🧠 Adversarial Poetry as a Universal Jailbreak for LLMs
🔗 Read more 🔗
🤖 Rethinking AI Agents and the Model Context Protocol
A grounded and thoughtful take that pushes beyond AI hype to highlight practical lessons from building real agent systems.
A reflective post exploring misconceptions about AI agents and the MCP standard. The author argues that effective agents rely on structured problem design, not just LLM capabilities, and shares insights from hands-on experimentation.
🔗 Read more 🔗
🖥️ 50th Anniversary of BITBLT
A concise and thoughtful overview of GPU evolution — why more bandwidth isn’t always the answer to performance bottlenecks.
A reflection on how GPU architecture has evolved over 50 years. Modern GPUs boast immense compute power but face limitations from memory bandwidth, unlike early systems that had more balanced trade-offs between speed and throughput.
🔗 Read more 🔗
🌀 Procedurally Generating Infinite Marble Runs
🔗 Read more 🔗