Source: https://internal-agents.com/agents/strongdm-software-factory # StrongDM — Software Factory An internal agent system in which natural-language seeds and end-to-end scenario harnesses drive agents that write, validate, and iterate on code under standing rules that humans neither write nor review code. - Approach type: Agent system - Deployment stage: Deployed - Autonomy: Autonomous - Evidence strength: Detailed primary - Status: Mixed - First reported year: 2026 - Work: Coding - Invocation: Background - Entry reviewed: 2026-09-15 ## Where people stay involved - **natural-language seed -> agents implement and validate against scenario harness -> converged software** — Exception only · Level 5 ## Overview ### Summary An internal agent system in which natural-language seeds and end-to-end scenario harnesses drive agents that write, validate, and iterate on code under standing rules that humans neither write nor review code. Fact · Reported · High confidence · `strongdm-software-factory--summary` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Definition: non-interactive development where specs + scenarios drive agents that write code, run harnesses, and converge without human review - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Rules: Code must not be written by humans; Code must not be reviewed by humans - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · Validation constraint: zero hand-written code and zero traditional review ## How it works ### Supporting component End-to-end user-story validations, often stored outside the codebase like a training holdout set, flexibly checked by an LLM to resist reward hacking and lazy test rewriting Fact · Reported · High confidence · `strongdm-software-factory--primitives-0` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Scenarios section ### Supporting component Probabilistic success measure replacing boolean test-green; the fraction of all observed scenario trajectories that likely satisfy the user Fact · Reported · High confidence · `strongdm-software-factory--primitives-1` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Satisfaction section ### Supporting component Behavioral clones of third-party services that replicate APIs, edge cases, and observable behaviors, so thousands of scenarios can run per hour without rate limits, abuse detection, or live-API cost Fact · Reported · High confidence · `strongdm-software-factory--primitives-2` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Digital Twin Universe section - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · Digital Twin Universe technique ### Supporting component Splitting work into interactive and fully specified classes; when specs, tests, and the existing app fully express intent, an agent runs end-to-end without back-and-forth Fact · Reported · High confidence · `strongdm-software-factory--primitives-3` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · Shift Work technique ### Supporting component Non-interactive coding agent structured as a graph of work phases with natural-language edges evaluated by the LLM; execution traverses the graph until convergence or termination, and the spec is open source Fact · Reported · High confidence · `strongdm-software-factory--primitives-4` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [5] [Attractor](https://factory.strongdm.ai/products/attractor) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-attractor-spec/content.md) · Graph structure and key properties - Supports · [7] [strongdm/attractor](https://github.com/strongdm/attractor) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-attractor-repo/content.md) · Repository description ## Supervision evidence ### Operating model assessment Level 5 for natural-language seed -> agents implement and validate against scenario harness -> converged software; human attention boundary: exception-only. Inference · Catalog judgment · Medium confidence · `strongdm-software-factory--operating-models-0` Confidence reason: The site documents zero human code-writing and review with automatic scenario validation, but not what happens when scenarios fail to converge; human authorship of seeds and scenarios sits before the run itself. Qualifications: - Observation date: 2026-09 Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Agents that write code, run harnesses, and converge without human review - Supports · [2] [Principles](https://factory.strongdm.ai/principles) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-principles/content.md) · Termination: the loop runs until the holdout scenarios pass (and stay passing) - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · Grow from cascades of natural-language specifications; be validated automatically without semantic inspection of source ## Implementation details ### Sandbox unknown Inference · Catalog judgment · Medium confidence · `strongdm-software-factory--architecture-sandbox` Confidence reason: The preserved sources do not document an execution sandbox; unknown does not mean absent. Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · The overview describes the workflow end to end without documenting an execution sandbox ### Harness End-to-end scenario harness with user-story-style scenarios kept outside the codebase as a holdout set and checked by an LLM; satisfaction measured as the fraction of observed scenario trajectories that likely satisfy the user; Digital Twin Universe clones of Okta, Jira, Slack, Google Docs, Google Drive, and Google Sheets absorb scenario load Fact · Reported · High confidence · `strongdm-software-factory--architecture-harness` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Scenarios, Satisfaction, and Digital Twin Universe sections - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · Digital Twin Universe: validate at volumes and rates far exceeding production limits ### Model Multi-provider routing by task as of 2026-06 (gpt-5.5 for everyday and DevOps tasks, QA orchestration, and security review; consensus of opus-4.8 and gpt-5.5 for sprint planning; opus-4.8 for frontend aesthetics and writing; gemini-3-flash-preview for image comprehension and agentic dialogues; gpt-image-2 and gpt-realtime-2 for UX ideation and voice) Fact · Reported · High confidence · `strongdm-software-factory--architecture-model` Confidence reason: First-party configuration log current as of 2026-06-04. Qualifications: - Observation date: 2026-06 Evidence: - Supports · [6] [Weather Report](https://factory.strongdm.ai/weather-report) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-weather-report/content.md) · Current configuration table, 2026-06-04 ### Tool access Filesystem that agents read and write to self-manage context; Digital Twin Universe as a stand-in for third-party services Fact · Reported · High confidence · `strongdm-software-factory--architecture-tool-access` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · The Filesystem; Digital Twin Universe ### Knowledge CXDB, a self-hosted context store for AI agents with turn DAG, blob deduplication, dynamic types, and visual debugging Fact · Reported · High confidence · `strongdm-software-factory--architecture-knowledge` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [4] [Products](https://factory.strongdm.ai/products) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-products/content.md) · CXDB entry ### Credentials StrongDM ID, an identity system for humans, workloads, and AI agents with federated authentication and path-scoped sharing Fact · Reported · High confidence · `strongdm-software-factory--architecture-credentials` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [4] [Products](https://factory.strongdm.ai/products) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-products/content.md) · StrongDM ID entry ### Context management Pyramid Summaries with reversible summarization at multiple zoom levels; on-disk state as the memory substrate Fact · Reported · High confidence · `strongdm-software-factory--architecture-context-mgmt` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · Pyramid Summaries; The Filesystem ## Lessons and interpretation ### Lesson Treat generated code like an ML model snapshot; judge it only by externally observable behavior and never by semantic inspection of the source Opinion · Reported · High confidence · `strongdm-software-factory--lessons-learned-0` Confidence reason: The team states this stance directly on the Techniques page. Evidence: - Supports · [3] [Techniques](https://factory.strongdm.ai/techniques) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-techniques/content.md) · Code treated like an ML model snapshot, opaque and judged by externally observable behavior ### Lesson The seed, validation harness, and feedback loop run until the holdout scenarios pass and stay passing Opinion · Reported · High confidence · `strongdm-software-factory--lessons-learned-1` Confidence reason: The team states the termination condition directly on the Principles page. Evidence: - Supports · [2] [Principles](https://factory.strongdm.ai/principles) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-principles/content.md) · Termination condition ### Lesson Tokens are the fuel; if the factory has not spent at least $1,000 on tokens per human engineer in a day, it has room for improvement Opinion · Reported · High confidence · `strongdm-software-factory--lessons-learned-2` Confidence reason: A stated benchmark for factory operation, not a measured result. Qualifications: - Observation date: 2026-02 Evidence: - Supports · [1] [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/strongdm-factory-overview/content.md) · Token spend benchmark ## Sources 1. [Software Factories And The Agentic Moment](https://factory.strongdm.ai/) - Engineering blog · First party · Evidence - Original URL: - Publisher: StrongDM · Published: 2026-02-06 · Accessed: 2026-09-15 · Last verified: 2026-09-15 - Preserved copy in the repository: 2. [Principles](https://factory.strongdm.ai/principles) - Engineering blog · First party · Evidence - Original URL: - Publisher: StrongDM · Accessed: 2026-09-15 · Last verified: 2026-09-15 - Preserved copy in the repository: 3. [Techniques](https://factory.strongdm.ai/techniques) - Engineering blog · First party · Evidence - Original URL: - Publisher: StrongDM · Accessed: 2026-09-15 · Last verified: 2026-09-15 - Preserved copy in the repository: 4. [Products](https://factory.strongdm.ai/products) - Documentation · First party · Evidence - Original URL: - Publisher: StrongDM · Accessed: 2026-09-15 · Last verified: 2026-09-15 - Preserved copy in the repository: 5. [Attractor](https://factory.strongdm.ai/products/attractor) - Documentation · First party · Evidence - Original URL: - Publisher: StrongDM · Accessed: 2026-09-15 · Last verified: 2026-09-15 - Preserved copy in the repository: 6. [Weather Report](https://factory.strongdm.ai/weather-report) - Documentation · First party · Evidence - Original URL: - Publisher: StrongDM · Accessed: 2026-09-15 · Last verified: 2026-09-15 - Preserved copy in the repository: 7. [strongdm/attractor](https://github.com/strongdm/attractor) - Repository · First party · Evidence - Original URL: - Publisher: StrongDM · Accessed: 2026-09-15 · Last verified: 2026-09-15 - Preserved copy in the repository: