My Super Simple Software Factory (For Agentic Engineers)

My Super Simple Software Factory (For Agentic Engineers)

More

Descriptions:

The next level of agentic engineering can be summarized in two words: Software Factory.

But this jump is larger… As agents and code commoditizes its important you realize your agents alone are NOT enough. Agents plus code beats agents alone, and engineers still throwing everything into skills and sub agents will pay for it in cost, speed, hallucinations, and broken runs. I built the Super Simple Software Factory to prove it, and I’m giving it to you for free.

🔗 RESOURCES
• Super Simple Software Factory repo: https://github.com/disler/super-simple-software-factory

🎥 VIDEO REFERENCES
• Forget Loop Engineering: https://youtu.be/VQy50fuxI34
• Is Anthropic STEALING YOUR DATA: https://youtu.be/qh4vLlit97I

✅ PUSH YOUR AGENTIC ENGINEERING BEYOND
Tactical Agentic Coding (TAC): https://agenticengineer.com/tactical-agentic-coding?y=haUfb1ievTE

🚀 Software factories are massively misunderstood and massively underappreciated. They exist for ONE reason: they give you more leverage on your prompt. At the lowest level you chain a couple agents with some config. At the highest level you build a system of agents plus code that operates without you, just as well as you would, and sometimes better.

🛠️ Three principles are wired into this thing, non negotiable in the age of agents:

Observable: If you can’t measure your agents, you can’t improve them. Full swim lane coding agent observability across every phase, every compiled prompt, every tool call, every cost breakdown.

Customizable: One YAML config controls the core four (context, model, prompt, tool) for every agent in your roster. Master the core four and you master the agent.

Reusable: One /install command drops the entire factory into any codebase. It’s a skill, so an agent can operate it AND deploy it.

🤖 We run three workflows live: a single agent Scout, an intermediate plan build test, and the full simple SDLC with deterministic gate checks between every phase. Kimi K3 plans, Gemini 3.6 Flash builds, Claude Opus 5 reviews. Nobody serious is debating “which model is best” anymore. It’s about your model stack, and the software factory is how you actually USE that stack.

💡 The big idea engineers are about to miss: everyone is agent pilled and forgot that code costs nothing, runs at the speed of light, and YOU own it. You don’t own your models, you rent them. So tell me why you’d pass a passing test suite back into your agent’s context window. Code proposes the structure, agents propose the work, code disposes and validates. That’s harness engineering, the fine line between vibe coding and agentic engineering.

⚡ What you’ll take away:
– Why the software factory is the unit of leverage above prompt engineering and context engineering
– How to structure AI developer workflows as phases: plan, build, test, fix, review, revise, document, commit
– Handing off context between agents with typed envelopes and shared session directories
– Deterministic gate checks that send failed work back to the build agent automatically
– Staying “in distribution” (just Python, just YAML, just agents, just a skill) instead of inventing a DSL
– Specialized agent harnesses per role, with sub agent support only where it earns its keep
– Where this goes next: branches, sandboxes, isolated agent runs, cloud deployment, merge steps

🌟 I’m not thinking at the prompt level anymore, or the skill level, or even the team of agents level. I think about the full end to end developer workflow I would run to get the work done, then I build THAT with agents. That’s reproducible success, thousands of runs deep, without me in the loop.

🎯 I can already see the comments: “at what scale do you need this?” Think about production. Think about the thousandth run, not the first. At some point your system gets large enough that validation is the only way forward, and you will pay that cost somewhere, in time or in tokens.

Vibe coding is not knowing how your system works and not looking. Agentic engineering is knowing how your system works so well you don’t have to look. Rise with the ceiling of agentic engineering, not the floor of vibe coding.

Clone it, steal it, roll it into your own factory. It’s here for you.

Stay focused and keep building.
Dan

#agenticengineering #softwarefactory #agenticcoding

1 Item

Channels

1 Item

People