Worked Examples

Each row shows the base tier, the escalators that fired and the result. Use them to calibrate borderline calls.

Technical

Task Result Deciding signal
“fix typo in README” 🟢 Haiku Trivial single edit, no reasoning
“convert PDF to markdown” 🟢 Haiku Simple conversion, no decisions
“debug flaky integration test” 🟡 Sonnet Single-system debugging, moderate reasoning
“refactor auth across 15 files” 🔴 Opus Scope escalator on top of architectural decisions
“design database schema for e-commerce” 🔴 Opus Architecture with trade-offs and long-term impact
“plan microservices migration strategy” 🔴 Opus Complex architectural planning plus Irreversibility

Business and strategy

Task Result Deciding signal
“summarise this meeting transcript” 🟢 Haiku Text transformation under 2K words
“extract action items from notes” 🟢 Haiku Extraction, no reasoning
“write blog post about AI trends” 🟡 Sonnet Creative writing, moderate reasoning
“draft sales proposal for enterprise client” 🟡 Sonnet Persuasive writing, moderate reasoning
“analyse competitor pricing strategy” 🟡 Sonnet Research with a single framework
“plan market entry strategy for Europe” 🔴 Opus Irreversibility, Synthesis and Ambiguity
“design organisational restructuring plan” 🔴 Opus Sensitivity, Stakeholders and Irreversibility
“M&A due diligence analysis” 🔴 Opus Stakes and Synthesis
“crisis communication plan for a data breach” 🔴 Opus Sensitivity, Stakes and Stakeholders

Creative and analysis

Task Result Deciding signal
“translate paragraph to French” 🟢 Haiku Language transform, no reasoning
“brainstorm product names (single session)” 🟡 Sonnet Creative generation, one framework
“brainstorm with SCAMPER plus trade-off analysis” 🔴 Opus Multi-framework
“retrospect: analyse collaboration patterns” 🔴 Opus Pattern detection plus Bias identification
“identify blind spots in strategy” 🔴 Opus Bias identification plus Ethical reasoning
“plan 3-day conference with speakers” 🔴 Opus Many constraints and Stakeholders

Commands, skills and agents

Task Result Deciding signal
“command: convert EPUB to markdown” 🟢 Haiku Simple workflow, minimal reasoning
“command: save session context” 🟡 Sonnet Context management, serialisation logic
“command: brainstorm with research plus SCAMPER” 🔴 Opus Multi-framework plus strategic analysis
“command: retrospect domain learnings” 🔴 Opus Pattern detection across sessions plus Bias identification
“skill: format code with prettier” 🟢 Haiku Deterministic task
“skill: standard workflow implementation” 🟡 Sonnet Standard workflow, moderate reasoning
“agent: explore codebase architecture” 🔴 Opus Exploration plus architectural synthesis

Cases where the escalator decides

Task: "fix a typo in the production payment config"
Base: Haiku (typo fix). Fired: Stakes (production, payments).
Result: Sonnet. One tier up, not Opus.
Task: "rename a function across 40 files with a codemod"
Base: Sonnet. Fired: none. Scope is one mechanical decision, not 40.
Result: Sonnet. Check the diff before trusting the codemod.
Task: "write a customer apology email after an outage"
Base: Sonnet. Fired: Sensitivity (customer-facing crisis).
Result: Opus. Sonnet is enough for the internal incident summary.