Master generative AI through a 100-day immersive challenge covering prompt engineering, model architecture, and real-world deployment. This comprehensive catalog provides daily practical tasks to transform beginners into AI experts.
Explore the history and architecture of Large Language Models (LLMs). Identify the key differences between transformer-based models and traditional neural networks.
Interact with a major chat model to understand zero-shot learning. Observe how the model handles basic instructions without prior context.
Experiment with the 'Act As' framework. Assign a specific persona (e.g., 'Senior Software Engineer') to the AI to observe shifts in tone and precision.
Adjust generation parameters. Compare high temperature (creative/random) versus low temperature (deterministic/focused) outputs.
Instruct the AI to 'think step-by-step.' Analyze how breaking down logic reduces hallucinations in mathematical and reasoning tasks.
Provide 3-5 examples within a prompt to guide the AI toward a specific formatting style or classification pattern.
Use triple quotes, XML tags, or brackets to separate instructions from content. Practice organizing complex multi-part prompts.
Provide a long article and request a bulleted summary. Experiment with length constraints (e.g., 'summarize in exactly 50 words').
Build a prompt that classifies customer reviews as positive, negative, or neutral, outputting the result in a clean JSON format.
Use generative AI to co-write a short story where the user provides plot turns and the AI generates descriptive prose.
Explore text-to-image models. Learn the importance of descriptive adjectives and lighting keywords in visual prompts.
Use negative prompting to exclude specific elements (e.g., 'no trees', 'no blur') from generated images to refine visual output.
Apply the style of a famous painter to a modern photo description using generative diffusion models.
Prompt an AI to generate a basic HTML/CSS landing page layout. Evaluate the cleanliness and validity of the code.
Paste a broken code snippet and ask the AI to identify the error and suggest a fix. Practice iterative refinement.
Provide a Python function and prompt the AI to write a comprehensive suite of unit tests using the pytest framework.
Test prompts for gender or racial bias. Discuss how training data influences model stereotypes and how to mitigate them.
Prompt the AI for facts about non-existent events. Analyze why models 'hallucinate' and how to verify information using RAG concepts.
Take a complex scientific paper and prompt the AI to explain the core concept to a 5-year-old using simple analogies.
Translate a technical document into three different languages, focusing on maintaining cultural nuances and technical accuracy.
Explore the LangChain framework. Understand how to 'chain' multiple AI calls together to complete complex workflows.
Use a Python script to connect to an LLM API and maintain a conversation history in a loop.
Learn how AI models access external data. Explore the concept of providing 'context' from a PDF or database.
Research how vector databases (like Pinecone or Chroma) store text as numerical embeddings for fast similarity searches.
Generate numerical embeddings for a list of sentences and visualize how similar meanings cluster together in vector space.
Practical: Use a vector store to allow an LLM to answer questions about a specific set of local text files.
Experiment with token limits. See what happens when a prompt exceeds the model's maximum context length.
Design a 3-step prompt chain: 1) Brainstorm ideas, 2) Outline the best idea, 3) Write the introduction.
Upload a CSV file to an AI data analyst tool. Request a correlation matrix and a trend analysis visualization.
Study the pros and cons of fine-tuning a model on custom data versus using a RAG-based approach for knowledge retrieval.
Explore autonomous agents that can set their own goals and execute tasks like web searching and file management.
Practical: Create a script where an AI agent searches the web for a topic, synthesizes the findings, and writes a report.
Learn about frameworks like AutoGen or CrewAI where multiple AI agents (e.g., Writer, Editor, Researcher) work together.
Learn how users can 'hack' prompts to bypass safety filters. Practice writing 'system prompts' that are resistant to injection.
Understand the concept of RLHF (Reinforcement Learning from Human Feedback) and how it creates model guardrails.
Install an open-source model locally using Ollama. Compare its performance to a cloud-based API model.
Explore quantization techniques (GGUF, AWQ) that allow massive models to run on standard laptops with limited VRAM.
Learn how to navigate the Hugging Face Hub. Discover how to download models, datasets, and use 'Spaces' for demos.
Use Python to generate an image programmatically using the Hugging Face Diffusers library.
Understand the basic pipeline for fine-tuning: preparing a dataset, choosing a base model, and using LoRA (Low-Rank Adaptation).
Practical: Clean and format a custom dataset into the JSONL format required for model fine-tuning.
Use a powerful LLM to generate 100 training examples for a smaller, niche-specific model.
Learn about metrics like BLEU, ROUGE, and using another LLM as a judge (LLM-as-a-Judge) to score responses.
Configure persistent 'Custom Instructions' to ensure every response follows a specific tone, length, and format.
Build an automated email draft generator that takes meeting notes and turns them into professional follow-ups.
Use OpenAI's Whisper model to transcribe an audio file. Evaluate its accuracy against background noise and accents.
Convert generated text into high-quality human-like speech using ElevenLabs or open-source alternatives like Bark.
Upload an image and prompt a Vision model (like GPT-4V or LLaVA) to describe the scene or extract text from it.
