Video · IBM Technology ·
A good prompt isn't enough for good AI code. Legare Kerrison explains how context helps AI coding agents understand your repository, dependencies, tests, and intent. Learn how to give agents the right context without overwhelming them.
VideoIBM Technology
The video explains how PyTorch’s torch.compile optimizes models through Dynamo, AOT Autograd, and Inductor, including lazy compilation, graph capture, kernel fusion, and hardware-specific code generation. It also shows why speedups vary and how graph breaks, recompilation, and incorrect benchmarking can limit performance.
VideoIBM Technology
The video examines how AI is exposing weaknesses in web security, conversational model evaluation, cultural understanding, and enterprise data infrastructure. It argues that organizations must modernize legacy systems and rethink benchmarks as AI agents discover flaws faster and create new operational demands.
VideoIBM Technology
The video explains how quantization compresses LLM parameters into lower-precision formats to reduce memory use and improve inference speed. It shows the tradeoff between smaller models and potential accuracy loss, and compares post-training quantization with quantization-aware training.
VideoIBM Technology
The video examines PHP2Shell vulnerabilities affecting software embedded across up to 70% of the web, with exploit details leaking before patches and older PHP versions remaining unpatched. It also discusses Anthropic’s external AI evaluators and the hijacking of 333 Tor exit relays, emphasizing secure-by-design practices, independent testing, and basic security hygiene.