@dexhorthy Am I disqualified from being an AI fanboy if I refuse to use one agent, much less 20+ agents, and do all of my AI assisted coding in a web browser session?
@SebAaltonen@mytechceoo@albertadevs@getaxal Which is also the cost of three junior programmers, so I have to ask: How are we saving money by laying them off, only to fork over the same amount for AI? The code quality certainly hasn't improved.
@lilyofashwood@AnthropicAI I notice that it makes no mention of running this in 4.6!? So switching to the cheap brain damaged model is your only option when you say the wrong word?
@CometThief@SebAaltonen They are trained to do so, at least in the second phase of training that I like to call the Marketing phase (first phase is just data crunching), where by they're trained to be confident and plausible rather than saying "I Don't Know".
@SebAaltonen Are you sure about that? It might have changed since yesterday. Or it might be depressed and just doesn't feel like it? Have you take the renderer's feeling into account? Have some empathy!
@KatanaLarp This is an idiotic soft-stop that just tells you how many tools it's used so far. You just have to hit continue to get it to move. Now if they could just put up a button that crash stops it in the middle of a hallucination, that would be grand!
@KatanaLarp@pedma7 I've not run into the nerfing that every other Claude user seems to see. Is that an artifact from running coding agents on top of Claude during peak hours, or does it also happen during off hours?
@SebAaltonen If there is one thing that LLMs have taught me is the value of a damn good design document. 2 hours spent on a good doc will save me hours of back-n-forth down-the-rabbit-hole sessions with LLMs.