Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Anthropic ChatGPT Output Wrangling
This talk explores best practices for interacting with AI APIs, covering prompt design, idempotence, schema management, error detection, and practical demos using Anthropic and ChatGPT in Go.
The nature of LLM APIs requires new thinking in terms of how you interact with the API. Prompts are not interfaces, but they can be made to feel a bit more like it.
Should they? How are they different? How are the same?
How do we leverage the unique capabilities of an LLM to make it easier? Dynamic data formats are one thing and I’ll show you how.
What is idempotence in an AI? Will it even reliably give the same result? I’ll demo some examples.
Can you achieve anything like “functional” requests? What new types of errors will occur? I’ll demo bad data detection, schema definition, some concepts for schema management and ways to determine if the AI met your schema.
I’ll demo a few examples of an increasingly complex interface with a few well known APIs, specifically Anthropic and ChatGPT with code samples in Go.