if you're not working with unlimited tokens like @steipete and @bcherny, you could do your loop with claude code + caveman.
event -> trigger->action -> eval -> feedback
- event: create a "wiki" to render claude generated md files as context
- trigger: click "review with claude"
New post! Optimizing Tool Selection for LLM Workflows: Differentiable Programming with @PyTorch and @DSPyOSS
Training local, learnable routers can reduce token overhead, lower costs, and bring structure back to agentic workflows. This post shows how.
(Link below)
New post! Optimizing End to End Agent Workflows with @DSPyOSS optimizers: Differentiable Programming for Learnable Graphs
How learnable graphs reduce manual tuning, scale decision-making, and make LLM workflows self-improving.
(Link below)
New post! Behavioral consistency in LLM agents isn’t a prompt problem, it’s a graph optimization problem.
Jointly optimize behavior across entire @DSPyOSS agents, from routing decisions to how each branch preserves context and communicates outcomes.
(Link below)
i've always admired what .@DeGodsNFT, .@kevindegods, and .@frankdegods have done for the NFT ecosystem, shitposts and controversies notwithstanding! I'm excited to join the community today and build together.
im the founder of @solarplex_xyz - @solana based creator platform that uses gamification to build better relationships for their audience.
we love @guac_gg. we've integrated them into our tipping flow for creators.
they use our in-post minting feature for audience engagement.
New post! Benchmarking Local vs GPT-Based Routing in LLM Workflows using @PyTorch + @DSPyOSS
We run an eval on a tool controller vs GPT-4o for classifying 1000+ support tickets. Same accuracy. Half the cost.
Code, evals, DSPy pipeline, all inside.
(link below)