← All implementations

StrongDM · Agent system

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.

1 Supports1 Supports3 Supports

Approach type
Agent system
Work
Coding
Human involvement
Autonomous
Invocation
Background
Deployment stage
Deployed
Evidence strength
Detailed primary
Entry reviewed

How it works

The workflow the sources report for this implementation.

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

1

Probabilistic success measure replacing boolean test-green; the fraction of all observed scenario trajectories that likely satisfy the user

1

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

1 Supports3 Supports

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

3

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

5 Supports7 Supports

Where people stay involved

  • natural-language seed -> agents implement and validate against scenario harness -> converged softwareException only · Level 5

Level 5 for natural-language seed -> agents implement and validate against scenario harness -> converged software; human attention boundary: exception-only.

1 Supports2 Supports3 Supports

Observation date
2026-09

Implementation details

Sandbox

unknown

1

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

1 Supports3 Supports

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)

6

Observation date
2026-06
Tool access

Filesystem that agents read and write to self-manage context; Digital Twin Universe as a stand-in for third-party services

3

Knowledge

CXDB, a self-hosted context store for AI agents with turn DAG, blob deduplication, dynamic types, and visual debugging

4

Credentials

StrongDM ID, an identity system for humans, workloads, and AI agents with federated authentication and path-scoped sharing

4

Context management

Pyramid Summaries with reversible summarization at multiple zoom levels; on-disk state as the memory substrate

3

Reported results and limitations

The catalog records what the sources report, with the scope and the denominator of every figure. A qualification below limits the figure it sits under.

Lessons and interpretation

Treat generated code like an ML model snapshot; judge it only by externally observable behavior and never by semantic inspection of the source

3

The seed, validation harness, and feedback loop run until the holdout scenarios pass and stay passing

2

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

1

Observation date
2026-02

Sources and research details

Citations link to the original publisher. Each source also keeps a preserved copy in the repository, so a changed or removed page stays checkable.

  1. Software Factories And The Agentic Momenthttps://factory.strongdm.ai/Engineering blog · First party · Last source verification: 2026-09-15
    Preserved copy ↗
  2. Principleshttps://factory.strongdm.ai/principlesEngineering blog · First party · Last source verification: 2026-09-15
    Preserved copy ↗
  3. Techniqueshttps://factory.strongdm.ai/techniquesEngineering blog · First party · Last source verification: 2026-09-15
    Preserved copy ↗
  4. Productshttps://factory.strongdm.ai/productsDocumentation · First party · Last source verification: 2026-09-15
    Preserved copy ↗
  5. Attractorhttps://factory.strongdm.ai/products/attractorDocumentation · First party · Last source verification: 2026-09-15
    Preserved copy ↗
  6. Weather Reporthttps://factory.strongdm.ai/weather-reportDocumentation · First party · Last source verification: 2026-09-15
    Preserved copy ↗
  7. strongdm/attractorhttps://github.com/strongdm/attractorRepository · First party · Last source verification: 2026-09-15
    Preserved copy ↗
Research details for every claim on this page
  1. Summary
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  2. Sandbox
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    Confidence reason
    The preserved sources do not document an execution sandbox; unknown does not mean absent.
  3. Harness
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  4. Model
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    First-party configuration log current as of 2026-06-04.
    Observation date
    2026-06
  5. Tool access
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
    • SupportsTechniquesThe Filesystem; Digital Twin Universe
  6. Knowledge
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  7. Credentials
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  8. Context management
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
    • SupportsTechniquesPyramid Summaries; The Filesystem
  9. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  10. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  11. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  12. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  13. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  14. Lesson
    Statement type
    Opinion
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    The team states this stance directly on the Techniques page.
    • SupportsTechniquesCode treated like an ML model snapshot, opaque and judged by externally observable behavior
  15. Lesson
    Statement type
    Opinion
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    The team states the termination condition directly on the Principles page.
  16. Lesson
    Statement type
    Opinion
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A stated benchmark for factory operation, not a measured result.
    Observation date
    2026-02
  17. Operating model assessment
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    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.
    Observation date
    2026-09
    • SupportsSoftware Factories And The Agentic MomentAgents that write code, run harnesses, and converge without human review
    • SupportsPrinciplesTermination: the loop runs until the holdout scenarios pass (and stay passing)
    • SupportsTechniquesGrow from cascades of natural-language specifications; be validated automatically without semantic inspection of source

The catalog records no related implementation for this entry yet.

Back to the directory →