Quick Reference: Domain 1 — Solution Design & Architecture

The professional sequence, in order: quantify the problem → capture constraints → define the success criterion → choose the solution shape. Inverting any pair produces designs that fit the technology rather than the problem.

• Success criterion = metric + baseline + target, stated before the build. No criterion, no architecture — just a demo.
• The fit test: language understanding, synthesis, or judgement over unstructured input → Claude. Fully rule-specified (tax, validation, thresholds) → deterministic code. Most real systems are hybrids; drawing the boundary is the architect's job.
• Task shapes: classification, extraction, summarisation, generation, conversation, agentic action. Each implies its own architecture, validation, and evaluation.
• Scope narrow first. Target the measured cost driver, prove value, expand. The general-purpose assistant with no baseline is a planted wrong answer.

Input → Context assembly → Processing → Output → Delivery → Feedback loops.

• Model output is untrusted until validated. Schema validity ≠ business validity.
• Design for failure: timeouts, retries with backoff, idempotency, and an explicit degraded path. Silent failure is the one forbidden outcome.
• Blast radius determines validation and review depth: wrong internal draft ≠ wrong customer refund.
• "Quality is inconsistent and nobody knows why" = missing feedback loop, not a weak model.

Three tiers: augmented LLM (one call + retrieval/tools/memory) → workflows (predefined code paths) → agents (model directs its own process). Escalate only when the simpler tier fails a named requirement.

The exam punishes over-engineering as hard as under-engineering. Fixed steps + audit requirement → workflow, never agent.

• Justified by: breadth-first parallel exploration, tasks exceeding one context window, distinct specialisms. Not by ambition.
• Cost: agents ≈ 4× chat tokens; multi-agent ≈ 15×. The value case must clear the multiplier.
• Subagents share no memory. No inherited conversation, no peer visibility. All context passes explicitly in the brief.
• A complete brief: objective + output format (with claim–source fields) + tool guidance + scope boundaries.
• Failure signatures: duplication + gaps = unpartitioned coordinator decomposition. Fix the briefs, not the agent count or model size.
• Long runs need: checkpointed state (resume, not restart) and per-agent decision observability.

• Fixed (design-time pipeline) when steps are known; dynamic (orchestrator/agent) when structure emerges during the work.
• Four axes: by stage, by section, by perspective, by specialism.
• Attention dilution: one pass over too much material → inconsistent depth, contradictory judgements on identical patterns. Fix = per-unit local passes + a cross-unit integration pass. Both halves mandatory. A bigger context window is the planted distractor — it solves fit, not attention.
• Interfaces: structured intermediate outputs (schemas with source locations) + acceptance gates between stages. Prose interfaces lose attribution and testability.
• Don't decompose a task that fits one call with no dilution symptoms; every split adds latency, cost, and error surface.

Efficiency · Transformation · Productivity · Cost · Performance SLAs

• Efficiency vs transformation discriminator: did the capability already exist? Faster existing process = efficiency. Newly possible = transformation.
• Productivity attaches to people assisted, not processes automated away.
• Model selection = per-task cost/performance decision against your own benchmarks; smallest model that meets the bar.
• Prompt caching: ~90% off cache reads + lower time-to-first-token. Serves cost AND latency SLA. Request-path lever.
• Batch processing: ~50% off asynchronous volume. Off-path lever. Distractors swap the two.
• SLA engineering: decompose p95 into per-stage budgets (a chain that doesn't sum inside the target changes); availability targets force designed fallbacks; quality SLAs need evaluation machinery.
• Value case: baseline before build; hypothesis (pillar, metric, target, measurement) stated up front and verified after.