# Deployment Run — FDE learning game A forward deployed engineering game and field guide: practise customer discovery, integration, evaluation, incident recovery, and handoff. Take responsibility for a customer deployment. Inspect evidence, choose an approach, review its consequence, and leave a service the customer can operate. All names, messages, timings, and outcomes are fictional. [Play Deployment Run](https://fdelab.dev/#game) ## Your three customers ### The Monday queue Northstar Services — Dispatch operations Get repair requests to eligible technicians faster. Every assignment still needs a dispatcher’s approval. [Read the Northstar Services mission brief](https://fdelab.dev/missions/northstar/) ### The missing invoice Cedar Supply — Finance operations Reduce invoice matching work. The pilot may create draft matches; payments and bank-detail changes stay outside its authority. [Read the Cedar Supply mission brief](https://fdelab.dev/missions/cedar/) ### A very confident answer Harbor Engineering — Internal knowledge Help staff find the right procedure faster. Preserve source authority, current versions, and each caller’s access. [Read the Harbor Engineering mission brief](https://fdelab.dev/missions/harbor/) ## What a delivery must prove - Finish five decisions within 14 simulated days. There is no real-time timer. - End with customer trust of at least 60 and reliability of at least 70. - Recover the incident and demonstrate handoff to an owner and backup. - A smaller supported service or controlled pause can count as a successful delivery. - Only the best result for each mission contributes to campaign XP. [Read the full scoring and provenance notes](https://fdelab.dev/research/deployment-run.md) ## Continue learning - [Learn](https://fdelab.dev/learn/) - [Read](https://fdelab.dev/reading/) - [Missions](https://fdelab.dev/missions/) - [Resources](https://fdelab.dev/resources/) [Agent guide](https://fdelab.dev/agent-guide.md). Hosting access permissions apply to all resources. # How to become a forward deployed engineer Build one customer workflow from discovery through handoff. These original exercises ask for working evidence, not just completed reading. The milestones are self-assessed and do not guarantee hiring. [Use the interactive learning plan](https://fdelab.dev/workspace/#learn) - [Build a software foundation](https://fdelab.dev/learn/#foundation) - [Understand the workflow first](https://fdelab.dev/learn/#discovery) - [Connect data and systems](https://fdelab.dev/learn/#integration) - [Build an AI workflow you can evaluate](https://fdelab.dev/learn/#ai) - [Make it work after the demo](https://fdelab.dev/learn/#production) - [Turn the work into a credible application](https://fdelab.dev/learn/#career) ## Build a software foundation Make code a tool you can rely on. If you already build software professionally, use the exercises as a diagnostic and spend time only on the gaps. Outcome: A small application that you can run, test, debug, and explain without following a tutorial. ### Write and debug a useful program Learn variables, functions, lists and dictionaries, file I/O, exceptions, and unit tests. Read stack traces and use a debugger. Choose Python for this path; it is a practical starting choice, not a universal FDE requirement. Practice: Build a command-line tool that reads a CSV of support tickets, validates required fields, groups tickets by category, and exports a report. Add empty-file, malformed-row, and duplicate-ID cases. Evidence: You can explain each transformation and show tests for both normal and malformed inputs. - [CS50’s Introduction to Programming with Python](https://cs50.harvard.edu/python/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-programming) ### Work in a shared codebase Practise commits, branches, diffs, merges, and reverting a change. Learn shell navigation, environment variables, dependency isolation, and how to keep secrets out of a repository. Practice: Put your ticket tool in Git. Make a feature branch, introduce and fix a failing test, then write a setup guide. Ask a peer to run it from a fresh checkout. Evidence: A second person can install and run it, and the history explains the changes. - [Pro Git](https://git-scm.com/book/en/v2) [Work on this milestone](https://fdelab.dev/workspace/#lesson-workflow) ### Make the result usable Understand the browser/server boundary, HTTP methods and status codes, JSON, and basic HTML. Learn enough JavaScript to send a request and present loading, empty, success, and error states. Practice: Build a small page that accepts a ticket file or text input and shows the results. Make an error understandable to someone who did not write the code. Evidence: The interface can be used with a keyboard and makes failures visible. - [Learn web development](https://developer.mozilla.org/en-US/docs/Learn_web_development) - [Full Stack Open](https://fullstackopen.com/en/part0/general_info/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-web-basics) ### Direct AI work and verify it Agent Factory describes two uses of a general agent: solving a task in a session and helping build a worker that runs independently. Its 10-80-10 rule is a planning heuristic for human direction, AI execution, and human review; it is not a measured allocation of time. Practice: Give an assistant the ticket-tool task with three acceptance examples. Inspect its diff, run the examples, and explain one incorrect assumption. Log time spent specifying, generating, and correcting. Evidence: You can reproduce the result and explain the code, its boundaries, and the checks you used. Faster generation alone is not the success measure. - [The AI Agent Factory: About](https://agentfactory.panaversity.org/docs/about) - [Pro Git](https://git-scm.com/book/en/v2) - [How to Learn from This Book](https://agentfactory.panaversity.org/docs/how-to-learn-from-this-book) [Work on this milestone](https://fdelab.dev/workspace/#lesson-direct-and-review) Stage review: Change the input format without following a tutorial. Can you isolate the failure, update the implementation, and prove the fix with a test? ## Understand the workflow first An FDE needs to discover what is worth building. Practise listening, mapping the current process, and narrowing a request into a deliverable. Outcome: An evidence ledger, stakeholder map, approved delivery specification, and a justified workflow redesign. ### Run a discovery conversation Ask about the last time a task happened, rather than whether someone likes your idea. Understand the trigger, steps, exceptions, current tools, and who is accountable. Record assumptions separately from observations. Practice: Interview