SoftwareFactory : 0 --> 1 : Harness Engineering from First Principles

​Harness Engineering from First Principles

​Most software teams use AI to generate snippets.

​What happens when we design a system where AI agents can reliably help build, ship, track, and improve entire software products?

​Software Hacktory: 0 → 1 is a practical workshop on harness engineering: building the surrounding systems, workflows, guardrails, integrations, and feedback loops that turn capable models into reliable software-building agents.

​Together, we’ll live-code a software factory using the DeepSeek harness and explore how to design an end-to-end software development lifecycle for agents—from an idea or ticket, through implementation and validation, to visibility into what happened along the way.

​We’ll cover:

• ​How to build a software factory and agent-ready software stack from first principles

• ​How agents can work with product and engineering workflows through tools such as Jira and Linear

• ​How to structure planning, implementation, testing, review, and delivery across an agentic SDLC

• ​How to track work, inspect agent decisions, and build observability around cost, traces, tool calls, failures, and outcomes

• ​How to run the same workflow across different model tiers—from local SLMs to frontier-model subscriptions

• ​The practical trade-offs: local vs cloud, cost vs capability, speed vs reliability, autonomy vs control

​This is not a “press a button and let AI build your startup” session.

​It is a workshop on the engineering work that makes agentic software delivery actually useful: clear interfaces, good context, strong feedback loops, reliable tooling, evaluations, and human control where it matters.

​Expect a mix of first-principles explanation, live building, and hands-on experimentation. You’ll leave with a clearer mental model of harness engineering—and a working sense of how to begin building your own software factory from 0 to 1.