This is a short, ten-minute manifesto on the purpose of my 10-day AI bootcamp and the value it might create for you if you decide to join me on my journey to catch up with my technical cofounder, and to gain a sound understanding of the most fascinating, pivotal invention in my life time and possibly in human history.

My brief love story with ChatGPT

I've always been an avid user of ChatGPT; it's been an incredibly sweet tool for me.

Like everyone else, I had doubts at first. I remember testing the model with some historical topics I was well familiar with (being a history nerd myself). It felt surreal to have an artificial being conversationally narrate Napoleon's Ulm campaign of 1805. Though ChatGPT 3.5's explanation seems shallow by today's standards, it captured my imagination immediately, and I knew AI was here to stay.

When GPT-4 arrived, the answers improved dramatically. It helped me excel during an internship at a consulting startup. Though, privately, I worried I was cheating. Even some math questions were no longer a GPT gotcha—it basically carried me through the end of college.

Then came 4o and o1, who acted as my personal chief of staff and chief strategist during my first full-time role. They were phenomenal. I stopped feeling embarrassed about leaning on AI and embraced it wholeheartedly. It was clear to me that society would be permanently, fundamentally altered, for better or worse, and I needed to be prepared.

Recently, however, I've hit the ceiling of today's AIs. As smart and competent as they stand today, they refuse to go the extra mile and become truly agentic because they lack deep, long-term context about my work. This painpoint led me to start Lab 24 with my technical cofounder — Kevin. We're building better context for AI by re-thinking the entire knowledge structure for LLMs.

But here's the strange part: despite co-founding an AI company and being a power user of ChatGPT, I still constantly wrestle with FOMO. I feel my understanding is naïve and shallow in some areas, and the constant noise in the market can be overwhelming. To make matters worse, I've always hated coding.

Yet I've always loved exploring deep, fascinating concepts. To me, learning how to code from scratch feels constraining (my co-founder would strongly disagree). On the other side, understanding the inner workings of AI feels liberating, since I'm literally unearthing how entire systems learn, reason, adapt, and grow into something as powerful as they are today. The 10-day bootcamp (designed by myself with the technical guidance from my co-founder) is set to achieve exactly that. Right now I'm still an AI user; as I embark on the journey of an AI builder, I need to catch up—and I welcome you along for the ride.

The plan

• 4 hours of deep learning everyday starting from July 3rd.
• 4th of July is my only off day
• Each day tackles a core theme in AI
• Summary and reflection of my learning each day (Sign up to stay updated)

Why now?

Because why not. Running an AI company, even from the business side, requires real technical depth. I've gotten by with "good enough" knowledge for too long. Yes, the FOMO is real and the market noise is unbearably loud, but those are everywhere. The only way out is to cut through them instead of tuning them out.

I'm purposely going public with my learning because I believe I'm not alone. If you are a:

• sales lead battling imposter syndrome demoing an AI solution you barely understand
• founder who nods through your CTO's technical proposal
• marketer writing GPT-powered copy without understanding why some prompts flop
• PM drowning in feature requests to "just add AI"
• curious individual who loves exploring the ins and outs of something new

Then you will probably find this 10-day bootcamp super useful.

Agenda and materials

Day 1 - Evolution of AI: Symbolic, CNN, RL and GPT-4o

Understand the background and evolution of AI through key stages, going from symbolic AI in the 1950s to the advent of GPT-4o.

Essays:

• Attention Is All You Need
• GPT-4o System Card
• Introduction to Reinforcement Learning

Blogs & Articles:

• The brief history of artificial intelligence: the world has changed fast — what might be next?
• Convolutional Neural Networks: A Brief History of their Evolution
• OpenAI's GPT-4o: The Multimodal Revolution That Changes Everything

Videos:

• Lecture 1 | Introduction to Convolutional Neural Networks for Visual Recognition
• AlphaGo - How AI mastered the hardest board game in history

Day 2 - AI Landscape 2025

Grasp the current competitive landscape of AI startups, unicorns and enterprises.

Blogs & Articles:

• AI 100: The most promising artificial intelligence startups of 2025
• The 2025 AI Index Report
• Here are the 24 US AI startups that have raised $100M or more in 2025
• Forbes 2025 AI 50 List
• The Coming 'Collapse' of the AI Startup Bubble
• 99% of AI Startups Will Be Dead by 2026 — Here's Why

Videos:

• The Gartner Hype Cycle for Generative AI 2025
• Perplexity CEO Aravind Srinivas: From Academic to $9B AI Pioneer | HBS Entrepreneurship Summit 2025

Day 3 - Transformer Mechanics & Token Economics

Grasp how transformers turn tokens into context and how tokenization and context length drive cost and latency.

Essays:

• The Effect of Sampling Temperature on Problem Solving in Large Language Models

Blogs & Articles:

• What is a context window?
• How to Understand "Tokens" in AI Large Language Models?
• A Comparative Analysis of Byte-Level and Token-Level Transformer Models in Natural Language Processing
• Understanding OpenAI's "Temperature" and "Top_p" Parameters in Language Models
• Understanding the Cost Economics of GenAI Systems: A Comprehensive Guide

Videos:

• The math behind Attention: Keys, Queries, and Values matrices

Day 4 - Prompt Engineering & Evaluation

Master the principles of crafting, refining, and automatically evaluating prompts to steer LLM behavior with reliability and minimal trial and error.