a willing user for 20 minutes, or run a clearly labelled simulation. Ask them to walk through a recent task. Write a workflow map and identify one costly handoff. Evidence: Your notes distinguish what the user said, what you observed, and what you still need to verify. - [Using in-depth interviews](https://www.gov.uk/service-manual/user-research/using-in-depth-interviews) [Work on this milestone](https://fdelab.dev/workspace/#lesson-interview-user) ### Define a small, testable outcome Choose one user and one workflow. Define the current baseline, the desired outcome, and how you will measure it. Ask about data access, approval, budget, and the person who accepts the delivery. Practice: Write a one-page brief: problem, user, current process, proposed slice, non-goals, data access, success test, and owner. For a simulation, choose a synthetic dataset and state its limits. Evidence: A reviewer can decide whether the project succeeded without relying on your demo narrative. - [Using in-depth interviews](https://www.gov.uk/service-manual/user-research/using-in-depth-interviews) [Work on this milestone](https://fdelab.dev/workspace/#lesson-scope) ### Explain a technical trade-off Compare a manual process, a deterministic automation, and an AI-assisted workflow. Explain trade-offs in accuracy, cost, latency, and maintenance. Learn to report a blocker with a decision or next step. Practice: Write a short decision record choosing the simplest viable approach. Then send a simulated weekly update: result, next step, blocker, and decision needed. Evidence: Your recommendation is understandable to a nontechnical decision-maker and states what would change your mind. - [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents) - [Is This an Agent Problem?](https://agentfactory.panaversity.org/docs/is-this-an-agent-problem) [Work on this milestone](https://fdelab.dev/workspace/#lesson-tradeoffs) ### Validate the brief before committing Buehling calls unverified understanding discovery debt. Compare the sponsor’s request with operator evidence, technical constraints, and the real reason for the deadline. Connect goals, scope, risk, time, and budget before promising delivery. Practice: Create an evidence ledger with claim, source, confidence, owner, and next check. Test three claims from a fictional AI-written brief. Map the sponsor, daily operator, system owner, and acceptance owner; record one disagreement. Evidence: Every critical assumption has evidence or an unresolved decision. Write a conditional commitment: “We can deliver this slice if access is approved by this date; otherwise we reduce scope or pause.” - [Forward Deployed — Richard Buehling (February 2026)](https://fdelab.dev/workspace/#ref-benmore) - [Using in-depth interviews](https://www.gov.uk/service-manual/user-research/using-in-depth-interviews) [Work on this milestone](https://fdelab.dev/workspace/#lesson-discovery-debt) ### Write a specification a reviewer can test Write observable behavior before implementation. Link each requirement to evidence and an acceptance check. Keep the approved intent and the actual implementation aligned when new information changes the work. Practice: Specify one ticket-routing slice: input schema, output, denied actions, error handling, latency and cost budget, reviewer, and three acceptance examples. Change one customer constraint and update the affected check and decision record. Evidence: A peer can test the requirement from the document. Open questions stay visible instead of becoming invented requirements. - [Spec-Driven Development](https://agentfactory.panaversity.org/docs/spec-driven-development-crash-course) - [Forward Deployed — Richard Buehling (February 2026)](https://fdelab.dev/workspace/#ref-benmore) [Work on this milestone](https://fdelab.dev/workspace/#lesson-delivery-spec) ### Decide what to preserve, change, or remove Observe the old workflow, then question why each step exists. Agent Factory’s vertical-design method distinguishes enduring obligations from steps created by older tools or organizational constraints. Validate proposed changes with the responsible operator. Assign each step to a person, AI, or a reviewed collaboration. Explain the choice using reversibility, the cost of error, and who is accountable. Practice: Map the three eligibility lookups in your project. For each, record its purpose, evidence, and a keep/change/remove recommendation. Preserve the eligibility decision while testing whether one repeated data-entry step can disappear. Name who reviews, what they check, and when review happens before the action takes effect. Give the map an owner and a review date. Evidence: The revised flow preserves required controls and has an owner-approved test. An unexplained step remains unresolved, not automatically deleted. - [Designing the Vertical System of Record](https://agentfactory.panaversity.org/docs/ecosystem/designing-the-vertical-sor) - [Workflow Design & Diagnosis](https://agentfactory.panaversity.org/docs/workflow-design-diagnosis-crash-course) [Work on this milestone](https://fdelab.dev/workspace/#lesson-redesign-work) Stage review: Have the customer change a critical assumption. Update scope, acceptance checks, time and budget implications; explain whether to proceed, narrow, or pause. ## Connect data and systems Connect your project to persistent data and an external interface. Reliability starts at the boundaries: schemas, identities, permissions, and retries. Outcome: A tested API, recoverable data flow, and governed policy register with denied-access evidence. ### Model and query the data Learn tables, primary keys, foreign keys, joins, aggregates, transactions, and indexes. Identify missing values, duplication, inconsistent timestamps, and fields that should not enter your application. Practice: Load your project data into a relational database. Create a schema, a migration, a duplicate-detection query, and a query that measures your workflow baseline. Evidence: You can explain the grain of each table and avoid double-counting when joining data. - [PostgreSQL tutorial](https://www.postgresql.org/docs/current/tutorial.html) [Work on this milestone](https://fdelab.dev/workspace/#lesson-data-model) ### Build an API with clear boundaries Define request and response schemas. Distinguish authentication from authorisation. Return meaningful status codes and avoid exposing secrets in logs or responses. Practice: Add endpoints to create, view, and update a record. Validate inputs, add a per-user permission check, and test allowed and denied requests. Use test identities locally. Evidence: Tests demonstrate that one user cannot read or change another user’s