Figma

Start with Jesse Lumarie’s Figma MCP server talk for a concrete account of moving from a one-day-a-week experiment to a launch in about three months. Follow the decisions around representing design context, reusing existing code components, evaluating results, and accommodating uneven MCP client support. Lumarie describes launching locally before building a remote server; read that as the recorded launch sequence, not a statement about today’s server availability.

Start with Eyal Blum’s organizational adoption talk if your challenge is getting beyond early agent successes on small tasks. Blum describes trust breaking down on larger production problems and recommends deterministic verification, red-to-green TDD, and plans divided into reviewable phases with validation gates. His account also covers uneven adoption, experienced engineers’ quality-control burden, reduced developer agency, and verbose AI-assisted communication. Treat these as his recorded observations and recommendations. His estimated 5× speedup comes with an unreconciled timing breakdown, so it should not serve as a reliable planning benchmark.

For a broader engineering path, watch the MCP launch account followed by the coding-agent adoption account. Compare Lumarie’s automated evaluations and integration constraints with Blum’s emphasis on tests, planning, and human review. This pairing connects a specific integration project with organizational workflow questions; it does not establish that both speakers describe one shared initiative or a single company-wide policy.