@Moore Folks still seem intimidated so I hope the takeaway from this series is just how surmountable all of this stuff really is. The only prerequisite is that you're curious enough to try.
Agent harnesses wait for no man. Updated my article to add two new steps: a testing pass and an optional security scan. The chain continues to surprise me.
Early thoughts on Opus 4.8:
The interesting part isn’t that Anthropic made the model smarter. It's how intentional the focus has been on honesty. There's a lot less over promise and a lot more careful collaborator.
Feels like a more mature direction. This is my fave model now.
https://t.co/lnmrVWlulH ← new new new
introducing imprint: hebbia's design team across product & brand. we've been pushing relentlessly to craft a unique, forward-facing pov on design in finance & tech.
see how many easter eggs you can spot. we're having a good time over here ⛵️
@visiblemiles Amen to that. Notable shift in the right direction. I suspect the next few months hold even more significance for Opus and Mythos as dominating models in SWE.
@mylaststar Yeah exactly, it really shows. I’ve built workarounds for this issue and they themselves are now hardened as a result. I’ve gone through 4 build chains with it now and noticed a huge difference in the approach to bug catching.
@krispuckett@claudeai It always changes that's why it seems prudent to invest time and energy in tools like @cursor_ai. Multi model with strong orchestration means you never have to pick a side.
Cannot understate how big this news is. Has potential to be the killer design/IDE tool we’ve been waiting for.
After trying EVERY tool over recent months I always return to Cursor and Figma. This might be the thing that replaces Figma in that equation.