Blog – AI
Insights and Ideas from Our Team
Filter By:
The Bigger-Model Trap: Why We Stopped Chasing the Largest LLM
Learn why defaulting to the largest open-source LLM backfires and how right-sizing four specialized models built a faster, more reliable AI app.
The AI Journey – A Complete Decision Trajectory
Learn how to outline your AI options, make deliberate decisions, and determine controls and data access to your intellectual property.
Deterministic AI: How to Build an AI Application That Gives the Same Answer Every Time
Learn why reliable AI applications are mostly code, not AI models, and how to design a system that gives consistent, repeatable results.
Skills, Agents, and Tools: Why FileMaker AI Agentic Coding Needs a Harness
Learn why FileMaker AI agentic coding needs a harness as well as the skills, agents, and tools that get you from A to Z.
Different Ways of Obtaining AI-Generated Descriptions for Images Stored in FileMaker
Learn how to obtain AI-generated captions or descriptions for images stored in FileMaker fields from Claris FileMaker and AI experts.
The Claris AI Model Server and AI-Generated Captions for Images Stored in FileMaker
Learn about two new script steps, Insert Image Caption and Insert Image Captions in Found Set, introduced in FileMaker 2026.
Quick FileMaker Up-skill with Your LLM
Learn how to feed your LLM of choice your FileMaker DDL for a quick context upgrade, an easy up-skill from our FileMaker and AI experts.
From Concept to Production: Lessons from Building a Production AI Assistant
What it takes to build an AI assistant that works in production. Real architecture decisions, honest lessons, and the practices that keep it trustworthy.
Model Context Protocol (MCP): The Protocol That Connects AI to Your Business
MCP connects your AI applications to real business data. Here’s what it is, how it works, and what it means for your AI strategy.
How to Use AI to Generate WebViewer Code in FileMaker
Learn how AI-assisted development tools (Claude, ChatGPT) helps FileMaker developers generate WebViewer code in minutes instead of hours.