Harnesses
3 videos across 2 channels
In AI, harnesses are the meta-tools and orchestrations that build, test, and deploy agent-powered systems, acting as the glue between language models and real-world tasks. The latest discussions explore diagnosing regressions in harnessed setups, launching open development harnesses, and the promise of meta-systems like Poetic that automate problem-solving layers on top of base models for faster, cheaper outcomes. Together, they highlight a shift from model-centric tweaks to scalable, code-driven orchestration that can empower startups to deploy capable agents with less retraining.
Did Claude really get dumber again?
The video argues that Claude based systems are showing measurable regressions across models, APIs, and harnesses, and di
It's finally here.
The video previews T3 Code, an open source desktop and web tool for building highly concurrent agent powered apps, outli
The Powerful Alternative To Fine-Tuning
Ian discusses the rapid evolution of AI, introducing Poetic and its recursive self-improving meta-system that builds hig