Prompt an AI to create a slide-by-slide outline and content for a 10-minute presentation on a technical topic.
Build a functional system that allows you to chat with your own personal notes from the last 50 days.
Explore text-to-video tools. Understand how prompts influence motion, camera angles, and temporal consistency.
Generate a visual storyboard for a 30-second commercial using consistent character prompts across multiple images.
Practice the AIDA (Attention, Interest, Desire, Action) framework by generating high-converting ad copy for a product.
Use AI to analyze a blog post for keyword density and suggest improvements for search engine ranking.
Practical: Write a script that scrapes headlines, summarizes them, and sends a daily digest email via AI.
Deep dive: Learn how the 'Attention' mechanism allows models to weigh the importance of different words in a sentence.
Learn how models keep track of word order and how different tokenizers (BPE, WordPiece) affect model efficiency.
Learn to use formulas and plugins that integrate LLMs directly into spreadsheets for mass data cleaning.
Create a prompt template that classifies support tickets and generates a draft response based on a knowledge base.
Master 'Function Calling'—instructing the AI to output arguments for a specific function rather than conversational text.
Practical: Use Function Calling to allow an AI to 'call' a weather API and report real-time data to the user.
Explore LlamaIndex for connecting LLMs to diverse data sources like Slack, Notion, and Discord.
Learn how to use a 'Re-ranker' model to improve the accuracy of documents retrieved from a vector database.
Compare traditional keyword searching to semantic (meaning-based) searching using embeddings.
Practical: Feed a textbook chapter into an LLM and have it generate 10 multiple-choice questions with an answer key.
Setup a roleplay prompt where the AI acts as a language tutor, correcting your grammar in real-time.
Learn why and how to version your prompts like code using tools like PromptLayer or simple Git workflows.
Calculate the cost of running a high-volume AI application. Learn strategies for optimizing token usage to save money.
Study the lifecycle of an AI model from development to production monitoring and continuous improvement.
Use an evaluation framework to run 'unit tests' on your LLM outputs to ensure they remain consistent after prompt changes.
Analyze the technology behind deepfakes and discuss the societal implications of AI-generated misinformation.
Practice removing PII (Personally Identifiable Information) from a dataset before sending it to a cloud LLM API.
Explore how generative AI is used for medical note summarization and drug discovery within strict regulatory limits.
Prompt an AI to identify high-risk clauses in a standard Terms of Service or Non-Disclosure Agreement.
Practical: Integrate an LLM with Discord to create a community assistant that answers member questions.
Use AI to generate user personas, wireframe descriptions, and microcopy for a new mobile application.
Explore tools that generate royalty-free background music or sound effects from text descriptions.
Study the mechanics of voice cloning. Discuss the balance between creative utility and the risk of fraud.
Learn about 'Lost in the Middle' phenomena and how to structure prompts so the most important info is at the beginning or end.
Learn the ReAct (Reason + Act) pattern where the AI documents its thoughts before taking an action.
Practical: Use tools like Perplexity or custom API scripts to automate the gathering of citations for a blog post.
Build a prompt that scans user-generated comments for hate speech, spam, or self-harm, flagging them for human review.
Learn how to read benchmarks like MMLU, GSM8K, and HumanEval to choose the right model for your task.
Understand the basics of fine-tuning an image model (like Stable Diffusion) on a specific face or art style using LoRA.
Explore experimental models that generate 3D meshes and textures from simple text descriptions.
Use AI to generate a cohesive brand: Logo descriptions, color palettes, and a mission statement for a fictional company.
Practical: Use AI to write a simple Chrome extension that summarizes the current webpage with a single click.
Design an AI Dungeon Master that manages inventory, health points, and narrative flow for a text-based RPG.
Learn how AI can generate alt-text for images, simplify complex text for cognitive disabilities, and provide audio descriptions.
Study international frameworks (like the EU AI Act) and how they impact the development of generative models.
Learn about load balancing for AI APIs, caching common responses, and using 'small models' for simple tasks.
Design a complex workflow using a state machine where the AI moves through different 'modes' based on user input.
Use the RAGAS framework to measure Faithfulness, Answer Relevance, and Context Precision in your RAG system.
Use AI to perform a SWOT analysis (Strengths, Weaknesses, Opportunities, Threats) for a business idea.
Explore the theoretical concepts of Artificial General Intelligence (AGI) and what it means for the future of work.
Integrate an LLM into a workflow using tools like Make.com or n8n without writing custom backend code.
Curate your best projects from the last 97 days into a GitHub repository or personal website.
Learn how 'Teacher' models (large) can train 'Student' models (small) to perform specific tasks with high efficiency.
Construct a complete, end-to-end autonomous agent that solves a specific problem (e.g., a travel planner or code reviewer).
Review the journey. Document your key learnings and set goals for staying updated in the ever-evolving field of AI.
Disclaimer
This content may have been generated with the assistance of AI tools. While we strive for accuracy, please verify important information independently. If you notice any errors or have concerns, please contact us.
Archived for the Perpetual Record
