---
title: "Sarah Pan (@sarahpan) | Hashnode"
description: "Software engineer exploring AI, games, and creative workflows. Read the latest articles by Sarah Pan on Hashnode."
url: "https://hashnode.com/@sarahpan"
---

@sarahpan

# Sarah Pan

@sarahpan · Los Angeles, USA · Joined July 2026

Software engineer exploring AI, games, and creative workflows.

### [When AI Models Disagree, Don’t Use Majority Vote](https://sarahtokenbay.hashnode.dev/when-ai-models-disagree-don-t-use-majority-vote?utm_source=hashnode&utm_medium=feed)

Asking a second AI model is one of the easiest ways to check an uncertain answer. But what happens when the second model disagrees with the first? You ask a third model. Now one model says A, another

6 min read · [sarahtokenbay.hashnode.dev](https://sarahtokenbay.hashnode.dev/when-ai-models-disagree-don-t-use-majority-vote?utm_source=hashnode&utm_medium=feed)

[Jul 23](https://hashnode.com/posts/when-ai-models-disagree-don-t-use-majority-vote/6a61b8e373b2ddb6d2fc9820)

### [Building a Model-Agnostic Agent Harness: What Breaks When You Switch LLMs](https://sarahtokenbay.hashnode.dev/building-a-model-agnostic-agent-harness-what-breaks-when-you-switch-llms?utm_source=hashnode&utm_medium=feed)

Six months of production lessons from multi-model systems

7 min read · [sarahtokenbay.hashnode.dev](https://sarahtokenbay.hashnode.dev/building-a-model-agnostic-agent-harness-what-breaks-when-you-switch-llms?utm_source=hashnode&utm_medium=feed)

[Jul 20](https://hashnode.com/posts/building-a-model-agnostic-agent-harness-what-breaks-when-you-switch-llms/6a5de27a351e9e515766653d)

### [The API Call Succeeded. The Workflow Still Broke.](https://sarahtokenbay.hashnode.dev/the-api-call-succeeded-the-workflow-still-broke?utm_source=hashnode&utm_medium=feed)

We're building TokenBay, a unified API that routes requests across GPT, Claude, Gemini, DeepSeek, and a few dozen other models. A few weeks ago we ran a fairly boring internal test: send the same stru

6 min read · [sarahtokenbay.hashnode.dev](https://sarahtokenbay.hashnode.dev/the-api-call-succeeded-the-workflow-still-broke?utm_source=hashnode&utm_medium=feed)

[Jul 16](https://hashnode.com/posts/the-api-call-succeeded-the-workflow-still-broke/6a5887d9e9b3777b5902c9ed)

### [The Best AI Model Is Usually the Wrong Question](https://sarahtokenbay.hashnode.dev/the-best-ai-model-is-usually-the-wrong-question?utm_source=hashnode&utm_medium=feed)

What matters more is how easily your workflow can survive when the answer changes.

5 min read · [sarahtokenbay.hashnode.dev](https://sarahtokenbay.hashnode.dev/the-best-ai-model-is-usually-the-wrong-question?utm_source=hashnode&utm_medium=feed)

[Jul 15](https://hashnode.com/posts/the-best-ai-model-is-usually-the-wrong-question/6a5729b79ad941045aa4d845)

### [Three AI APIs Now Look Almost Compatible. They Still Aren’t Interchangeable.](https://sarahtokenbay.hashnode.dev/three-ai-apis-now-look-almost-compatible-they-still-aren-t-interchangeable?utm_source=hashnode&utm_medium=feed)

OpenAI, Claude, and Gemini now look surprisingly similar at the API level. In some cases, switching providers requires little more than changing an API key, a base URL, and a model name. But this comp

4 min read · [sarahtokenbay.hashnode.dev](https://sarahtokenbay.hashnode.dev/three-ai-apis-now-look-almost-compatible-they-still-aren-t-interchangeable?utm_source=hashnode&utm_medium=feed)

[Jul 14](https://hashnode.com/posts/three-ai-apis-now-look-almost-compatible-they-still-aren-t-interchangeable/6a55db195930630cc5ded5d3)