Proof standard
The exact task, artifact, acceptance authority, correction record, delivered surface, and rejected attempts remain visible.
Visualization practitioners
Use assistance without giving up intent or control.
For analysts, designers, developers, and graphics practitioners deciding what to delegate, what to inspect, and when to stop or revert.
Decision to support
Where does AI reduce real work, and which parts still require human judgment, correction, and acceptance?
Working output
A bounded workflow plan naming what to delegate, what to inspect, who accepts it, and when to stop or revert.
Composed research
6 source sections, assembled for this decision.
These are the original sections from the visual research experiences, not summaries or reference cards. Expand each step to read it here, or open its stable page for direct linking.
01 · PracticeThe jobsWhat work is hidden inside make a chart?
Name the whole job and its acceptance point before choosing a tool.
Open this section as its own page →02 · PracticeThe tools and work surfacesWhat can each work surface see, change, and leave behind?
Match the surface to the visibility, editability, and recovery the job needs.
Open this section as its own page →03 · Capability modelCurrent common-core scorecardWhere is the field closest to and farthest from the ideal?
Set a realistic boundary for what assistance can currently carry.
Open this section as its own page →04 · Human evidenceCreator experienceHow does assistance change the full creator workflow?
Count inspection, repair, discarded attempts, and retained control as work.
Open this section as its own page →05 · EvaluationA practical evaluationHow should the complete creator-to-artifact-to-reader episode be measured?
Freeze the acceptance contract and test the rendered result.
Open this section as its own page →06 · ContextContext and lifecycle modesHow should the common evidence spine change by role and lifecycle state?
Keep delivery, failure, repair, and maintenance connected after release.
Open this section as its own page →Ask before acting
Three questions that keep the route honest.
- 01
Can I state the claim, reader, stakes, and acceptance authority before generating anything?
- 02
Can I inspect and repair the data, semantics, design, and rendered behavior without starting over?
- 03
Will rejected attempts, final acceptance, and later changes remain attributable?
Focused analysis
Complete arguments to keep nearby.
- Human skills and banked gainsSeparate work that becomes less mechanical from judgment that becomes more consequential.
Named evidence
Reference pages connected to this route.
- DashArenaA benchmark, browser executor, and human-calibrated judge for open-ended interactive dashboard generation.
- Dashboard2CodeA benchmark for reconstructing interactive Plotly Dash applications from screenshots, optional DOM, and interaction.
- Text2VisAn end-to-end benchmark combining data, a question, an answer, chart code, and annotated visual evidence.