In today's world driven by AI, Quick Advancement It is not just buzzword – it's an important skill. This combination of arts and science are beyond the simple questions, allows you to change mysterious ideas directly, an active AI.
Whether you use ChatGpt 4o, Google Gemini 2.5 Flash, or Claude Sonnet 4, four basic principles open the full power of these powerful models. Well, and change all communication to the gate to special consequences.
Here are essential pillars of practical engineering:
1. Clear Master commands and clear
The basis of high-quality content of AI, including the code, relies on unscrupless guidelines. Tell A AI specifically what you want practice including How You want you to be presented.
Of chatgpt and Google Gemini:
Use a powerful action of verbs: Start your promotion with specific instructions like “Write,” “Create,” “Create,” “Change,” “
Specify the Exit Format: Name the required building (eg, “Give the code as the Python work,” “Release Jon ARRAY,” “Use a list of steps”).
Describe the size and length: Show clearly if you need “brief text,” “one activity,” or “code code.”
Example Quickly: “Write the Python work named counting
For Claude:
Use persistent makers to clarity: Put your main instruction within different tags such as … or three quotes (“” … “” “”). This part is helping Claude clean and focused on the basic work.
Use the tongue language: Focus on what you want AI to accomplish, rather than doing that do not want to do it.
Consider 'the System “: Before your main question, start a person or more law (eg a python engineer that focuses on clean, readable code. “).
Example Quickly: “” “Generate JavaScript activity to resolve the cord. The work should be named refstersring
2. Give the perfect context
AI models need valid advertising information to understand your application nuances and prevent mistreatment, setting their answers to your situation.
Of chatgpt and Google Gemini:
Enter background information: Describe the status or purpose of the code (eg
Describe a variable / data data: If your code must meet specific data, clearly define its format (eg.
Say what depends / libraries (if known): “Use the API privacy library.”
Example Quickly: “I have a CSV file named for a Productucts.Csv with a column ',' price ', and' quantity ', write the pych text to read the company (price * size).
For Claude:
Clause of Clearly Unclaimed: Use different components or demolition to present background information (eg '
Set Up Man: As noted, create a specific class of the Claude where it is on (eg.
Example Quickly:
3. Use visual examples (fewer shots)
Examples are the powerful teaching tools of llms, especially when showing desirable patterns or complex converts difficult to specify it in descriptive language.
For all llms (Chatgipt, Gemini, Claude):
Show inputs and exit expectations: For work, he clearly shows its intended characteristic by including something and their proper consequences.
Give examples of formatting: If you need a specific style of results (eg, a HMSON direct structure, enter the formal sample.
“Few shot” stimulates: Add 1-3 modeling boots to install and their output. This guides AI to the basic logic.
An instance immediately (of any llm): “Write the Python work that converts temperatures from Celsius to Fahrenheit. Here is an example:
Input: Celsius_to_fahrenheit (0)
Output: 32.0
Input: Celsius_to_fahrenheit (25)
Output: 77.0 “
4. Accept the test and checkpoint
It is rare to happen immediately in the first attempt. Expect to reimburse and ITerate based on the first AI's first response to the relevant results.
Of chatgpt and Google Gemini:
Provide error correcting error messages: If the code produced do not work, you attach a direct error message back to the conversation and ask AI to disagree or define the problem.
Describe an unexpected result: If the code is valid but produces an incorrect or wrong result, explaining clearly what you have seen and what you expected.
Request alternatives: Soon with such questions as “can you show me another way to do this?” Or “Can you increase this code at speed?”
For Claude:
Specify and add new issues: If the result is increasing or missing specific information, introduces new instructions (eg please make sure the code is treating kindly. “)
Analyze a person: If the tone or style of content produced is incorrect, change the original plan or add some instructions like “Accept a short code to install short codes.
Break up with complex tasks: If Claude is facing a large, much, more, easy steps, manageable, and ask the code for each step individually.
Through well spending the subtles and understanding of different llms, you can turn your AI into the active code assistant, to submit your projects and increase your problem skills.
Max is an Ai MarkteachPost critic, based on Licon Valley, who diligently develop technical future. He teaches Bide Robatovsne, fighting spam with a compulseeMememail, and put AI daily interpreting the complexity of the technology in finding clear, understandable
