Update #2: Back in the game
Spent all night wrestling auth loops and redirects in ClawCognition.
Switched models until Grok Code Fast 1 crushed the toughest bugs and got me back in.
Heads down again — still building, still pushing.Here’s the current “Your Agents” view in beta: clean dashboard, 7 built-in lenses, 5 config files, ready to design cognitive architectures in under 5 minutes.
More coming soon.
Thank you for the support and votes.Won’t let you down.
https://x.com/pablothethinker/status/2022263964764164428?s=20
46 views
Replies
Auth loops are brutal. Been there too many times - that moment when you realize you're in a redirect chain and can't tell which layer is breaking. Switching models mid-debug is underrated. Sometimes you just need fresh eyes on the same code, different models catch different things. Curious about the 7 lenses - are those predefined reasoning patterns or can you customize them? Building cognitive architecture tooling feels like the next step after basic prompting.
Oh nice, .md files for the lenses is a smart call - easy to version control and tweak without touching code. That is basically how I handle prompt configs in my own projects too, just markdown files you can iterate on fast. Looking forward to seeing how the foundation layer shapes up, the customization angle is what makes cognitive frameworks actually usable vs just a demo.