In-depth insights into software architecture, best practices, and lessons learned from real production systems.
A practical analysis of when microservices make sense and when they do not. Based on real successes and failures in distributed systems architecture.
These videos are carefully curated external resources selected by Profectus Lab to help engineers understand modern AI and system design principles.
~1:56:00 A rare hands-on explanation of GPT internals, connecting theory with real code. Ideal for engineers who want to truly understand how modern LLMs are built.
~1:00:00 Karpathy's full hour walkthrough of how LLMs actually work — inference, training, fine-tuning, and the emerging LLM-OS picture. The clearest single-shot mental model on the internet for engineers new to the field.
~27:00 3Blue1Brown's signature visual treatment of the transformer architecture. The single best 30-minute primer for engineers who want the intuition before the math.
~16:00 StatQuest's hallmark step-by-step teaching style applied to the attention mechanism. Removes the magic and leaves you with a working intuition for self-attention.
~14:00 3Blue1Brown's visual proof of how networks actually learn. The clearest mental model of gradient descent and the chain rule available in 15 minutes.
~18:00 Breaks the "neural networks are black boxes" myth. Step-by-step walk through the math with the simplest possible example — the right starting point if neural nets feel impenetrable.
~50:00 Mosh Hamedani's hands-on intro that walks from zero through a real ML pipeline in Python. The fastest end-to-end ML primer for engineers who already know Python.
~15:00 Michael Phi pairs every transformer concept with a clean animation. The right second video to watch after 3Blue1Brown — same intuition, different angle.
~6:00:00 The MCP track keynote with the Anthropic team. If you want to understand why MCP became the industry-standard protocol for connecting LLMs to tools in 2025, this is the single best primary source.
~2:00:00 A hands-on workshop from Anthropic on building production agents with the Claude Agent SDK — tool use, sub-agents, hooks, MCP servers, and the patterns that scale beyond the demo.
~8:00:00 The keynote stream from the largest technical AI conference of 2024. A snapshot of the state of AI engineering — what shipped, what worked, what didn't — straight from the teams building it.