Project Ops

Systems Thinking and AI Engineering — Historical Track Audit

Project operations source: Systems Thinking and AI Engineering — Historical Track Audit.

This is historical evidence, preserved on 2026-09-09 from source checkpoint caf2fe6. The matching Astro checkpoint is b91dfa3. The original outline below preserves planning inputs, not a second curriculum inventory or a commitment to build every named idea.

The current Track Guide and shared Track Map carry the approved composition: How Technology Changes the Ecology of Communication as signature, three systems routes, selected conversation and context Labs as shared practice, and optional technical and evaluation depth.

Next Artifact Pass — Evidence and Unresolved Questions

  • How Technology Changes the Ecology of Communication is Sketch / Planned, owner Unassigned, primary to AI Literacy and related to AI Readiness. Its folder contains only module.md. Its Communication Ecology Map is an embedded activity concept; the worksheet, example packet, standalone Activity, Notebook, and facilitator package have not been created. The module includes facilitator notes, but not a separate facilitator package. Keep the Systems signature assignment at Track level until reciprocal metadata receives its own review.
  • The nontechnical entry is a communication situation, not a prompt-engineering survey or architecture course. Preserve the module’s caution: earlier technology shifts can orient discussion but do not prove AI is ordinary, harmless, or destined to improve the situation.
  • First package gap: a reviewed low-risk scenario, a one-page map, a self-contained Activity and Notebook, and facilitator guidance. Test the short route only after these materials and learning design receive review; do not treat a proposed demonstration as an existing asset.
  • Conversation Engineering and Context Engineering remain Sketch / Planned, owned by Renaissance Solutions, primary to AI Literacy. Stay, Steer, or Start Over and Build the Brief Before the Work remain Beta, owned by Mike, primary to their respective modules and related to AI Literacy only. Systems placement is a candidate cross-Track relationship, not a metadata reassignment.
  • Both selected Labs have standalone Notebooks and learner/facilitator materials. Their current learning-design, accessibility, and pilot review needs remain open. The Track uses their Lab estimates of 8–12 and 30–40 minutes, not pilot evidence.
  • Context Engineering’s recommended estimate of 20–30 minutes differs from Build the Brief Before the Work’s 30–40 minutes. Context metadata names Conversation Engineering as a dependency while its learner page allows entry without required preparation. Preserve these sources and reconcile them during the module/Lab pass, not this composition pass.
  • Both shared Labs allow alternatives to an individual learner account, but their working steps include corrections or output comparisons. A facilitator must prepare an appropriate comparison set or live operator; a supplied initial scenario alone does not provide every later response. Sorting a brief alone is not completing a two-output comparison.
  • The old “build a context packet and verify whether it was used” idea is served in part by Build the Brief Before the Work. Its output comparison does not establish that a model used every source. Keep source selection, observed effects, model self-report, and claims about internal behavior separate.
  • Decision Ownership Before Delegation remains Seed / Planned, primary to AI Readiness, with Friction Mapping as its stated dependency. Its Decision Ownership Map and Delegation Ladder and Control Plan are candidate artifacts, not supplied Labs or worksheets. The broader permissions-review Lab idea is proposed; review overlap before creating another package.
  • Designing Systems That Can Admit They Are Wrong remains Seed / Deferred, primary to AI Readiness, with its dependencies unchanged. It is an ideation placeholder requiring research, governance, and learning-design review, not a ready system-design framework.
  • When the Average Lies: Simpson’s Paradox in AI Evaluation remains Sketch / Deferred and primary to Systems. Its imported PDFs, learner and instructor workbooks, and synthetic datasets are preserved unchanged. Statistical-methods, accessibility, learning-design, and pilot review remain open. Keep instructor answers separate; do not equate packet size with validated curriculum or synthetic data with evidence about real systems.
  • Memory and Retrieval, Observability, AI Runtimes, Tools and Agents, Designing AI Systems, and Deploying Version 2 remain proposed module directions without source pages. System decomposition, permissions/oversight review, and vendor-pattern comparison remain proposed Lab ideas without packages or settled build commitments.
  • The old Prompt Engineering idea is retained below, not turned into another core. The old Governance and Safety and Human-in-the-Loop Design ideas inform the authority and evaluation routes; no new modules are invented for those labels. Technical prerequisites from the outline become optional depth, not barriers to the nontechnical signature.
  • Paper mapping, discussion, supplied examples, and facilitator observation are valid. Live AI is optional for the signature, and no learner account, coding, connected system, deployment, or execution authority is required. Use public or fictional examples and distinguish live, observed, supplied, and hypothetical evidence.
  • The Track remains Draft / Drafting and Unassigned. Additional Reading remains unchanged. All module and Lab sources, dependencies, ownership, statuses, and review contracts remain unchanged.

Original Track Guide Snapshot

The snapshot retains its original relative links. Its AI Literacy dependency link now opens the neighboring historical audit; use the current AI Literacy Track Guide for today’s learner route.

Systems Thinking and AI Engineering — Original Outline

Status: Working track outline, not finished curriculum.

This advanced track builds from the published “from prompts to systems” framing. It teaches learners that AI engineering matures in layers: prompts, conversations, context, runtimes, memory, tools, agents, observability, governance, and trustworthy system design.

The invitation is explicit: this is not only for software engineers. Teachers, artists, project managers, healthcare professionals, business owners, students, retirees, and future consultants can all learn the patterns. Curiosity matters more than credentials. The point is to understand the layers, practice the habits, and find a responsible place to participate.

Audience: Curious advanced learners, managers, educators, builders, future consultants, technical contributors, and non-engineers who want to understand AI systems.

Purpose: Teach durable AI systems concepts that outlast vendor terminology.

Likely Modules

Starter Lab Ideas

  • Decompose an AI assistant into its system parts.
  • Compare vendor names to shared architectural patterns.
  • Build a context packet and verify whether it was used.
  • Review a system for permissions, memory, observability, and human oversight.

Additional Reading

No additional reading is listed yet.

Dependencies

  • AI Literacy for the Rest of Us.
  • Prompting is a conversation.
  • Context and knowledge-base basics.
  • Agentic literacy.
  • Trustworthy deployment principles.

MVP Holes

  • Decide whether this track is article-led, class-led, or lab-led first.
  • Define the minimum nontechnical path through the concepts.
  • Decide where AI runtime examples should come from without making the track vendor-specific.
  • Create a glossary that connects systems terms to beginner-friendly analogies.

Source: ops/track-audits/systems-thinking-and-ai-engineering.md