It’s interesting how “tool rules” in Letta make this something of a sliding scale to my understanding; the more you constrain the agent the closer it becomes to a “workflow” and vice versa; Claude code has hooks which are much more limited but have a somewhat related purpose
September 5, 2025 at 11:34 AM
It’s interesting how “tool rules” in Letta make this something of a sliding scale to my understanding; the more you constrain the agent the closer it becomes to a “workflow” and vice versa; Claude code has hooks which are much more limited but have a somewhat related purpose
Reminds me of a funny quote I found at some point: The saying, ‘Life is just one damn thing after another,’ is a gross understatement. The damn things overlap. -The Cincinnati Enquirer
August 21, 2025 at 4:36 AM
Reminds me of a funny quote I found at some point: The saying, ‘Life is just one damn thing after another,’ is a gross understatement. The damn things overlap. -The Cincinnati Enquirer
W/ stackattack I took the approach of implementing components as simple functions that can be easily copy/pasted and extended in your own codebase (w/o needing to recreate anything), and keeping the configuration that can be done on components relatively limited
July 30, 2025 at 5:05 PM
W/ stackattack I took the approach of implementing components as simple functions that can be easily copy/pasted and extended in your own codebase (w/o needing to recreate anything), and keeping the configuration that can be done on components relatively limited
They also seem to straddle a sort of middle ground where their components try to remove boilerplate but still not be too opinionated, so I've found that putting together real infra stacks still requires a lot of code, and there isn't great documentation for their components
July 30, 2025 at 5:05 PM
They also seem to straddle a sort of middle ground where their components try to remove boilerplate but still not be too opinionated, so I've found that putting together real infra stacks still requires a lot of code, and there isn't great documentation for their components
I have used their awsx package in the past, and I don't love some of the design decisions; the biggest thing is they use Pulumi's class-based "component resources", which actually become part of the resource state and thus make it quite hard to migrate away from them without recreating things
July 30, 2025 at 5:05 PM
I have used their awsx package in the past, and I don't love some of the design decisions; the biggest thing is they use Pulumi's class-based "component resources", which actually become part of the resource state and thus make it quite hard to migrate away from them without recreating things
Do you have to learn the special key combos to enter the mathy symbols or are those ligatures?
I loved doing proofs in college, been getting enough fomo from how much fun you’ve been seeming to have w/ lean that I’m sort of itching to give it a shot
July 29, 2025 at 5:08 AM
Do you have to learn the special key combos to enter the mathy symbols or are those ligatures?
I loved doing proofs in college, been getting enough fomo from how much fun you’ve been seeming to have w/ lean that I’m sort of itching to give it a shot
The reason I made this is just because setting up infra as code for a full infra stack initially is incredibly tedious and usually takes me a lot of iterations to get right. Once you have it working, it’s pretty magical though
July 8, 2025 at 11:14 PM
The reason I made this is just because setting up infra as code for a full infra stack initially is incredibly tedious and usually takes me a lot of iterations to get right. Once you have it working, it’s pretty magical though
One of the design choices I made was to use functions for components instead of pulumi’s component resources; I really like this approach because it means that the components aren’t represented in Pulumi’s resource graph (except via naming conventions)
July 8, 2025 at 11:14 PM
One of the design choices I made was to use functions for components instead of pulumi’s component resources; I really like this approach because it means that the components aren’t represented in Pulumi’s resource graph (except via naming conventions)