Deep Dive
Monday 23 March 2026
AI Adoption Challenges
Attendees
| Name | Background |
|---|---|
| Vikram Pawar | Host — backend engineer, tech lead, trainer |
| Shubham Jain | Community member |
| Dmitriy Disskiy | Community member |
| Aditya Nair | Community member |
| Mohan | Community member |
| Rajan Patel | Community member |
| Alagirisingam Pullaiyannan | Community member |
| Rushikesh Ramesh Gavhane | Community member |
| Martim Loureiro | Community member |
| Ben Butterworth | Developer, community member |
| Oliver Cooke | Founder, ebitlabs |
| Erika Tsai | Product manager in legal tech |
Session Summary
The first CCCL Deep Dive explored real-world AI adoption challenges, covering image generation constraints in Claude Code, external APIs for consistent image quality, secure sandboxing practices with Docker, and knowledge management tools including Obsidian.
Key Topics Discussed
1. Image Generation Constraints
- An educator working with connected Google and OpenAI APIs discussed difficulties generating flat vector illustrations and photographs in Claude Code
- Historical context: Claude Code was originally text and coding-focused, making it inherently weaker with images than models like Gemini
- The gap between coding-focused and multimodal models was explored as a practical adoption challenge
2. External APIs and Sandboxing
- External image generation APIs such as Replicate and Foul.ai were suggested for generating and refining prompts for consistent character styling
- Cost concerns with the OpenAI API noted as a practical barrier for smaller teams
- Docker images recommended for complete file system isolation in corporate environments — a key concern for enterprise AI adoption
3. Knowledge Management and Custom Skills
- Participants shared custom Claude Code skills for triaging Slack messages and automating competitor research
- Obsidian recommended for non-technical users to store and visualize interconnected knowledge using Markdown files
- Discussion on bridging the gap between technical and non-technical team members through shared tooling
Action Items
Vikram — Share the Figma Claude Code deep dive link within meeting notes
Konstantinos Papadopoulos — Research image consistency: contact friend building consistent characters and detail the process for Brenda