Internal Agents Map

Internal coding agent (unnamed)

Uber runs an internal coding agent that writes complete code changes with no human authoring, and engineers review and approve them. Its CTO reports that engineers now delegate tasks to the agent rather than accept its suggestions, and that its share of all code changes has grown from under 1%.

Company
Uber
Approach type
Agent
Work
Coding
Human involvement
Drafts reviewed
Invocation
Interactive
Deployment stage
Deployed
Evidence strength
Secondary only
Entry reviewed
Copy as MDAsk ChatGPT

How it works

Delegated coding task to engineer-approved code change.

Delegate the task

Engineers delegate coding tasks to the agent rather than accepting its suggestions[1]

Write the complete change

The agent writes the entire code change with zero human authoring[1]

Review and approve

An engineer reviews and approves the agent-written change[1]

Where people stay involved

People review, approve, or guide the agent’s work here.

Reported: Engineers review and approve every agent-written change; the CTO describes zero human authoring.

  • Supervision level

    Work-product review

    A person reviews the draft or implementation.

    • coding request → complete code change[1]

How supervision levels work

Implementation details

Model
Not reported
Harness
Not reported
Sandbox
Not reported
Tool access
Not reported
Knowledge
Not reported
Context management
Not reported
Credentials
Not reported
Interfaces
Not reported

Validation and failure handling

Reported: Review and approve

Reported observations

These results come from the sources. Notes below each result explain what it covers and any reported limits.

Headline claim

~1,800 complete code changes per week (~8% of changes)[1]

Reported by
Uber
Scope
Code changes written entirely by the internal coding agent and human reviewed
Denominator
All Uber code changes for the reported 8% share
About this result
Category
Adoption output
Evidence type
Reported measurement
What it covers
Weekly code changes written entirely by the internal coding agent, and their share of all Uber code changes
Key observation

95% of engineers use AI tools monthly[1]

Reported by
Uber
Scope
Monthly use of AI tools among Uber engineers; not internal-agent-specific adoption
Denominator
Uber engineers
About this result
Category
Adoption output
Evidence type
Reported measurement
What it covers
Monthly AI-tool use among Uber engineers generally, not specific to this agent

Lessons

Reported opinion: Uber's CTO describes the engineer's role shifting from writing every line to architecting systems and reviewing AI-generated code.[1]

Reported opinion: Uber's CTO attributes the strongest adoption to engineers quietly experimenting, not to a top-down push.[1]

Back to agents
Improve this record

Give your coding agent the context and instructions to prepare a source-backed update.

Suggest on GitHub

Paste the prompt into your coding agent, then describe the correction or add a public source.

Preview prompt

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. Uber's CTO on AI coding agents (Business Insider)https://www.businessinsider.com/uber-cto-ai-coding-agentic-software-engineers-2026-3News · Independent secondary · Last source verification: 2026-08-31
    Preserved copy
Question coverage and scope
purpose
Reported
workflow
Reported: The three steps carry the full delegation-to-approval arc the article describes; no other stage is documented.
human involvement
Reported: Engineers review and approve every agent-written change; the CTO describes zero human authoring.
implementation
Not reported
validation
Reported: The documented check of the work is the human review gate cited with human involvement; no automated check is described.
observations
Reported
lessons
Reported: Two attributed opinions of the CTO, relayed by a single secondary article.
Research details for every claim on this page
  1. Summary
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    Only limited public evidence supports this claim.
  2. Headline claim
    Statement type
    Metric
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    Only limited public evidence supports this claim.
    Reported by
    Uber
    Scope
    Code changes written entirely by the internal coding agent and human reviewed
    Denominator
    All Uber code changes for the reported 8% share
    Method
    Not reported
    Observation date
    Not reported
  3. Delegate the task
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    A single secondary article relays the CTO's statement about delegation versus suggestions.
  4. Write the complete change
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    The CTO's zero-human-authoring statement is relayed verbatim by one secondary article.
  5. Review and approve
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    One secondary article relays the CTO's review-and-approve description without naming the review venue or policy.
  6. Key observation

    ~1,800 complete code changes per week (~8% of changes at the time)

    Duplicate representation of Headline claim. Same weekly count and 8% share from the same passage; the at-the-time wording only makes explicit the point-in-time framing the headline growth statement already carries, so it adds no separate information.

    Statement type
    Metric
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    Only limited public evidence supports this claim.
    Reported by
    Uber
    Scope
    Code changes written entirely by the internal coding agent and human reviewed
    Denominator
    All Uber code changes for the reported 8% share
    Method
    Not reported
    Observation date
    Not reported
  7. Key observation
    Statement type
    Metric
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    Only limited public evidence supports this claim.
    Reported by
    Uber
    Scope
    Monthly use of AI tools among Uber engineers; not internal-agent-specific adoption
    Denominator
    Uber engineers
    Method
    Not reported
    Observation date
    Not reported
  8. Lesson
    Statement type
    Opinion
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    An attributed view of one CTO on the engineer's role, relayed by a single secondary article.
  9. Lesson
    Statement type
    Opinion
    Provenance
    Reported
    Confidence
    Medium
    Confidence reason
    An attributed adoption observation of one CTO; no adoption method or measurement is described.
  10. Operating model assessment
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    Confidence reason
    The CTO states there is zero human authoring and that engineers review and approve the agent-written code, which locates human attention at review of the work product.
    Observation date
    2026