← All implementations

Zup · Task agent

CodeGen

CodeGen is Zup's internal coding agent, described in a paper by its builders. They report that targeted tool design, such as string-replacement edits over full-file rewrites, and layered safety guardrails mattered more than prompt engineering. Progressive human oversight drove adoption.

1

Approach type
Task agent
Work
Coding
Human involvement
Human in loop
Deployment stage
Research
Evidence strength
Detailed primary
Entry reviewed

How it works

The workflow the sources report for this implementation.

Tools that limit what the agent can change, vs. unconstrained generation

1

Levels of human review that increase trust over time

1

Where people stay involved

  • constrained coding task → human-supervised editContinuous steering · Level 2

Level 2 for constrained coding task → human-supervised edit; human attention boundary: continuous-steering.

1

Observation date
2026

Implementation details

Harness

Constrained editing tools; state-management concerns; progressive levels of human oversight

1

Model

Not specified (see paper)

1

Tool access

Constrained editing tools rather than free-form code generation

1

Credentials

Layered safety controls

1

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

Targeted tool design and layered safety controls matter more than prompt tweaks

1

Progressive levels of human oversight help build trust before granting autonomy

1

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. Building an Internal Coding Agent at Zuphttps://arxiv.org/abs/2604.09805Paper · First party · Last source verification: 2026-08-31
    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. Harness
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  3. Model
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  4. Tool access
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  5. Credentials
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  6. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  7. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  8. Lesson
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    Confidence reason
    The catalog derives this observation from the linked sources.
  9. Lesson
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    Confidence reason
    The catalog derives this observation from the linked sources.
  10. Operating model assessment
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    Confidence reason
    The research source describes progressive human oversight but does not establish background delegation.
    Observation date
    2026

The catalog records no related implementation for this entry yet.

Back to the directory →