# Agentic Engineering Field Curriculum **Edition 2026.08 · Reviewed 2026-08-10** A source-disciplined learning system of four published paths connecting agentic engineering concepts, practitioner perspectives, implementation patterns, exercises, independent verification, and evidence. > Learn the system. Practice the judgment. Produce the evidence. ## Published paths 4 published paths carry 26 finite stages. Every stage ends in a field exercise, an independent challenge, and named evidence. ### Engineering with agents without losing engineering Move from model-assisted coding to a bounded, inspectable, independently challenged delivery system. Level: practitioner · Designed for Engineer, Architect, Platform · Estimated study: 9 hours 01. **See the whole system:** Describe an agent as a bounded system of model, loop, context, tools, harness, substrate, identity, verification, and human authority. 02. **Align before implementation:** Turn an ambiguous request into durable intent, explicit decisions, acceptance signals, and a bounded specification without treating the specification as a substitute for code comprehension. 03. **Shape an agent-readable codebase:** Design stable seams, compact interfaces, local feedback, and durable repository context that both humans and agents can navigate. 04. **Slice for fast feedback:** Turn a durable destination into vertically sliced, dependency-aware work that fits bounded contexts and exposes feedback early. 05. **Build against evidence:** Design deterministic tests, evaluations, typed outputs, and outcome checks that let construction advance in small verified increments. 06. **Verify independently:** Design an independent verifier with named ownership, trigger, scope, block authority, method, and recorded disposition. 07. **Scale without losing control:** Balance generation with product decisions, verifier capacity, integration, evidence, runtime intervention, and named human release authority. Full path: [Engineering with agents without losing engineering](https://aisdlc.ai/agentic-engineering/learn/agentic-engineer-core) · [Markdown edition](https://aisdlc.ai/agentic-engineering/learn/agentic-engineer-core/index.md) ### Private AI and owned inference Choose deliberately among hosted, dedicated, local, self-hosted, adapted, and trained model paths without treating deployment location as a complete privacy claim. Level: advanced · Designed for Engineer, Architect, Platform, Risk · Estimated study: 9 hours 01. **Trace where the data comes to rest:** Trace every place prompts, retrieved context, memory, tool arguments, telemetry, backups, identities, and operator access carry data across a boundary. 02. **Choose how much of the stack to own:** Compare hosted APIs, dedicated endpoints, self-hosted open weights, adapted models, and pretraining against a written constraint rather than a preference. 03. **Run inference on hardware you control:** Match model format, quantization, memory, context length, and latency to a workload you measure on hardware you control. 04. **Serve it like production infrastructure:** Engineer capacity, isolation, routing, patching, observability, and failure recovery for an inference endpoint you own. 05. **Adapt only with evidence:** Choose among prompting, retrieval, parameter-efficient tuning, full fine-tuning, and pretraining from a reproduced failure rather than an assumption. 06. **Govern the model lifecycle:** Track license, provenance, evaluation, vulnerability, approval, recertification, and retirement for every model a system calls. Full path: [Private AI and owned inference](https://aisdlc.ai/agentic-engineering/learn/private-ai-owned-inference) · [Markdown edition](https://aisdlc.ai/agentic-engineering/learn/private-ai-owned-inference/index.md) ### Agent platform engineering Specify the identity, substrate, context, tool, orchestration, verification, evidence, and lifecycle services that bounded agents require, as contracts an independent reviewer can challenge. Level: advanced · Designed for Architect, Platform, Risk, Operations · Estimated study: 10 hours 01. **Specify the execution substrate:** Specify isolated workspaces, scoped credentials, controlled egress, and reproducible recovery for every agent run. 02. **Give every agent standing:** Design an identity record that binds each agent to a purpose, a sponsor, an accountable owner, scoped permissions, and an expiry that removes standing by default. 03. **Version context and memory:** Design context assembly so instructions, retrieval, and working state carry a version, a provenance record, a trust class, and a retention rule. 04. **Draw the tool boundary:** Design typed tool contracts that separate inspect, propose, execute, and commit capability, each with its own authorization and audit record. 05. **Design durable runs:** Design multi-step agent work with durable state, bounded retries, explicit handoffs, spend and time budgets, and a finite set of terminal dispositions. 