records. - [FastAPI tutorial](https://fastapi.tiangolo.com/tutorial/) - [Testing FastAPI applications](https://fastapi.tiangolo.com/tutorial/testing/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-api-contract) ### Make the integration recoverable Assume remote calls can fail and events can repeat. Use bounded retries and backoff where appropriate. Define what happens after a timeout and how a caller can safely retry a write. Practice: Connect a mock customer API. Simulate paginated results, a rate-limit response, duplicate events, and a partial failure. Add a reconciliation script for the failed records. Evidence: A replay does not duplicate work, and a failed batch can resume without losing completed records. - [Testing FastAPI applications](https://fastapi.tiangolo.com/tutorial/testing/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-resilience) ### Give domain knowledge an accountable owner A searchable document collection needs ownership, review, versioning, permissions, and citations to serve as governed knowledge. Agent Factory separates source evidence, a map of what exists, and complete procedures that must be followed together. Keep three source types distinct: approved rules, current values from the owning system, and supporting conversations. An index helps find evidence; it does not become the authority. Practice: Create a synthetic policy register with document ID, version, owner, approval date, review date, access group, and superseded version. Add one complete eligibility procedure. Test expired guidance, a revoked reader, and a missing required step. Assemble one context packet with the applicable policy version, live ticket facts, supporting notes, missing evidence, and citations. A chat message cannot override the approved policy. Evidence: The answer cites an approved version; unavailable authority leads to escalation. Permissions are enforced outside the model, and a procedure is never executed from a partial search excerpt. - [The FDE AF Model](https://agentfactory.panaversity.org/docs/ecosystem/fde-af-model) - [Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) - [Excessive agency](https://genai.owasp.org/llmrisk/llm062025-excessive-agency/) - [System of Context](https://agentfactory.panaversity.org/docs/ecosystem/system-of-context) [Work on this milestone](https://fdelab.dev/workspace/#lesson-governed-knowledge) Stage review: Replay the same event twice, deny the caller permission, and make the dependency time out. Show exactly what changed in the database for each case. ## Build an AI workflow you can evaluate Learn model APIs and tool use in the context of a useful workflow. Add autonomy only when you can show that it improves the outcome. Outcome: A bounded AI worker, a non-AI baseline, a calibrated evaluation report, and an approval-and-recovery interface. ### Start with a bounded model call Learn messages, tokens, structured output, tool calling, and schema validation. Separate a model’s suggestion from an operation your application executes. Set a usage budget before using a paid API. Practice: Add ticket classification or draft suggestions. Compare a rule-based baseline with one model call. Validate the returned schema and route uncertain or invalid outputs to human review. Evidence: The workflow behaves sensibly when the model is unavailable or returns an invalid response. - [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents) - [AI Agents Course](https://huggingface.co/learn/agents-course/en/unit0/introduction) [Work on this milestone](https://fdelab.dev/workspace/#lesson-model-workflow) ### Retrieve the right context safely Learn when keyword search, indexed retrieval, or live exploration fits. Preserve source identity, apply access controls before returning content, and handle missing or contradictory evidence. Practice: Add a small knowledge collection with source links. Include stale documents, two conflicting answers, and a document that a test user cannot access. Evidence: Your test user never receives the restricted document and can trace an answer to its source. - [Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) - [Excessive agency](https://genai.owasp.org/llmrisk/llm062025-excessive-agency/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-retrieval) ### Measure outcomes and inspect failures Create representative tasks with expected outcomes. Separate development examples from evaluation cases. Use deterministic checks when possible and calibrate subjective judgements with human review. Practice: Create an initial set of 20–30 diverse test cases as a project exercise. Include ambiguity and adversarial input. Record completions, failure categories, latency, and cost; explain the small sample’s limits. Evidence: Your report includes failed cases and the actual resulting system state, not just a convincing model response. - [Demystifying evals for AI agents](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents) [Work on this milestone](https://fdelab.dev/workspace/#lesson-evaluation) ### Turn a session task into an operable worker A useful session result does not establish that a workflow can run unattended. Specify who owns a recurring job, what wakes it, what it can change, where progress lives, and how it stops. Keep human judgment explicit at consequential boundaries. Practice: Run the ticket task from a synthetic queue. Persist job state, cap runtime and retries, require review before assignment, and expose pause/resume. Interrupt a job after a suggested route is saved and demonstrate safe recovery. Evidence: A duplicate trigger causes no duplicate assignment. The operator can see why work stopped and resume without guessing what already happened. - [The AI Agent Factory: About](https://agentfactory.panaversity.org/docs/about) - [Eval-Driven Development for AI Employees](https://agentfactory.panaversity.org/docs/eval-driven-development-crash-course) - [From One-Off to Worker](https://agentfactory.panaversity.org/docs/from-one-off-to-worker) - [Harness Engineering](https://agentfactory.panaversity.org/docs/harness-engineering-crash-course) [Work on this milestone](https://fdelab.dev/workspace/#lesson-worker-contract) ### Check the evaluator before trusting its score Tests verify specific properties; evaluation estimates behavior on a sample. A model judge can miss a bad outcome or reward a persuasive explanation. Check the final state, compare with human judgments, and keep evaluation examples separate from tuning. Practice: Label 12 held-out routing runs as correct or incorrect with a peer. Include fluent but wrong suggestions, correct but terse results, and failed writes. Compare the judge with those labels, list false