Synt-E: The Protocol for Talking to AIs 🚀
Synt-E is a "language" designed to give instructions to Artificial Intelligences (LLMs) as efficiently as possible. Instead of writing long sentences, you use short, dense commands that the AI understands better, faster, and at a lower cost.
🤔 Why Does Synt-E Exist? The Problem
When we talk to an AI like ChatGPT, we use human language, which is full of words that are useless to a machine.
BEFORE (Natural Language):
"Hi, could you please write me a Python script to analyze data from a CSV file?" (Too many words, too many "tokens", risk of ambiguity)
AFTER (Synt-E):
task:code lang:python action:analyze_data format:csv(Few words, zero ambiguity, maximum efficiency)
✨ How Does the Magic W…
Synt-E: The Protocol for Talking to AIs 🚀
Synt-E is a "language" designed to give instructions to Artificial Intelligences (LLMs) as efficiently as possible. Instead of writing long sentences, you use short, dense commands that the AI understands better, faster, and at a lower cost.
🤔 Why Does Synt-E Exist? The Problem
When we talk to an AI like ChatGPT, we use human language, which is full of words that are useless to a machine.
BEFORE (Natural Language):
"Hi, could you please write me a Python script to analyze data from a CSV file?" (Too many words, too many "tokens", risk of ambiguity)
AFTER (Synt-E):
task:code lang:python action:analyze_data format:csv(Few words, zero ambiguity, maximum efficiency)
✨ How Does the Magic Work? The Logic Behind Synt-E
The secret is simple: modern AIs have been trained on almost the entire Internet, and most of the Internet is in English.
They have seen billions of patterns of code, terminal commands, configuration files, and technical texts in English. For them, technical English is not a language; it is their native language.
- Technical English is a highway: Giving a command in Synt-E is like getting on the highway. The request reaches its destination quickly and smoothly.
- Other languages are country roads: The AI understands them, but it has to "translate" and "interpret" more, wasting time and resources.
The Concrete Advantages
- 💰 Token Savings (and Money): Fewer words mean fewer "tokens" to pay for if you use a paid service. Locally, it means less load on your CPU/GPU.
- ⚡ Superior Speed: The AI doesn’t have to think about how to interpret your pleasantries. It gets straight to the point, giving you an answer faster.
- ✅ Better Answers: By eliminating ambiguity, you reduce the risk of the AI misunderstanding and giving you a wrong or incomplete answer.
💻 Try It Now on Your PC! (with Ollama)
This project includes a simple Python program that transforms your sentences in Italian (or any other language) into the Synt-E protocol, using an AI that runs free and offline on your computer.
Step 1: Prerequisites
- Python: Make sure you have it installed. If you don’t, download it from python.org.
- Ollama: Install Ollama to run AIs locally. Download it from ollama.com.
Step 2: Choose the Right Model (IMPORTANT)
Not all AI models are suitable for this task.
- "Assistant" Models (like Llama 3.1 Instruct): They are too "helpful." If you ask them to translate a request to write code, they will write the code instead of translating it. They are the least suitable.
- "Raw" or "Unfiltered" Models (like GPT-OSS or Dolphin): They are more flexible and obedient. They understand their role as a "compiler" and do not try to perform the task for you. They are the best for this script.
From your list, the winner was gpt-oss:20b.
Step 3: Install and Run
Download the model: Open the terminal and run this command.
ollama pull gpt-oss:20b
Install the library: In the project folder, run this command.
pip install ollama
Run the script: Make sure Ollama is running, then run the program.
python synt_e.py
Usage Examples
Now you can write your requests. The program will send them to your local model and return the translation in Synt-E.
Example 1: Technical Request
YOU > Write a Python script that uses Keras for sentiment analysis.
AI >
task:write_script language:python libraries:keras model:RNN dataset:movie_reviews task:sentiment_analysis
Example 2: Creative Request
YOU > Generate an image of a red dragon, in watercolor style.
AI >
task:generate_image subject:red_dragon style:watercolor
Example 3: Complex Request
YOU > Prepare a PowerPoint presentation for the quarterly meeting with the CEO on the topic of sales.
AI >
task:create_presentation format:powerpoint event:quarterly_meeting audience:ceo topic:sales
🏗️ The Future of the Project
This script is just a prototype. The complete architecture of Synt-E (which we have explored) includes:
- A hybrid engine that uses fast rules for simple commands.
- A security system to block sensitive data.
- An ecosystem with extensions for editors like VS Code.
Have fun compiling your thoughts!