> ## Documentation Index
> Fetch the complete documentation index at: https://docs.logic.inc/llms.txt
> Use this file to discover all available pages before exploring further.

# The Logic Bits Publication

> Logic Bits is the official publication for Logic, where you can find essays, experiments, and ideas on AI, product design, and more.

## Logic Bits

We occasionally publish our thoughts, experiments, and lessons learned while building Logic. You can find them on [Logic Bits](https://bits.logic.inc), our public Substack.

The topics vary, but include guides on how to use AI as a tool successfully, experiments with AI models, and insights into Logic development.

### Bits Posts

You can find posts on a variety of topics, including the following:

* **Your agent works. Now what?**: What comes after building an agent — deployment, iteration, and making it reliable in production. [Read it here](https://bits.logic.inc/p/your-agent-works-now-what).
* **AI Is Forcing Us To Write Good Code**: How AI development tools are reinforcing best practices like testing and documentation. [Read it here](https://bits.logic.inc/p/ai-is-forcing-us-to-write-good-code).
* **Codex is a Slytherin, Claude is a Hufflepuff**: A comparative look at how different AI coding agents behave in practice. [Read it here](https://bits.logic.inc/p/codex-is-a-slytherin-claude-is-a).
* **Fine-Tuning Is (Probably) A Trap**: Why fine-tuning LLMs often isn't the right answer, and what to do instead. [Read it here](https://bits.logic.inc/p/fine-tuning-is-probably-a-trap).
* **How to Ship Confidently When Your Backend Makes Things Up**: Strategies for handling LLM hallucinations and unreliable outputs in production. [Read it here](https://bits.logic.inc/p/how-to-ship-confidently-when-your).
* **Getting gpt-4o-mini to perform like gpt-4o**: Using few-shot learning to bring smaller model performance closer to larger models. [Read it here](https://bits.logic.inc/p/getting-gpt-4o-mini-to-perform-like).
* **Making a List and Checking It Twice**: How structured output works with LLMs, and the limitations it has. [Read it here](https://bits.logic.inc/p/making-a-list-and-checking-it-twice).