passes and false failures, then revise the rubric without editing the held-out cases. Evidence: You can show a disagreement table, explain which errors the grader misses, and state where human review is still needed. Twelve examples are a calibration exercise, not production reliability evidence. - [Trusting the Checker](https://agentfactory.panaversity.org/docs/trusting-the-checker-crash-course) - [Eval-Driven Development for AI Employees](https://agentfactory.panaversity.org/docs/eval-driven-development-crash-course) - [Demystifying evals for AI agents](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents) [Work on this milestone](https://fdelab.dev/workspace/#lesson-check-the-checker) ### Design a workflow people can supervise An approval screen should show the proposed change, supporting evidence, affected records, and available alternatives. Distinguish queued, running, waiting for review, failed, and completed work. Design recovery and operator access before expanding autonomy. Practice: Build a routing review screen with approve, edit, and reject actions. Seed a stale recommendation and an interrupted write. Ask a peer to use only the interface to explain the change, correct it, and recover the interrupted task. Test the same task with a keyboard and at a narrow viewport. Evidence: The reviewer sees what will happen before approving. A stale decision is revalidated, a repeated click cannot repeat the write, and the person can find the next action after a failure. Record completion time and the confusing step. - [Designing Agent Experiences](https://agentfactory.panaversity.org/docs/designing-agent-experiences-crash-course) - [Workflow Design & Diagnosis](https://agentfactory.panaversity.org/docs/workflow-design-diagnosis-crash-course) - [Learn web development](https://developer.mozilla.org/en-US/docs/Learn_web_development) [Work on this milestone](https://fdelab.dev/workspace/#lesson-review-and-recover) Stage review: Replace the model or prompt, rerun the same evaluation, and explain the change in quality, cost, and latency. Include at least one regression. ## Make it work after the demo Give your project an operational life. Make it possible to see a failure, recover from it, and hand the system to another person. Outcome: A release-gap register, reproducible deployment, recovery drill, and accepted ownership and support pack. ### Deploy reproducibly Separate code from environment-specific configuration. Pin the dependencies your build needs. Understand health checks and what a deploy changes in application and database state. Practice: Containerise the service or document an equivalent reproducible setup. Add an automated test step, a health check, and a rollback procedure. Keep a local demo mode to control cost. Evidence: You can deploy a known version and return to the previous working version. - [Get started with Docker](https://docs.docker.com/get-started/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-deploy) ### See and diagnose a failure Connect a user-visible failure to the requests and dependencies involved. Record useful identifiers without exposing sensitive data. Choose measurements that correspond to the customer workflow. Practice: Add structured logs and measure task completion, p95 latency, and cost per task. Simulate a slow dependency. Create one actionable alert with an owner and a response. Evidence: You can find the failing request, identify the dependency, and describe the user impact. - [Observability signals](https://opentelemetry.io/docs/concepts/signals/) - [The Site Reliability Workbook](https://sre.google/workbook/table-of-contents/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-observe) ### Transfer ownership responsibly Document setup, permissions, common failures, recovery, and operational ownership. Observe a user doing the workflow and remove friction. Describe what the system cannot yet do. Practice: Write a runbook and a short incident report from your failure exercise. Run a handoff session, ask the other person to operate the service, and capture one improvement from feedback. Evidence: A named person can operate the project without relying on your memory. - [The Site Reliability Workbook](https://sre.google/workbook/table-of-contents/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-handoff) ### Close the gap between prototype and release Buehling separates prototype, specification completion, and production readiness. A convincing interface can still contain stubbed integrations, unsupported rules, and untested recovery. Keep a visible record of those gaps. Practice: Make a gap table: requirement, current behavior, placeholder, risk, evidence needed, and owner. Replace one mock dependency with a test integration. Recheck earlier acceptance cases and review the full workflow before proposing a pilot. Evidence: Every release-critical gap has been closed with evidence or explicitly excluded from the pilot. Tests and permission checks start with implementation and continue through release. - [Forward Deployed — Richard Buehling (February 2026)](https://fdelab.dev/workspace/#ref-benmore) - [Spec-Driven Development](https://agentfactory.panaversity.org/docs/spec-driven-development-crash-course) [Work on this milestone](https://fdelab.dev/workspace/#lesson-prototype-gap) ### Agree the support boundary and prove handoff Buehling’s handoff stack combines an operational runbook, a system overview, and a maintained specification tree. Receiving teams learn through changes and recovery exercises. Support needs explicit responsibilities, coverage, and an end or review point. Practice: Have a peer diagnose a seeded failure, roll back a release, and implement a small requirement change. Record the drill outcome, primary and backup owner, support hours, escalation route, exclusions, and next review date. Evidence: The receiver can operate and change the system. A named owner accepts remaining issues, and ongoing support distinguishes defect fixes from new scope. - [Forward Deployed — Richard Buehling (February 2026)](https://fdelab.dev/workspace/#ref-benmore) - [The Site Reliability Workbook](https://sre.google/workbook/table-of-contents/) - [Human-Agent Teams](https://agentfactory.panaversity.org/docs/human-agent-teams-crash-course) [Work on this milestone](https://fdelab.dev/workspace/#lesson-support-exit) Stage review: Ask a peer to recover from a simulated outage using only the runbook. Record where they got stuck and improve the instructions. ## Turn the work into a credible application Make your contribution easy to assess. Connect engineering decisions to the user’s problem and apply to roles whose responsibilities match your