Logic Bits
We occasionally publish our thoughts, experiments, and lessons learned while building Logic. You can find them on Logic Bits, 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.
- AI Is Forcing Us To Write Good Code: How AI development tools are reinforcing best practices like testing and documentation. Read it here.
- Codex is a Slytherin, Claude is a Hufflepuff: A comparative look at how different AI coding agents behave in practice. Read it here.
- Fine-Tuning Is (Probably) A Trap: Why fine-tuning LLMs often isn’t the right answer, and what to do instead. Read it here.
- How to Ship Confidently When Your Backend Makes Things Up: Strategies for handling LLM hallucinations and unreliable outputs in production. Read it here.
- 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.
- Making a List and Checking It Twice: How structured output works with LLMs, and the limitations it has. Read it here.