06. **Design the verifier plane:** Design independently owned preventive, inline, gate, and continuous checks so that authority to stop the work exists before the point of consequence. 07. **Close the evidence and lifecycle loop:** Design one addressable record that links runs, findings, approvals, releases, incidents, recertification, and retirement. Full path: [Agent platform engineering](https://aisdlc.ai/agentic-engineering/learn/agent-platform-engineering) · [Markdown edition](https://aisdlc.ai/agentic-engineering/learn/agent-platform-engineering/index.md) ### Lead AI-native delivery Redesign the operating system around higher generation speed while protecting strategic judgment, independent assurance, and outcome ownership. Level: practitioner · Designed for Executive, Engineer, Risk, Operations · Estimated study: 9 hours 01. **Measure the system, not the activity:** Build a measurement set that separates local generation activity from delivered outcomes, stability, learning, and human cognitive load. 02. **Find the absorption constraint:** Map the decision, review, verification, integration, release, and operating queues that determine how much generated work becomes delivered value. 03. **Redesign work and attention:** Set batch size, work limits, decision packets, and protected review time from measured capacity so human authority stays real under higher generation speed. 04. **Scale verification and evidence with generation:** Plan independent verification and release evidence that grows with generation capacity, with named ownership and a stop condition when capacity is exceeded. 05. **Bind governance to execution:** Connect delegated standing, policy decisions, intervention, and named disposition so written rules reach the point where an action can be refused, held, or escalated. 06. **Close the loop from production:** Turn production outcomes, incidents, corrections, and drift into revised intent, durable evaluation cases, and reopened autonomy decisions. Full path: [Lead AI-native delivery](https://aisdlc.ai/agentic-engineering/learn/lead-ai-native-delivery) · [Markdown edition](https://aisdlc.ai/agentic-engineering/learn/lead-ai-native-delivery/index.md) ## Independent practitioner sources Practitioner inclusion identifies a relevant independent source. It does not imply contribution, partnership, endorsement, sponsorship, or approval of AISDLC unless a separate public agreement explicitly says so. ### Matt Pocock Developer educator and creator of AI Hero and Total TypeScript. Relationship: independent-source. - [“Software Fundamentals Matter More Than Ever” — Matt Pocock](https://www.youtube.com/watch?v=v4F1gFy-hqg) — Practitioner perspective. This is an attributed conference position, not a controlled study, consensus standard, universal productivity law, or enterprise release method. - [mattpocock/skills: A complete AI Coding workflow, end-to-end](https://www.youtube.com/watch?v=M6mYodf0dJM) — Practitioner perspective. This demonstrates one practitioner workflow and tool ecosystem; it does not establish independent verification, governed authority, or fitness for every team and codebase. - [/wayfinder: Nothing is too big to plan anymore](https://www.youtube.com/watch?v=F3lL98Pj90o) — Practitioner perspective. Wayfinder is an evolving practitioner method. Its planning map does not replace accountable product decisions, architecture review, implementation controls, or release evidence. - [Software Design playlist](https://www.youtube.com/playlist?list=PLLjGHW6KybtUXPYeFEBTM6pEVKemaGXXT) — Practitioner perspective. Matt Pocock is featured in the collection but is not represented as its curator or publisher. A playlist is a discovery surface, not evidence for a claim; individual items require review and attribution. - [Skills for Real Engineers](https://github.com/mattpocock/skills) — Open-source implementation reference. The repository is a practitioner implementation, not an enterprise control standard. Substantial reuse must preserve its MIT copyright and permission notice. - [AI Hero](https://www.aihero.dev/) — Practitioner home. Public descriptions can establish attribution and available offerings, but paid lessons, course exercises, visual assets, and brand identity are not AISDLC content. Matt Pocock is included as an independent practitioner source. AISDLC owns this synthesis and does not represent him as a contributor, partner, endorser, or approver. ## Boundaries This curriculum is original AISDLC editorial synthesis informed by cited research, official documentation, standards, open-source references, and attributed practitioner perspectives. It is learning material—not a standard, production approval, compliance determination, or proof that a depicted control is implemented. Progress and completion are self-directed learning records. They are not certification, professional standing, production evidence, or authorization to deploy an agentic system.