experience. Outcome: A reproducible case study, a bounded engagement proposal, practice interviews, and a targeted application shortlist. ### Write the project case study Explain the original workflow, your scope, the architecture, a difficult trade-off, the evaluation, and what happened after delivery. Distinguish real user outcomes from results on synthetic data. Practice: Publish a concise case study with a diagram, a runnable repository, a demo, a failure example, and the runbook. Make setup instructions the first thing a reviewer can find. Evidence: A reviewer can reproduce your evidence and understand what you personally owned. - [The AI-native interview](https://sierra.ai/blog/the-ai-native-interview) [Work on this milestone](https://fdelab.dev/workspace/#lesson-case-study) ### Practise explaining and building Practise a timed debugging task and a solution-design conversation with missing requirements. State assumptions, ask focused questions, and explain alternatives. Check AI permissions for the actual assessment. Practice: Do two mock interviews with a peer. In one, debug a broken integration. In the other, design a workflow and respond to a changed constraint. Review where your reasoning became unclear. Evidence: You can explain your code and decisions, respond to feedback, and work within the stated tool rules. - [OpenAI interview guide](https://openai.com/interview-guide/) - [Candidate AI guidance](https://www.anthropic.com/candidate-ai-guidance) - [The AI-native interview](https://sierra.ai/blog/the-ai-native-interview) [Work on this milestone](https://fdelab.dev/workspace/#lesson-practice-interview) ### Build a targeted application shortlist Read the exact requisition. Compare what you would build, the customer interaction, seniority, location eligibility, and travel. Tailor your CV around relevant ownership and outcomes. Practice: Shortlist five roles using the directory, recheck each employer page, and map three requirements per role to concrete evidence. Rewrite two CV bullets without inventing impact numbers. Evidence: Every application has a clear match, an evidence link, and explicit gaps you can discuss. - [OpenAI interview guide](https://openai.com/interview-guide/) [Work on this milestone](https://fdelab.dev/workspace/#lesson-applications) ### Present a bounded engagement or portfolio proposal Frame your work around an accepted result and the evidence you will deliver. Employment, independent client work, and a domain product require different commercial arrangements. Do not infer income or demand from a teaching framework. Practice: Write a one-page proposal for the synthetic routing pilot: sponsor, baseline, fixed scope, assumptions, access needs, acceptance, support boundary, and exclusions. Explain which generic adapter or check could be reused without customer data. Evidence: A reviewer can distinguish the deliverable, acceptance decision, recurring support, and optional expansion. Label simulated results and explain who owns any proposed reusable material. - [Forward Deployed — Richard Buehling (February 2026)](https://fdelab.dev/workspace/#ref-benmore) - [The FDE AF Model](https://agentfactory.panaversity.org/docs/ecosystem/fde-af-model) [Work on this milestone](https://fdelab.dev/workspace/#lesson-delivery-offer) Stage review: Give a peer five minutes with your case study. Can they explain the problem, your contribution, the result, and the most important limitation? --- # Benmore and Agent Factory: an FDE reading path Read with a concrete decision to make. Close the source, recall the idea, then build the artifact. These prompts are an independent synthesis; the original Benmore PDF is not hosted. [Track reading progress](https://fdelab.dev/workspace/#reading) [Read the source-review limits](https://fdelab.dev/research/material-review.md) ## 1. Understand the job What are you responsible for after the demo? Start with a person’s task and a result they can verify. Use AI to help execute the work while retaining responsibility for the decision, the evidence, and the operating relationship. Keep this distinction: Benmore describes its consulting practice. Agent Factory focuses on AI workers. FDE roles can also involve conventional software and data systems. Neither book defines every employer’s role. Benmore: Chapter 1, Compressed Like a DJ, printed pages 2–8; Chapter 11, What is Forward Deployed?, printed pages 72–75; Chapter 12, The Deployed Workflow, printed pages 75–77. - [The AI Agent Factory: About](https://agentfactory.panaversity.org/docs/about) - [How to Learn from This Book](https://agentfactory.panaversity.org/docs/how-to-learn-from-this-book) Recall: Explain the difference between the human FDE and an AI worker without using either acronym. ### Build: Responsibility map - Customer problem and daily user - What I own: discovery, delivery, adoption, operations - What needs another specialist or decision owner - One outcome and how we will observe it [Apply this in the learning guide](https://fdelab.dev/learn/#direct-and-review) ## 2. Discover before promising Which part of the brief is still a guess? Treat a polished request as the beginning of an investigation. Replay recent work with its operator. Separate observed behavior, reported needs, and assumptions before agreeing a delivery date. Keep this distinction: A sponsor’s preferred solution contains useful context, but it is not proof of feasibility. A generated requirement does not become a fact because it appears in several documents. Benmore: Chapter 2, Disco Debt, printed pages 8–14; Chapter 3, Yes, Yes! Yes if…, printed pages 14–17; Chapter 4, What Discovery Actually Is, printed pages 18–23; Chapter 5, Understanding the Client, printed pages 23–32. - [Workflow Design & Diagnosis](https://agentfactory.panaversity.org/docs/workflow-design-diagnosis-crash-course) Recall: Name one observation that would make you change your proposed solution. ### Build: Evidence ledger - Claim and source - Observed, reported, or assumed - Impact if wrong - Owner and next verification step [Apply this in the learning guide](https://fdelab.dev/learn/#discovery-debt) ## 3. Make a conditional commitment What can you deliver within the real constraints? Choose a small outcome with a baseline, a test, and an owner. Compare ordinary automation with a model call before committing to an agent. Make access dependencies and stop conditions visible. Keep this distinction: Respectful challenge protects the relationship. Treat Benmore’s escalation and separation advice as one firm’s practice; follow the actual engagement’s authority and agreements. Benmore: Chapter 3, Yes, Yes! Yes if…, printed pages 14–17; Chapter 6, The