Show HN: I rebuilt a 4-year-old app in 5 days using many agents – here's harness
2 days ago
- GPT-6 Astra demonstrates advanced LLM capabilities in programming, computer use, browsing, math, science, cybersecurity, abstract reasoning, and long-context reasoning.
- Agents need easy access to app interfaces, documentation, SDKs, logs, databases, session records, past decisions, external resources, issue trackers, and team channels, while restricting access to sensitive production data.
- Agents can coordinate with each other autonomously, triggered by events like bug reports, support tickets, code reviews, and monitoring alerts, reducing the need for direct human supervision.
- Letting go of control is essential; many routine tasks can be trusted to agents, with limited consequences for mistakes, and hierarchical agent structures can manage complex workflows.
- Maintaining documentation and specifications at scale can be automated by organizing specs in folders, using flexible templates, referencing commits, and leveraging AI to capture screenshots and sync docs.
- A personal example shows rebuilding a four-year-old side project in five days using an agent-driven setup with Grok Bot, Pi, Herdr, and a VPS, demonstrating near-autonomous development.
- The takeaway: product development is moving toward full autonomy, requiring expertise but offering massive productivity gains; start small with vision and style guides, and iterate.