Good AI Task

AI compatibility

AI can scaffold the Apollo caching work, but hitting those performance targets needs a human in the loop.

Possible with caveats

Workable, but read the conditions.

Average across 1 submission.

58
avg / 100

The honest read

An AI code agent can meaningfully assist with Apollo Client caching patterns, fragment extraction, and query batching boilerplate — these are well-documented techniques with clear implementations. However, the specific 60% reduction and sub-2s startup targets require profiling real runtime behavior, understanding the app's actual data graph, and making judgment calls about cache normalization trade-offs that depend on live context the agent cannot access without deep integration into the dev environment. This is a strong AI-assist task, not a fully autonomous one.

Aggregated across 1 submission.

The five dimensions

Repeatability

Medium

Apollo caching patterns and fragment reuse are structurally repeatable, but the specific query waterfall problems across 8 screens require unique analysis of this app's data graph and component tree. Each screen's fix is a judgment call, not a template.

Ambiguity Tolerance

Low

Success criteria are quantified (60% fewer network calls, startup under 2s), which sounds crisp, but verifying those numbers requires runtime profiling tools and real device testing the agent cannot perform autonomously. The agent cannot confirm it's done without external measurement.

Data & Tool Availability

Low

The agent needs full access to the codebase, GraphQL schema, existing query definitions, and ideally network profiling traces — none of which are provided here. Without the actual code and schema, the agent can only produce generic patterns, not targeted fixes.

Error Cost

High

Incorrect cache normalization or aggressive batching can introduce stale data bugs, broken UI states, or race conditions that are subtle and hard to detect in QA. These errors can ship to production and degrade user experience in ways that are costly to diagnose and reverse.

Human Judgment Required

High

Deciding which queries to batch, how to normalize cache keys, and where fragment reuse is worth the abstraction cost requires understanding the product's data freshness requirements and team conventions. These are architectural decisions with long-term maintenance implications.

What an agent would need

  • Full read access to the React Native codebase including all 8 screens and their GraphQL query definitions
  • The complete GraphQL schema to understand entity relationships and design cache normalization keys
  • Network profiling traces or Apollo DevTools output showing current waterfall patterns and redundant fetches
  • Ability to run the app and measure startup time and network call counts before and after changes
  • Clear product requirements around data freshness to make safe decisions about cache TTLs and polling strategies

Best-matched agent type

Code Agent

The kind of agent this work would call for if it were a fit. For this task, it isn't.

Run your own fit check

Get a calibrated read on your specific task in under a minute.

Check a task