Relationship, printed pages 32–35; Chapter 7, The Discovery Process, printed pages 36–50; Chapter 8, Walking Away, printed pages 51–56. - [Is This an Agent Problem?](https://agentfactory.panaversity.org/docs/is-this-an-agent-problem) Recall: If access arrives two weeks late, what changes: scope, date, or the decision to continue? ### Build: Pilot brief - User, baseline, and desired outcome - Included work and explicit non-goals - Access, time, and budget assumptions - Acceptance owner and pause condition [Apply this in the learning guide](https://fdelab.dev/learn/#scope) ## 4. Write intent that can be tested Could someone build the wrong thing and still pass? Connect project intent to a feature, then to a small implementation task and its acceptance examples. Update both the specification and the checks when new evidence changes the plan. Keep this distinction: A complete-looking specification can preserve a wrong assumption. Review the intent with the operator as well as checking the resulting code. Benmore: Chapter 7, The Discovery Process, printed pages 36–50; Chapter 9, General Trends in AI-Driven Development, printed pages 57–62. - [Spec-Driven Development](https://agentfactory.panaversity.org/docs/spec-driven-development-crash-course) Recall: Close the document. Describe one required behavior and the exact observation that proves it. ### Build: Behavior specification - Input and required output - Rules, permissions, and denied actions - Happy path and two failure examples - Links to evidence and changed decisions [Apply this in the learning guide](https://fdelab.dev/learn/#delivery-spec) ## 5. Decide who owns each step What should change, and what must still be protected? Study the current workflow to understand it, then redesign around the outcome. Keep obligations and useful controls; challenge repeated handoffs and data entry. Assign responsibility at each consequential step. Keep this distinction: Copying an old workflow can preserve waste. Removing a step without understanding its purpose can remove a necessary control. Resolve uncertainty with the accountable operator. Benmore: Chapter 4, What Discovery Actually Is, printed pages 18–23; Chapter 5, Understanding the Client, printed pages 23–32; Chapter 7, The Discovery Process, printed pages 36–50. - [Designing the Vertical System of Record](https://agentfactory.panaversity.org/docs/ecosystem/designing-the-vertical-sor) - [Workflow Design & Diagnosis](https://agentfactory.panaversity.org/docs/workflow-design-diagnosis-crash-course) Recall: Why is “a human will review it” an incomplete control? ### Build: Delegation map - Step and purpose - Keep, change, or remove, with evidence - Human, AI, or collaboration, with reason - Reviewer, check, timing, and review date [Apply this in the learning guide](https://fdelab.dev/learn/#redesign-work) ## 6. Give every answer a source Is this a rule, a live fact, or someone’s interpretation? Bring together approved knowledge, current operational facts, and relevant case notes. Preserve source identity, version, access boundaries, and missing evidence when assembling context for a task. Keep this distinction: Search relevance does not establish authority. A model’s memory, an indexed copy, or a chat comment cannot silently replace the system or expert that owns the answer. Benmore: Chapter 7, The Discovery Process, printed pages 36–50. - [The FDE AF Model](https://agentfactory.panaversity.org/docs/ecosystem/fde-af-model) - [System of Context](https://agentfactory.panaversity.org/docs/ecosystem/system-of-context) Recall: What should happen when a helpful note conflicts with an approved procedure? ### Build: Context packet - Task, actor, and allowed action - Applicable rule, version, owner, and citation - Live facts from the owning system - Supporting notes, conflicts, and missing evidence [Apply this in the learning guide](https://fdelab.dev/learn/#governed-knowledge) ## 7. Use the prototype to learn What looks finished but is still simulated? Use a working slice to test the flow with a person. Keep a gap register that distinguishes demonstrated behavior, placeholders, missing integrations, and release evidence. Keep this distinction: Benmore’s prototype percentages are illustrative, not acceptance targets. Start security and permission checks with implementation; a later readiness review does not replace them. Benmore: Chapter 9, General Trends in AI-Driven Development, printed pages 57–62; Chapter 10, An AI-Driven Development Process, printed pages 63–71. - [Spec-Driven Development](https://agentfactory.panaversity.org/docs/spec-driven-development-crash-course) Recall: Show one prototype behavior you would not yet expose to a real user, and explain why. ### Build: Release-gap register - Requirement and current behavior - Mock or incomplete dependency - Risk, owner, and evidence needed - Pilot exclusion or closure decision [Apply this in the learning guide](https://fdelab.dev/learn/#prototype-gap) ## 8. Prove the result and the checker Which wrong answer still earns a passing grade? Compare a simple baseline with the proposed workflow on held-out tasks. Inspect the resulting state and failure groups. Check a model grader against independent labels before using its score to decide on release. Keep this distinction: A deterministic test proves a bounded property; an evaluation describes sampled behavior. A high average can hide a serious failure in a small but important group. Benmore: Chapter 10, An AI-Driven Development Process, printed pages 63–71. - [Trusting the Checker](https://agentfactory.panaversity.org/docs/trusting-the-checker-crash-course) - [Eval-Driven Development for AI Employees](https://agentfactory.panaversity.org/docs/eval-driven-development-crash-course) Recall: Give an example where the response sounds correct but the task actually failed. ### Build: Evaluation decision - Baseline and held-out cases - Expected and actual final state - False passes, false failures, and serious errors - Release decision, limitations, and reviewer [Apply this in the learning guide](https://fdelab.dev/learn/#check-the-checker) ## 9. Design approval and recovery Can the operator understand, change, and recover the action? Place the proposed change, evidence, and alternatives at the decision point. Show where work is waiting and who can act. Practise a wrong recommendation and an interrupted operation through the interface. Keep this distinction: Correct model output does not guarantee usable software. Measure the attention and corrections the complete workflow demands from people. Benmore: Chapter 7, The Discovery Process, printed pages 36–50; Chapter 10, An AI-Driven Development Process, printed pages 