Protocol selection

LLM_ADVISOR

Choose the right AI coding assistant. Manual decision guide for newsroom deployments.

LAUNCH_INTERACTIVE_MODE

01 // TASK_ANALYSIS

OBJECTIVE PARAMETERS

Data analysis

Cleaning spreadsheets, merging datasets, statistical analysis.

OPTIMAL_TOOL
Python
Claude 4.5 Opus / Gemini 3.0 Pro

Web scraping

Extracting data from websites, automating downloads.

OPTIMAL_TOOL
Python (BS4)
Claude 4.5 Sonnet

Interactive viz

Charts, maps, interactive graphics for stories.

OPTIMAL_TOOL
JavaScript
Claude 4.5 Sonnet

File automation

Batch renaming, format conversion, organization.

OPTIMAL_TOOL
Bash / Python
Gemini 3.0 Flash

Database work

SQL queries, data extraction, reporting.

OPTIMAL_TOOL
SQL
Claude 4.5 Sonnet

02 // COMPLEXITY_LEVEL

PROCESSING LOAD
LEVEL_1

Simple

Single script, one task.

USE: Claude.ai Free
LEVEL_2

Medium

Multiple files, some logic.

USE: Claude 4.5 Sonnet
LEVEL_3

Complex

Multi-step, debugging intensive.

USE: Claude 4.5 Opus

03 // RESOURCE_ALLOCATION

BUDGETARY CONSTRAINTS
Tier Option Cost Use Case
FREE Claude.ai $0 Learning, simple scripts
FREE ChatGPT $0 Quick questions
PAID Claude Pro $20/mo Heavy daily use
PAID GitHub Copilot $10/mo Real-time coding

04 // DECISION_MATRIX

LOGIC GATES
Best coding/writing Claude 4.5 Opus
Front-end design Gemini 3.0 Pro
OpenAI coding Codex (GPT 5.1)
In-editor help GitHub Copilot

05 // MODEL_SPECS

TECHNICAL SPECIFICATIONS

Claude 4.5 Opus

BEST_FOR_CODE
STRENGTHS Best model for coding and writing by far, excellent at debugging.
TARGET_USE All coding tasks, large projects, long-form writing.

Gemini 3.0 Pro

FRONTEND
STRENGTHS Best for front-end design, industry-leading context window.
TARGET_USE Front-end work, large document analysis, multimodal projects.

Codex (GPT 5.1)

OPENAI
STRENGTHS Specialized for coding, excellent multi-file projects.
TARGET_USE All OpenAI coding work, multi-file projects.

GitHub Copilot

INLINE
STRENGTHS Real-time suggestions, integrated in VS Code/editors, line-by-line autocomplete.
TARGET_USE In-editor coding, autocomplete, small functions. $10/mo (free for students/educators).

06 // OPTIMIZATION

TIPS FOR BETTER RESULTS

Include context

  • > Operating system (Windows, Mac, Linux)
  • > Language version (Python 3.11, Node 18)
  • > Frameworks or libraries you're using

Show sample data

Include examples of your input and expected output. CSV columns, JSON structure, etc.

Ask for chunks

  • 1. Core function first
  • 2. Error handling second
  • 3. Edge cases third

Verify everything

  • > Test with real data
  • > Check edge cases
  • > Verify calculations manually

07 // FAILOVER_PROTOCOL

WHEN TO SWITCH LLMs

Same error 3+ times?

Try a different model. Each has different training data and approaches.

Too slow?

Drop from Opus to Sonnet, or use Gemini 3.0 Flash for quick tasks.

Too expensive?

Use free tiers for exploration, paid for production code.

Need real-time help?

GitHub Copilot for in-editor, conversational LLM for debugging.