Model-independent enterprise AI harness for autonomous software development
Build a platform that lets enterprise teams develop software with AI agents while remaining independent of any single model provider. The harness maintains workflow state, dynamically allocates intelligence across models, supports open and local models, expands effective context through compaction, and lets humans and AI develop a new software-development methodology together.
From 20VC with Harry Stebbings — Should American Enterprises Work With Open-Source Chinese Models? | Only 10% of Neo-labs survive at 06:25
Problem: Enterprises cannot reliably use a single model for every task because model quality, cost, latency, and suitability vary by use case. Gateway routing alone does not provide enough context for agentic workflows, and outsourcing intelligence to model providers can give those providers knowledge of a company's workflows and business.
For: Enterprise software-development teams that need autonomous AI coding capabilities while retaining control over their workflows, data, intelligence, deployment model, and model-provider choices.
Products from this video
arena.ai Bolt Claude Code Cursor Droids Factory Linear Lovable OpenAI Codex OpenRouter Salesforce
Examples
- Factory: described as building autonomous software-development capabilities and as using model routing, on-premise deployment, project-level credit measurement, and outcome-based evaluation.