Learn
The software factory library
Plain-language guides to software factories, coding agents, and the verification that makes them safe to run.
The AI SDLC: what actually changes in each phase
AI in the SDLC means two different things: assistants that make each phase faster, and agents that run phases without a person present. A phase-by-phase look at planning, coding, testing, review, deployment, and maintenance, honest about which level of automation is real in 2026.
Looking for a Factory AI alternative?
Factory (factory.ai) sells Droid, an agent-native platform you work inside. Darkshift is a background software factory: tasks in, pull requests out, with recorded browser QA. An honest comparison, including when Factory is the better buy.
Coding agent security: assume the agent is compromised
Prompt injection makes a coding agent’s judgment untrustworthy by design. The fix is architectural: brokered credentials the agent never holds, per-repo scoped tokens, host-side git, and evidence on every change. How Darkshift implements it.
Looking for a Devin alternative?
Devin is a hosted coding agent priced in ACUs. Darkshift is a software factory you can own: agent-agnostic, self-serve at $99 per user per month plus AI at provider cost, or installed on your own infrastructure. An honest comparison of the two operating models, including when Devin is the right choice.
Claude Code vs Codex
We run Claude Code and Codex side by side, in the same sandboxes, on the same backlogs, with the same checks on the output. How the two coding agents differ in practice, when we route tasks to each, and why the comparison is usually a false choice.
What is a forward-deployed engineer?
A forward-deployed engineer (FDE) works embedded with a customer, adapting a product to that customer’s environment, data, and workflows. Where the role came from, how it differs from solutions engineering and consulting, and why AI products keep hiring for it.
The agentic SDLC, and the stage everyone skips
The agentic SDLC, sometimes called the ADLC, redraws the software development lifecycle around coding agents. Most versions automate every stage except the one that makes the rest safe: verification. The argument for a verification-first ADLC.
What is agentic coding?
Agentic coding is delegating implementation to an AI agent that plans, edits, runs, and iterates on real code while you own the outcome. How it differs from autocomplete and from vibe coding, what changes in a team’s workflow, and where it breaks.
What is a background coding agent?
A background coding agent works on tasks in an isolated environment, without a person watching, and returns a finished pull request. How the category differs from IDE assistants, who the players are, and why the sandbox matters more than the model.
The AI software factory
An AI software factory turns a backlog into verified pull requests: coding agents implement in isolated sandboxes, checks and recorded browser QA run on every change, humans review the output. What separates a factory from an agent with commit access, and how to run one on your own infrastructure.
What is a software factory?
A software factory applies manufacturing discipline to shipping code: standardized pipelines, quality gates at every station, measurable throughput. Where the term comes from, how it differs from DevOps, and what changes now that coding agents do the implementation.