Descriptions:
Bart Slodyczka walks through the newly launched Projects feature inside Claude Co-work, Anthropic’s agentic coding environment, explaining how it extends the project concept already familiar from Claude’s chat interface into a full workspace with persistent memory, custom instructions, scheduled tasks, and local file context.
The most substantive section covers memory architecture. Slodyczka distinguishes between what he calls “hard memories” — explicit notes written to a learnings.md or memory.md file by the agent — and “soft memories,” the conversational nuance across many sessions that previously got lost. Projects adds two mechanisms to capture both: scoped memory search that queries only sessions within the active project folder, and an automated 24-hour background sweep that extracts key context and stores it as a persistent memory resource accessible to every future session.
The video closes with a practical demonstration using a fictional client called Apex Digital, showing how to write project-level instructions, attach a proposal document with phased deliverables and due dates, and structure a long-running consulting engagement inside a single project. Viewers working with Claude Code or similar agentic tools on multi-session client work will find the memory management strategies directly applicable.
📺 Source: Bart Slodyczka · Published March 22, 2026
🏷️ Format: Tutorial Demo
