Interesting but can you elaborate?

I’d def want to hear your thoughts on this. Can we chat sometime?

Third, Anthropic sent an email just now describing how they are gonna crack down on the platform abuse and this makes me a bigger fan of Anthropic and more glad I’m using them.

So glad I’m back to building instead of troubleshooting the past few weeks.

#JustBuilding

Second, connecting my project direct to GitHub instead of internal documents has significantly improved the quality of Claude’s output (except I still don’t use Opus) with 3.7 Sonnet and it responds more directly and specifically to my prompt instead of assuming what I what when I just said it.

…over the head with instead of actually trying to help or produce a more impactful model.

I mean, have you even tried to use the Opus model yet?

Opus is an example of applying a third layer of logic over the previous failed two layers and then charging us $100 a month for.

…extends the error to another layer of logic, processing, and your results and never address or fixes the root cause.

In the insurance world, we call this product liability and product negligence, but in the legal world it’s not defined yet, so it’s a liability Anthropic beats us…

…and that Anthropic is not doing anything to fix its or even knows how at this point.

The flaw is in the first layer of logic Claude executes, so when the Anthropic dev’s think they are fixing it by applying a second layer of logic that “corrects” the first layer, it actually just…

Love it, this helps significantly with building @prismmediapro.bsky.social right now. And it will help with every other project I’m building too!

Love to see it!