@vojtechSkip the Netflix episode and dive into Karpathy’s one-hour Stanford lecture on AI engineering. It is a solid weekend bookmark. He breaks down the reality: an LLM only delivers 10%, which is merely the starting point rather than the final product. Prompting can push that metric to 30%, but that is where most developers halt their progress. The real work involves agents, loops, and systems to bridge the gap. The graph represents the full 100% required for things to actually survive production. Most engineers obsess over the 10% (the model) and the 30% (the prompt). Karpathy dedicates the session to the remaining 70%. Afterward, check out my Kimi K3 guide, From Loops to Graphs, to see how I implemented these concepts.

@vojtechSkip the Netflix episode and dive into Karpathy’s one-hour Stanford lecture on AI engineering. It is a solid weekend bookmark. He breaks down the reality: an LLM only delivers 10%, which is merely the starting point rather than the final product. Prompting can push that metric to 30%, but that is where most developers halt their progress. The real work involves agents, loops, and systems to bridge the gap. The graph represents the full 100% required for things to actually survive production. Most engineers obsess over the 10% (the model) and the 30% (the prompt). Karpathy dedicates the session to the remaining 70%. Afterward, check out my Kimi K3 guide, From Loops to Graphs, to see how I implemented these concepts.