I'm running a 2-week validation experiment before building my next product.
Concept: Visual AI orchestration platform
- Drag blocks (GPT-5.1, Claude Opus 4.5, Llama Maverick 4, etc.)
- Connect them into workflows
- Parallel execution (not sequential like Zapier)
- Zero code required
Why: n8n is too complex for non-devs. Zapier doesn't do parallel execution. Gap exists.
The bet: IF 500 signups in 2 weeks → I build it (3 months). IF not → I pivot/kill it.
Questions for HN:
1. Is "Scratch for AI" clear or confusing?
2. Is 2 weeks enough validation?
3. What would make this actually useful vs just another automation tool?
Appreciate all brutal feedback.
0 comments
—
Show HN: Validating "Scratch for AI agents" before building · HackerTrans
Concept: Visual AI orchestration platform - Drag blocks (GPT-5.1, Claude Opus 4.5, Llama Maverick 4, etc.) - Connect them into workflows - Parallel execution (not sequential like Zapier) - Zero code required
Why: n8n is too complex for non-devs. Zapier doesn't do parallel execution. Gap exists.
The bet: IF 500 signups in 2 weeks → I build it (3 months). IF not → I pivot/kill it.
Current: 26/500 (Day 2)
Landing page: https://orchastra.org (mockups only - no product yet)
Questions for HN: 1. Is "Scratch for AI" clear or confusing? 2. Is 2 weeks enough validation? 3. What would make this actually useful vs just another automation tool?
Appreciate all brutal feedback.