Tianyang Liu's personal website. https://leoii22.com/feed.xml 2026-10-16T02:03:35+00:00 Tianyang Liu 2026-09-30T00:00:00+00:00 2026-09-30T00:00:00+00:00 https://leoii22.com/blog/2026/when-ai-designs-what-makes-ui-good/

What evidence it takes to call a generated interface good, and which of those checks can be handed to automation.

2026-09-17T00:00:00+00:00 2026-09-17T00:00:00+00:00 https://leoii22.com/blog/2026/when-ai-designs-aesthetics-style-and-taste/

What taste means when models design interfaces, and how preference training shapes the styles they keep returning to.

2026-09-08T00:00:00+00:00 2026-09-08T00:00:00+00:00 https://leoii22.com/blog/2026/the-user-is-part-of-the-agent/

Human judgment shapes an agent's goals, assurance, authority, and accountability.

2026-08-28T00:00:00+00:00 2026-08-28T00:00:00+00:00 https://leoii22.com/blog/2026/reinforcement-learning-in-llms/

How rewards become policy updates, and how data, credit assignment, and rollout systems come to define agentic RL.

2026-08-12T00:00:00+00:00 2026-08-12T00:00:00+00:00 https://leoii22.com/blog/2026/gpu-doing-with-my-tensors/

A mental model of GPU compute, memory, and communication, starting from the LLM operations researchers already know.

2026-07-23T00:00:00+00:00 2026-07-23T00:00:00+00:00 https://leoii22.com/blog/2026/modern-transformers/

How modern Transformers use positional mechanisms, compressed memory, conditional computation, and residual paths to reallocate model budgets.

2026-05-26T00:00:00+00:00 2026-05-26T00:00:00+00:00 https://leoii22.com/blog/2026/linear-algebra-in-llms/

A ground-up account of representation, transformation, interaction, scale, and low-rank structure inside LLMs.