When you're iterating on a complex prompt (1000+ tokens with multiple sections), how do you:
1. Track what changed between versions?
2. A/B test specific sections without rewriting the whole thing?
3. Share prompts with teammates while keeping some sections private?
4. Reuse common blocks (like persona definitions or output format specs) across different prompts?
I've been thinking about this a lot while building a prompt editor, and I'm not sure if most people even feel this pain yet or if it's just a power-user problem. Would love to hear your workflows.
I've built a prototype to help me manage requirements; you can check it out. I'd like to hear your feedback.
https://www.promptbuilder.space/
Hey HN! I'm Jaber, and I built Prompt Builder.
What it is:
A block-based editor for composing AI prompts. Think of it like Notion blocks, but purpose-built for prompt engineering—drag to reorder, toggle visibility, wrap in tags (XML/JSON/etc.), and see a live combined preview.
Why I built it:
After 6 months of juggling 100+ ChatGPT/Claude prompts across Notion, Slack DMs, Apple Notes, and random text files, I got tired of the chaos. Every developer and marketer I talked to had the same problem—no good place to save, organize, and reuse prompts. Most tools are either marketplaces (buy prompts) or plain text boxes. I wanted something that treats prompts like code: modular, version-controlled, and shareable.
What makes it different:
* Block-based composition: Each prompt is built from collapsible blocks (system message, instructions, examples, constraints). Reorder via drag-and-drop, toggle which blocks appear in the final output.
* Live preview with character count: See your combined prompt update in real-time as you edit.
* Voice-to-prompt: Click the mic icon on any block, speak, and it transcribes. Pro users can translate on the fly (50+ languages).
* Share without signup: Generate a shareable link (with optional password/expiration for Pro). No account needed to share or view.
* Tags & variables: Wrap content in custom tags (JSON schema, XML, whatever). Pro users get variables for parameterized prompts.
* Export/Import (Pro): Download as JSON, Markdown, or plain text. Import existing prompt libraries.
Try it (no signup required):
promptbuilder.space
What I'd love feedback on:
1. Does the block-based approach feel natural, or is it overengineered? I debated just doing folders + plain text, but blocks let you compose complex prompts systematically.
2. What features are missing? I built this for my own workflow, but I'm sure I'm blind to obvious gaps.
3. Pricing—does $5/month (50% launch discount) feel fair for Pro? Free plan gives you 3 prompts, 5 blocks each. Pro is unlimited + translation, variables, export, etc.
Built this mostly solo over the past month. Would genuinely love to hear what the HN community thinks—especially if you work with LLMs regularly.
When you're iterating on a complex prompt (1000+ tokens with multiple sections), how do you:
1. Track what changed between versions? 2. A/B test specific sections without rewriting the whole thing? 3. Share prompts with teammates while keeping some sections private? 4. Reuse common blocks (like persona definitions or output format specs) across different prompts?
I've been thinking about this a lot while building a prompt editor, and I'm not sure if most people even feel this pain yet or if it's just a power-user problem. Would love to hear your workflows.
I've built a prototype to help me manage requirements; you can check it out. I'd like to hear your feedback. https://www.promptbuilder.space/
reply