Essays:

• Understanding and Avoiding AI Failures: A Practical Guide
• Unleashing the potential of prompt engineering for large language models
• Reverse Prompt Engineering

Blogs & Articles:

• LLM Evaluation Metrics: The Ultimate LLM Evaluation Guide
• Few shot Prompting and Chain of Thought Prompting
• How to Secure AI-Based Systems — Preventing Prompt Injection and Reverse Engineering Attacks

Videos:

• ChatGPT Roles Explained: User, Developer (System), Assistant

Day 5 – Data Foundations, Embeddings & Retrieval

Learn why chunk size, embedding space, and similarity search choices determine retrieval precision when grounding models in proprietary data.

Essays:

• LLMs Will Always Hallucinate, and We Need to Live With This

Blogs & Articles:

• Chunk size and overlap
• Vector Similarity Explained
• Embedding space and static embeddings
• Getting Started with Hybrid Search
• Unstructured Data Isn't Just for Embeddings: Hidden Structure can Improve RAG
• Time Complexity and Space Complexity
• Finding the Perfect Chunk Size & Overlap for Rec Systems
• 7 Chunking Strategies in RAG You Need To Know
• Understanding Cosine Similarity and Word Embeddings
• Hybrid Search: Combining BM25 and Semantic Search for Better Results with Langchain
• An Exhaustive Guide to Using Pinecone for Vector Databases
• Searching algorithms

Day 6 - RAG Conceptual Pipeline

Build a mental model of the full RAG flow (ingest, embed, search, cite, generate) and recognize common failure modes such as drift and hallucination.

Blogs & Articles:

• How to Minimize Latency and Cost in Distributed Systems
• What Is Ground Truth in Machine Learning?
• Optimizing Chunking, Embedding, and Vectorization for Retrieval-Augmented Generation
• How to Evaluate RAG If You Don't Have Ground Truth Data
• When to Apply RAG vs Fine-Tuning

Videos:

• RAG vs. Fine Tuning

Day 7 - Temporal Knowledge Graphs & Context Layers

Appreciate how temporal knowledge graphs encode evolving facts and how they can extend LLM memory and reasoning beyond static text corpora.

Essays:

• Relational Deep Learning - Graph Representation Learning on Relational Databases
• Enhancing Business Process Execution with a Context Engine
• A Survey on Temporal Knowledge Graph: Representation Learning and Applications

Blogs & Articles:

• Event-Centric Temporal Knowledge Graph Construction: A Survey
• Graphiti by Zep: Advanced Temporal Knowledge Graphs for Your Data
• How to Improve Multi-Hop Reasoning With Knowledge Graphs and LLMs
• How to Implement Graph RAG Using Knowledge Graphs and Vector Databases

Day 8 - Safety, Alignment & Evaluation

Internalize the major alignment challenges (toxicity, bias, privacy), review leading evaluation frameworks, and outline practical risk-mitigation steps

Essays:

• Large Language Model Safety: A Holistic Survey
• Align in Depth: Defending Jailbreak Attacks via Progressive Answer Detoxification
• EXPLAINING AND HARNESSING ADVERSARIAL EXAMPLES

Blogs & Articles:

• SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decoding
• Evals the Lifeline of AI Agents
• AI's Regulatory Reckoning — EU AI Act and Ripple Effects on U.S. Technology Policy
• AI's Regulatory Reckoning — EU AI Act and Ripple Effects on U.S. Technology Policy

General Knowledge Repository:

• Awesome LLM-Safety Repository

Day 9 – Deploy, Scale & Cost

Gain hands-on insight into packaging an LLM service, containerizing it, deploying to cloud, and optimizing for latency, concurrency, and dollar cost.

Blogs & Articles:

• LLM Inference as-a-Service vs. Self-Hosted: Which is Right for Your Business
• The Costly Open-Source LLM Lie
• Commercial vs. Self-Hosted LLMs: A Cost Analysis & How to Choose the Right Ones for You
• Running GPU Inference on Kubernetes Without Breaking the Bank
• Best practices for serverless inference
• Using serverless GPUs in Azure Container Apps
• The Costly Open-Source LLM Lie

Day 10 - Future SoTA & Business Implications

Forecast frontier AI trends (MoE, multimodality, synthetic data, agents) and convert them into a concise business pitch and integration roadmap.

Essays:

• Soup-of-Experts: Pretraining Specialist Models via Parameters Averaging
• Multimodal Foundation Models: From Specialists to General-Purpose Assistants
• Examining the Expanding Role of Synthetic Data Throughout the AI Development Pipeline

Blogs & Articles:

• State-of-the-Art (SOTA) AI Models: LLMs, NLP, and Computer Vision

Videos:

• The Next Breakthrough In AI Agents Is Here
• What is Multimodal AI? | The AI Research Lab - Explained
• How do Multimodal AI models work? Simple explanation
• The Rise of Multimodal AI Agents: What You Need to Know
• AI is replacing AI Agencies… Now what?
• State-Of-The-Art Prompting For AI Agents

Kevin (Left) & Frank (Right) the day Frank starts his day 1