Cole Medin's workflow routes Opus 4.8 and Gemini 3.5 Flash to their strengths
Cole Medin demonstrates a multi-provider AI coding workflow that uses Gemini 3.5 Flash for UI design and Opus 4.8 for planning, integrations, and reasoning to build full-stack web applications.
Score breakdown
The workflow demonstrates a concrete, cost-aware approach to composing multiple frontier models by phase — using each model where it outperforms the other — rather than relying on a single model for the entire development pipeline.
- 01Two models are composed in the workflow: Gemini 3.5 Flash for UI design and Opus 4.8 for planning and integrations.
- 02Gemini 3.5 Flash is priced at $1.50 per million input tokens, making it cheaper for the token-heavy design phase.
- 03Opus 4.8 is used for planning, page copy, and integrations due to its stronger reasoning and lower hallucination rate.
Cole Medin describes a multi-provider AI coding workflow built in response to two models shipping within a week of each other: Gemini 3.5 Flash from Google and Opus 4.8 from Anthropic. Rather than treating them as competitors, Medin routes each phase of frontend development to the model best suited for it. Gemini 3.5 Flash handles the UI design phase, which Medin found produces interfaces that look less obviously AI-generated than those from other coding agents. Opus 4.8 handles the planning stage, page copy, and backend integrations, where its stronger reasoning capability reduces hallucinated content — a problem Medin observed when letting Gemini handle the full workflow.
Medin also references Archon as a tool that automates the full chain end-to-end.
The workflow also has a cost rationale: Gemini 3.5 Flash at $1.50 per million input tokens is significantly cheaper than Opus 4.8, making it practical to spend more tokens on the design phase without inflating overall costs. The full workflow runs on top of Claude Code and Pi (a coding agent harness) and is published as a GitHub repository at `coleam00/frontend-mix`. Medin also references Archon as a tool that automates the full chain end-to-end. The video walks through the workflow in phases: exploration and spec, strategic planning, UI design, and integrations.
Key facts
- 01Two models are composed in the workflow: Gemini 3.5 Flash for UI design and Opus 4.8 for planning and integrations.
- 02Gemini 3.5 Flash is priced at $1.50 per million input tokens, making it cheaper for the token-heavy design phase.
- 03Opus 4.8 is used for planning, page copy, and integrations due to its stronger reasoning and lower hallucination rate.
- 04Medin observed that Gemini 3.5 Flash produced good-looking UIs but hallucinated page copy when used alone.
- 05The workflow runs on top of Claude Code and Pi as the coding agent harness.
- 06The full workflow and example are published at the GitHub repo `coleam00/frontend-mix`.
- 07Archon is referenced as a tool that automates the entire workflow chain end-to-end.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 9, 2026 · 17:05 UTC. How this works →