thousands of dollars in token usage. Cursor accesses these models through API, which is hella expensive.
@claudeai@ClaudeDevs Just noticed that when using Sonnet and asking it to perform a task by delegating to a subagent using the same model and going into that subagent's chat interface, it says it's a Fable agent. Not sure if it's a UI bug, but got scared for my tokens lol.
@vmg I feel like this is one of those problems that needs people with years of experience to be solved, so I’m curious on how it was architected in this era of applying AI to solving complex engineering challenges.
@vmg Hey Vicent! Loved the article and the detail you put into it. I had a simple question: when developing a novel approach to a problem like solving the problems GitHub has, which is a huge challenge in terms of many engineering domains, how does agentic coding come to play?
@PremiumGoblin If you were a CS student/aspiring engineer today, what would you do to learn the fundamentals properly, when the easy fix is always looking at you around the corner (a quick Claude prompt)?