Context Engineering in 2026 — Louis-François Bouchard, Omar Solano & Samridhi Vaid, Towards AI
Agent failures in 2026 are primarily caused by context rot, not model quality degradation
“the context is filling up and it's getting worse and worse. The results are getting worse because of it”
Towards AI founders present practical context engineering techniques—compaction, memory retrieval, and context management—drawn from experiments on their open-source AI tutor. The core thesis is that agent degradation is typically a context problem, not a model capability problem, reframing how practitioners should debug and architect long-running AI interactions. The session includes an open-source repo and live experiment results on Hugging Face, making it a practitioner-facing engineering resource rather than a research or product announcement.