63–71; Chapter 13, Knowledge Transfer and Graduation, printed pages 77–81. - [Designing Agent Experiences](https://agentfactory.panaversity.org/docs/designing-agent-experiences-crash-course) Recall: What information must be visible before a reviewer approves an assignment? ### Build: Supervision brief - Action preview and supporting evidence - Approve, edit, reject, and stale-decision behavior - Pending, failed, and completed states - Recovery path and keyboard-only task result [Apply this in the learning guide](https://fdelab.dev/learn/#review-and-recover) ## 10. Turn repeated work into a worker What replaces the help you give during a session? Promote a proven recurring task by making its specification, checks, escalation rules, and runtime durable. Give jobs identity, bounded execution, persistent state, and an accountable operator. Keep this distinction: A successful session is useful evidence, but it does not establish unattended reliability. Keep the simpler workflow if repetition and value do not justify operating a worker. Benmore: Chapter 10, An AI-Driven Development Process, printed pages 63–71; Chapter 14, The Support Phase, printed pages 82; Chapter 17, Support in the AI Era, printed pages 88–90. - [From One-Off to Worker](https://agentfactory.panaversity.org/docs/from-one-off-to-worker) - [Harness Engineering](https://agentfactory.panaversity.org/docs/harness-engineering-crash-course) Recall: If the worker stops after a remote write succeeds, how does the next run avoid repeating it? ### Build: Worker operating contract - Trigger, job ID, and durable state - Allowed actions, runtime, and retry limits - Escalation reason, recipient, and evidence - Pause, restart, and duplicate-trigger demonstration [Apply this in the learning guide](https://fdelab.dev/learn/#worker-contract) ## 11. Transfer the ability to operate Could another person fix and change the system? Prepare a runbook for incidents, a system overview for design decisions, and a maintained specification. Ask the receiver to recover a failure and make a small change using those assets. Keep this distinction: Documentation delivery is not the same as knowledge transfer. Choose technology and operating practices that the receiving team can maintain. Benmore: Chapter 11, What is Forward Deployed?, printed pages 72–75; Chapter 12, The Deployed Workflow, printed pages 75–77; Chapter 13, Knowledge Transfer and Graduation, printed pages 77–81. - [Human-Agent Teams](https://agentfactory.panaversity.org/docs/human-agent-teams-crash-course) Recall: Which step did the receiver need you to explain, and what did you change afterward? ### Build: Witnessed handoff - Primary and backup owner - Runbook, system overview, and current specification - Recovery and change drill results - Unresolved issues and accepted responsibilities [Apply this in the learning guide](https://fdelab.dev/learn/#support-exit) ## 12. Agree what happens after delivery Where does support end and new work begin? Match support to actual operating needs. Separate maintaining agreed behavior from new scope, name response ownership, and set a review point. Reuse only the assets you have the right to reuse. Keep this distinction: Benmore’s support percentage is its commercial example, not a market rate. Agent Factory’s vertical business model is an option, not a requirement for employment or a promise of earnings. Benmore: Chapter 14, The Support Phase, printed pages 82; Chapter 15, The Support Model, printed pages 82–85; Chapter 16, When to Recommend What, printed pages 86–87; Chapter 17, Support in the AI Era, printed pages 88–90. - [The FDE AF Model](https://agentfactory.panaversity.org/docs/ecosystem/fde-af-model) Recall: When would the honest recommendation be no additional support contract? ### Build: Support and reuse agreement - Coverage, hours, escalation, and exclusions - Maintenance versus new scope - Operating owner and next review date - Reusable asset, ownership, and customer-data boundary [Apply this in the learning guide](https://fdelab.dev/learn/#delivery-offer) --- # Books, courses, and resources for FDE practice Choose a resource for the decision or behavior you need to improve. Reading recommendations are editorial; access and source checks are dated, and full paid courses were not completed for this catalogue. [Search and save resources in the library](https://fdelab.dev/workspace/#build) ## Forward Deployed: Discovery, AI-Driven Development, and the Forward Deployed Engineer Richard Buehling · Benmore · book · intermediate A consulting practitioner’s method for discovery, specification-led development, deployment, handoff, and support. Use it for: Use Chapters 4–8 for engagement judgment and Chapters 10–15 for delivery and ownership. Practice: Build an assumption register, track prototype gaps, and ask a receiver to perform a rollback. Access: User-supplied PDF; no public full-text link verified. This entry opens the source note and chapter guide. The original PDF is not hosted. Limits: Benmore’s February 2026 perspective. Its support pricing and workflow percentages are company examples, not industry benchmarks. Checked 2026-09-14. Supplied PDF inspected: title, contents, discovery, specification, development and handoff chapters. Page references in the source guide use printed page numbers. [Open the source and chapter notes](https://fdelab.dev/workspace/#ref-benmore) ## The AI Agent Factory Panaversity · Panaversity · book · intermediate An online curriculum for specifying, building, and supervising AI workers. Use it for: Use its orientation to distinguish session assistance from recurring worker operation. Practice: Compare a one-off task with a recurring version; name the owner, trigger, limits, and review. Access: Public web reading; optional tools, model calls, and hosted infrastructure may cost extra. Limits: The 10-80-10 rule is a teaching heuristic. Economic predictions and Digital FTE terminology do not prove labor replacement. Checked 2026-09-14. Official page read on 14 September 2026. This records the author’s educational method, not independent validation of its market forecasts or ecosystem performance. [Read the original resource](https://agentfactory.panaversity.org/docs/about) ## Spec-Driven Development Panaversity · Panaversity · course · intermediate A course on agreeing intent, defining acceptance, and keeping specifications current. Use it for: Translate a broad customer request into checks that can survive a change of implementation. Practice: Write three acceptance examples before building; revise the spec when a constraint changes. Access: Public web reading; optional tools, model calls, and hosted infrastructure may