The team held a working session focused on getting everyone up to speed with Claude Code and AI-assisted development workflows. Much of the discussion centered on debugging AI skill behavior (Claude creating email drafts unexpectedly) and the importance of precise prompt engineering. The group discussed Anthropic's compute constraints and degraded model performance tied to the upcoming Mythos release, with optimism it will normalize. Key strategic decisions included establishing weekly scrum-style sync meetings, setting up a shared GitHub repo for the team, separating client accounts from development accounts, and building an automated task distribution system using email and Claude. The session closed with a commitment to hands-on "homework" — each member exploring personal projects to build practical AI development skills.
mindmap
root((April 19 Session))
AI Skill Debugging
Email draft bug
Prompt specificity fix
Effort level degradation
Mythos compute demands
Revenue growth outpacing infra
Team Workflow Setup
GitHub repo creation
Shared canvas for flow systems
SDLC structure in GitHub
Account separation
Client accounts isolated
Dev accounts per person
Weekly scrum cadence
Sync and workshop hybrid
Task Automation Vision
Email to info@systems
Claude parses action items
Auto-divvy to team members
Kanban board population
Trello or similar
Reverse daily digest
Poll shared inbox
Send calendar invites
Learning and Enablement
Hands-on homework
Personal project exploration
Build real bots
Resource sharing
Claude docs links
Copy markdown into Claude
Long-running personal bots
Phone-controlled endpoints
Always-on Mac as bot host
Context Transfer
Multi-device session concern
Skill to export session context
MD file summarization
Team subscription research
Max plan sharing limitations