PinnedInOpenCUIbySean Wu·Apr 4, 2023Interaction logic #3: a type-based approachThe flow-based approach can not the the right tool to build conversational user interface, and why you should switch to type-based…
Sean Wu·Feb 24Agents: just Software You can BuildAgent is getting more attention every day. Unfortunately, the most common definition — an agent interacts with an environment and can use…
Sean Wu·Dec 11, 2025Why Large Language Models are Trapped in Consensus HullReferring to transformer-based Large Language Models (LLMs) like ChatGPT as “Generative AI” is widely accepted, but this label quietly…
Sean Wu·Nov 25, 2025Kill First-Time Frustration, with Conversational InteractionsSo while you’re deploy conversational experience for your team, it’s time to also ask: what about my user?
Sean Wu·Jun 30, 2025Three Ways to program with LLMsLarge Language Models (LLMs) are changing how we build software. But just like any tool, the way you use them makes all the difference…
Sean Wu·May 28, 2025Why Prompt will Never Be as Deterministic as ProgrammingIn the age of large language models (LLMs), prompt engineering has emerged as a new way to define behavior. But as we embrace this new…
Sean Wu·Apr 23, 2025Why State Is Crucial for Agent FrontendsIn the era of LLM application, we still need principled software engineering, flow for backend, and statechart for frontend…
Sean Wu·Dec 2, 2024Logic in Chatbots: Business vs InteractionSeparating business logic from interaction logic is essential for building GUI applications. One could argue that the same separation is…
Sean Wu·Oct 18, 2024AI Agents #3: Collaboration in OpenAI SwarmMulti-agent systems (MAS) have experienced waves of popularity across various domains over the years, driven by their ability to model and…
Sean Wu·Oct 7, 2024Dialog Understanding 12: Multi-Task or Single-Task Instruction tuning?Instead of fine-tuning one model for multiple task, divide and conquer, approach, fine-tune one model for each task, it way to go.
