VIDEO
HIGHLIGHT
Log InSign up
Gemini 3.1 Pro (Fixed with KingMode) + GLM-5: This SIMPLE TRICK makes Gemini 3.1 PRO A BEAST!
SummaryTranscriptChat

Gemini 3.1 Pro (Fixed with KingMode) + GLM-5: This SIMPLE TRICK makes Gemini 3.1 PRO A BEAST!

Gemini 3.1 Pro: A Deep Dive into King Mode

Initial Impressions of Gemini 3.1 Pro

  • The speaker previously criticized Gemini 3.1 Pro for significant regressions in performance, particularly on aentic tasks, and advised against purchasing the model.
  • Viewer feedback prompted the speaker to test the model using "King Mode," which proved to enhance its functionality.

Understanding King Mode

  • King Mode is described as a system prompt designed to improve model discipline rather than intelligence, stripping away unnecessary fluff and enforcing focus.
  • The issues with Gemini 3.1 Pro are identified as discipline-related; it has strong capabilities but struggles with execution due to excessive planning.

Performance Improvements with King Mode

  • With King Mode's ultraink trigger, the model reduces planning time from 90 seconds to about 10 seconds before starting code execution.
  • Testing showed that without King Mode, Gemini wasted tokens on redundant thinking; with it, focused planning led to better outputs.

Front-End Development Capabilities

  • When using King Mode, Gemini excels at front-end tasks by producing clean Tailwind implementations and well-organized component structures.
  • The model demonstrates an understanding of responsive design and can maintain an entire design system in memory while building components.

Practical Application Demonstration

  • An example task was given: creating a modern portfolio website using Next.js14 and Tailwind, showcasing how quickly Gemini produces high-quality components under King Mode.
  • The output includes smooth transitions and visually appealing designs that reflect good design taste due to reduced planning time.

Limitations in Back-End Development

  • Despite strengths in front-end work, Gemini struggles with complex back-end tasks such as architecture and database design due to lack of architectural intuition.

Combining Models for Optimal Results

  • GLM5 is recommended for back-end work while Gemini handles front-end development; both can be run simultaneously within Verdant's framework.

Cost Efficiency of Using Multiple Models

  • Utilizing both models allows for efficient parallel processing at minimal cost since they are either free or low-cost options.

Tips for Effective Use of Models

  • Injecting the King Mode prompt into project rules enhances performance across agents; it's especially beneficial for Gemini due to its discipline issues.
  • Detailed prompts yield better results; specificity in visual direction significantly improves output quality when working with front-end tasks.
Video description

In this video, I'll be telling you about how KingMode completely fixes Gemini 3.1 Pro's biggest problems, turning it into a genuinely powerful frontend model and showing you how to pair it with GLM 5 for a free, full-stack AI workflow. -- Key Takeaways: 🔧 Gemini 3.1 Pro has discipline problems, not intelligence problems, and KingMode fixes them. ⚡ KingMode's Ultrathink trigger cuts planning time from 90 seconds down to about 10 seconds. 🎨 Gemini 3.1 Pro with KingMode is surprisingly strong at frontend and UI work. 🏗️ GLM 5 with KingMode remains the best choice for backend and architectural tasks. 🔀 You can run both models in parallel inside Verdent using isolated Git worktrees. 💸 The entire workflow is essentially free since both models are free or dirt cheap. 👍 With the right system prompt, Gemini 3.1 Pro goes from a skip to a solid tool in your stack.