Gists

Abstract

This paper presents a self-evolving framework, Recursive Knowledge Crystallization (RKC), designed to overcome the “Catastrophic Forgetting” inherent in autonomous AI agents. By persisting evolved technical insights into a universally readable SKILL.md file based on the Agent skills specification, this approach establishes long-term memory and cross-platform portability. The framework was empirically validated through the development of gas-fakes, a highly complex Node.js-to-Google Apps Script (GAS) emulation library. The results demonstrate that agents can autonomously internalize project-specific architectural patterns and environmental nuances. Consequently, the framework achieves Zero-Shot Knowledge Transfer across distinct toolchains (Google Antigravity and the Gemini CLI) while maintaining absolute 1:1 behavioral parity with the live GAS environment.

Gists

Abstract

Discover how to seamlessly integrate Google Workspace with GitHub Actions using the gas-fakes library. This guide demonstrates running Google Apps Script locally and within CI/CD pipelines without deploying Web Apps. Automate workflows, secure credentials, and effortlessly interact with Google Drive and Sheets directly from your repository.

Introduction

Google Apps Script (GAS) is a powerful low-code platform that enables developers to integrate, automate, and extend Google Workspace with ease. Ref Typically, executing GAS requires the script to be hosted on Google’s servers via the Script Editor. While tools like clasp allow for local development and synchronization, running scripts from outside the Google ecosystem—such as from a local environment or a different cloud provider—often involves complex setups relying heavily on the Apps Script API or Web Apps. Ref

I got an awarded for publishing a top Google AI post for this article. Recursive Knowledge Crystallization: A Framework for Persistent Autonomous Agent Self-Evolution

Gists

Abstract

In the development of autonomous agents using Large Language Models (LLMs), restrictions such as context window limits and session fragmentation pose significant barriers to the long-term accumulation of knowledge. This study proposes a “self-evolving framework” where an agent continuously records and refines its operational guidelines and technical knowledge—referred to as its SKILL—directly onto a local filesystem in a universally readable format (Markdown). By conducting experiments across two distinct environments featuring opaque constraints and complex legacy server rules using Google’s Antigravity and Gemini CLI, we demonstrate the efficacy of this framework. Our findings reveal that the agent effectively evolves its SKILL through iterative cycles of trial and error, ultimately saturating its learning. Furthermore, by transferring this evolved SKILL to a completely clean environment, we verify that the agent can successfully implement complete, flawless client applications in a single attempt (zero-shot generation). This methodology not only circumvents the limitations of short-term memory dependency but also pioneers a new paradigm for cross-environment knowledge portability and automated system analysis.

Gists

Abstract

This article demonstrates how to build an adaptive learning agent using Agent-to-User Interface (A2UI), Gemini, and Google Apps Script. We explore a system that generates personalized quizzes, tracks performance in Google Sheets, and dynamically adjusts difficulty to maximize learning efficiency within the Google Workspace ecosystem.

Introduction

A2UI (Agent-to-User Interface) represents a paradigm shift in how users interact with generative AI. Originally open-sourced by Google and implemented in TypeScript and Python Ref, A2UI becomes even more powerful when integrated with Google Apps Script (GAS). This combination enables seamless access to the Google Workspace ecosystem, transforming static documents into intelligent, agentic applications.

Gists

Abstract

This article explores A2UI (Agent-to-User Interface) using Google Apps Script and Gemini. By generating dynamic HTML via structured JSON, Gemini transforms Workspace into an “Agent Hub.” This recursive UI loop enables complex workflows where the AI builds the specific functional tools required to execute tasks directly.

Introduction: The Evolution of AI Interaction

The Official A2UI framework by Google marks a significant paradigm shift in how we interact with artificial intelligence. Short for Agent-to-User Interface, A2UI represents the evolution of Large Language Models (LLMs) from passive chatbots into active agents capable of designing their own functional interfaces. Building upon my previous research, A2UI for Google Apps Script and Bringing A2UI to Google Workspace with Gemini, I have refined this integration to support sophisticated, stateful workflows.

Gists

Abstract

This article details the development of Smart Stowage Optimizer, a web-based digital twin for logistics that bridges the gap between physical safety and artificial intelligence. By integrating Gemini 3 Pro, the system solves the 3D Bin Packing Problem (3DBPP) using advanced spatial reasoning. Built with React 19 and Three.js, the application visualizes physics-aware load stability in real-time, offering a comparative analysis between traditional heuristic algorithms and modern generative AI agents.

Gists

Abstract

This article explores implementing the Agent-to-User Interface (A2UI) protocol within Google Apps Script. It demonstrates utilizing Gemini’s structured output to render secure, dynamic, server-driven UIs—like booking forms and event lists—directly inside Google Sheets, streamlining workflows without complex external infrastructure.

Introduction

I recently published a sample implementation demonstrating how to bring the Agent-to-User Interface (A2UI) concepts to Google Apps Script (GAS). Ref

A2UI is an emerging open-standard protocol designed to enable AI agents to generate rich, interactive user interfaces that render natively across web, mobile, and desktop environments. Ref Unlike traditional approaches that often require executing arbitrary, high-risk code to generate UI on the fly, A2UI leverages a strict schema-based data format to describe UI components. This “secure-by-design” architecture effectively mitigates security risks like Cross-Site Scripting (XSS) while ensuring high performance and cross-platform consistency.

Gists

Abstract

The Gemini API now supports external file URLs, allowing developers to process data directly without uploading it first. This article demonstrates how to leverage this update to integrate Google Workspace resources—including Google Sheets, Docs, Slides, and Apps Script—into Gemini’s workflow, covering both public and secure private access methods.

Introduction

Recently, the limitations regarding inline file data in the Gemini API have been significantly updated Ref. The maximum file size has increased from 20 MB to 100 MB. Furthermore, external file URLs—both public and signed—can now be used directly as input.

Gists

Abstract

This article demonstrates how to implement Google’s A2UI (Agent-to-User Interface) using Google Apps Script (GAS). By porting official Python/TypeScript examples to GAS, we show how to create dynamic, AI-generated interfaces within Google Workspace, enabling flexible business automation and interactive user experiences without complex server infrastructure.

Introduction

Google recently released A2UI, a protocol designed for agent-driven interfaces. Ref A2UI enables AI agents to generate rich, interactive user interfaces that render natively across web, mobile, and desktop environments without executing arbitrary code.