How to Create an AI-Powered English Learning Workflow: A Complete Guide

Create a custom Gemini Gem to orchestrate a six-stage learning loop that covers speaking, reading, writing, and vocabulary, using supplementary AI tools like ChatGPT, Claude, and DeepL Write for specialized tasks.

The repository byoungd/English-level-up-tips provides a production-ready framework for transforming modern AI services into a personal English tutor. Instead of switching between random apps, this workflow creates a continuous Input → Output → Error-Correction → Review cycle using specifically configured prompts and custom Gems.

The Six-Stage AI Learning Architecture

According to the source code in docs/threads/part-1/7-ai.md, the workflow divides English acquisition into six modular stages, each handled by a specific AI configuration.

1. Course Definition with Gemini Gems

Stage 1 establishes your learning parameters. In 7-ai.md (lines 58-73), the repository describes creating a custom English Coach Gem that stores your CEFR level, goals, and feedback preferences. This Gem acts as the central controller for the entire workflow, maintaining context across sessions and tracking your progress over time.

2. Speaking Practice via Gemini Live

For real-time conversation training, the workflow uses Gemini Live to conduct turn-based spoken dialogues. As implemented in 7-ai.md (lines 100-103), the AI interrupts only when sentences are hard to understand and provides concise feedback on grammar, word choice, and pronunciation every three rounds.

3. Guided Reading and Listening

When processing articles or podcast transcripts, Gemini splits materials into incremental questions. The 7-ai.md file (lines 58-60) specifies that the AI should first ask for your interpretation of the main idea before explaining key expressions, ensuring active comprehension rather than passive translation.

4. Vocabulary and Grammar Reinforcement

The repository includes specific prompts for converting any input material into spaced-repetition flashcards. According to lines 33-34 in 7-ai.md, effective prompts request high-frequency vocabulary, collocations, and sentence patterns rather than rare difficult words.

5. Writing Improvement with Canvas

For writing tasks, the workflow leverages Gemini's Canvas feature. Lines 98-100 in 7-ai.md describe a two-phase correction process: first identifying weaknesses without rewriting, then providing an enhanced version after you attempt corrections yourself.

6. Specialized Tools for Advanced Tasks

The architecture supports optional "plug-in" AIs for specific needs:

• ChatGPT Study Mode: Deep grammar explanations and exam simulation (lines 54-57 in 7-ai.md)
• Claude: Long-form reading and RAG-based knowledge bases (lines 22-24 in 7-ai.md)
• Perplexity: Source material discovery (referenced in section 3.3.1 of 7-ai.md)
• DeepL Write: Final polish for tone and conciseness (lines 101-104 in 7-ai.md)

Implementing the Four-Step Learning Loop

The workflow follows a continuous cycle described in 7-ai.md (lines 96-98):

"输入、输出、纠错、复习能串起来"

This translates to Input → Output → Error-Correction → Review → Re-Input. Each session must contain all four elements to ensure retention and growth, creating a closed loop where every mistake becomes a targeted review item.

Configuring Your Gemini English Coach Gem

To activate the workflow, create a new Gem in Gemini's Gem editor and paste the following instruction block derived from docs/threads/part-1/7-ai.md (lines 58-73):

You are my English Level‑Up Coach.
I am a Chinese speaker at B1‑B2 level, aiming to improve listening, speaking, reading and writing in 12 weeks.
Your tasks:
1. Interact in English unless I ask for Chinese explanations.
2. Keep each session 20‑30 min with warm‑up, input, output, correction, and recap.
3. Record frequent errors and schedule weekly reviews.
4. When I say “start today”, launch the next lesson.
5. Convert any uploaded article, résumé or transcript into lesson material.
6. After each session output:
   • New expressions learned
   • Key errors I made
   • Homework
   • Focus for the next session

This configuration creates a persistent coach that maintains context across sessions and drives the four-step loop automatically.

Ready-to-Use Prompts for Each Workflow Stage

Copy these verified prompts directly from docs/threads/part-1/7-ai.md to execute specific tasks:

Speaking Session (lines 100-103):

Please act as my speaking coach. We will have a natural English conversation for 15 minutes. Keep your turns short. Interrupt me when necessary only if my sentence is hard to understand. After every 3 rounds, give me brief feedback on grammar, word choice, and pronunciation priorities.

Work Meeting Simulation (lines 104-107):

Let's simulate a weekly sync meeting in English. You are my teammate. Ask me one question at a time about project progress, blockers, next steps, and risks. After each answer, tell me how to make it sound more natural and concise.

Article to Quiz (lines 27-30):

Create a quiz about this material. Start with 5 easy comprehension questions, then 5 harder inference questions. After each answer, do not tell me only whether it is right or wrong. Explain why.

Canvas Writing Revision (lines 98-100):

Here is my draft. Do not rewrite everything immediately. First identify the most important mistakes and weak sentences. Explain why they are weak. Then ask me to revise them myself. After I revise them, show me a stronger version for comparison.

IELTS Exam Simulation (lines 54-57):

Act as an IELTS speaking examiner. Ask me one question at a time in Part 1, Part 2, and Part 3 order. After each answer, give me short feedback on fluency, grammar, vocabulary, and how to sound more natural.

Summary

• Create a central Gem in Gemini with the full instruction block from 7-ai.md (lines 58-73) to maintain a persistent learning context across weeks
• Cycle through the four stages: Input, Output, Error-Correction, and Review in every 20-30 minute session
• Use Gemini Live for speaking practice and Canvas for writing revision to enable real-time interaction and incremental improvement
• Integrate specialized tools: ChatGPT for exam prep, Claude for long-form reading, Perplexity for material discovery, and DeepL Write for final polishing
• Reference line-specific prompts in docs/threads/part-1/7-ai.md to ensure you are using the exact configurations tested by the repository maintainers

Frequently Asked Questions

Do I need a paid subscription to implement this workflow?

While the workflow functions with free tiers, Gemini Advanced unlocks the Gem creation and Canvas features essential for the persistent coaching experience. ChatGPT Plus and Claude Pro enhance the specialized tasks but are optional according to the repository's implementation notes.

Can I adapt this workflow for languages other than English?

The architecture in 7-ai.md is language-agnostic, though the specific prompts and error-correction patterns are optimized for English CEFR levels. You would need to modify the instruction block in lines 58-73 to reflect your target language's grammar and phonology rules.

How do I maintain progress between sessions?

The English Coach Gem configured in 7-ai.md automatically records frequent errors and schedules weekly reviews as specified in the instruction block. For additional persistence, export chat histories or use Claude's RAG features to build a personal knowledge base of your mistakes and learned expressions.

What is the difference between using Gemini Live versus ChatGPT for speaking practice?

According to the repository's analysis in 7-ai.md, Gemini Live provides real-time interruption and pronunciation feedback during natural conversation, while ChatGPT Study Mode excels at structured exam simulation and deep grammatical post-analysis. Use Gemini Live for fluency practice and ChatGPT for test preparation.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}