cost extra. Limits: Specifications need evidence and review; writing one does not establish that the customer problem is valid. Checked 2026-09-14. Official page read on 14 September 2026. This records the author’s educational method, not independent validation of its market forecasts or ecosystem performance. [Read the original resource](https://agentfactory.panaversity.org/docs/spec-driven-development-crash-course) ## The FDE AF Model Panaversity · Panaversity · article · intermediate A proposed model separating shared machinery, governed knowledge, domain assets, and customer delivery. Use it for: Distinguish reusable method from domain evidence and customer-specific integration. Practice: Give every source a reviewer and version, then identify one adapter worth reusing. Access: Public web reading; optional tools, model calls, and hosted infrastructure may cost extra. Limits: A proposed architecture and business model; named infrastructure and ecosystem claims are not independently tested here. Checked 2026-09-14. Official page read on 14 September 2026. This records the author’s educational method, not independent validation of its market forecasts or ecosystem performance. [Read the original resource](https://agentfactory.panaversity.org/docs/ecosystem/fde-af-model) ## Designing the Vertical System of Record Panaversity · Panaversity · article · intermediate A method for examining inherited work and rebuilding it around outcomes, evidence, and enduring obligations. Use it for: Avoid encoding an old workaround as a permanent requirement. Practice: Explain why one manual step can be changed while its underlying control remains intact. Access: Public web reading; optional tools, model calls, and hosted infrastructure may cost extra. Limits: Proposed redesign still needs operator validation; an agent is not assumed to follow every step reliably. Checked 2026-09-14. Official page read on 14 September 2026. This records the author’s educational method, not independent validation of its market forecasts or ecosystem performance. [Read the original resource](https://agentfactory.panaversity.org/docs/ecosystem/designing-the-vertical-sor) ## Eval-Driven Development for AI Employees Panaversity · Panaversity · course · intermediate A multi-track curriculum connecting worker behavior with evaluation and release decisions. Use it for: Use it beside a working project and a domain reviewer. Practice: Record severe failures separately from average success, and use the same held-out cases to compare revisions. Access: Public web reading; optional tools, model calls, and hosted infrastructure may cost extra. Limits: A sample exercise is not production assurance. Validate graders and cover the consequences of your particular workflow. Checked 2026-09-14. Official page read on 14 September 2026. This records the author’s educational method, not independent validation of its market forecasts or ecosystem performance. [Read the original resource](https://agentfactory.panaversity.org/docs/eval-driven-development-crash-course) ## CS50’s Introduction to Programming with Python Harvard · Harvard · course · beginner A structured introduction to Python with problem sets and a final project. Use it for: Build the coding, debugging, and testing foundation needed before deployment work. Practice: Turn a CSV of tickets into a validated report and test malformed rows. Access: OpenCourseWare is free. Optional verified credentials are a separate purchase. Limits: Focuses on Python rather than a complete computer-science or FDE curriculum. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://cs50.harvard.edu/python/) ## The Missing Semester of Your CS Education (2026) Missing Semester teaching team · MIT CSAIL · course · beginner Learn the shell, development tools, version control, debugging, packaging, and modern coding workflows. Use it for: Useful when you must diagnose an unfamiliar customer environment and leave a reproducible setup behind. Practice: Build a one-command setup script, inspect a failed process, and recover a deliberately broken Git change. Access: Public lectures and exercises; self-study does not confer MIT credit. Limits: Start with basic programming familiarity; this develops tool fluency rather than teaching your first programming language. Checked 2026-09-14. Read the official course home and 2026 lecture index. The current course is the 2026 offering; older annual archives remain available. [Read the original resource](https://missing.csail.mit.edu/2026/) ## Using in-depth interviews Government Digital Service · Government Digital Service · docs · beginner Guidance for preparing and conducting interviews about users’ actual experience. Use it for: Replace assumptions about a customer workflow with observations and evidence. Practice: Interview a willing user about their last task, then separate observations from assumptions. Access: The public service manual is freely accessible. Limits: Public-service research context; adapt recruitment and consent to the organisation. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://www.gov.uk/service-manual/user-research/using-in-depth-interviews) ## PostgreSQL tutorial PostgreSQL · PostgreSQL · docs · beginner An introduction to relational concepts, SQL queries, joins, and transactions. Use it for: Inspect customer data, model entities, and avoid incorrect reports from joins. Practice: Model tickets and customers; demonstrate a join that does not double-count. Access: Official tutorial and database software are free. Hosted database services may cost money. Limits: The current URL is versioned by PostgreSQL; match reference details to your installed release. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://www.postgresql.org/docs/current/tutorial.html) ## Building effective agents Erik S. and Barry Zhang · Anthropic Engineering · article · beginner A concise pattern language for chaining, routing, parallel work, orchestration and autonomous tool use. Use it for: Helps choose the simplest architecture that satisfies a customer's workflow. Practice: Draw your workflow, implement a deterministic baseline, and justify each point where model autonomy is necessary. Access: Public article; reproducing API examples can incur usage charges. Limits: Published December 2024; the page now explicitly warns that its tooling landscape has changed. Treat it as patterns, not a current SDK selection guide. Checked 2026-09-14. Read the original article, workflow/agent distinction, pattern descriptions, tool appendix and current freshness note. [Read the original resource](https://www.anthropic.com/engineering/building-effective-agents) ## AI Evals: Everything You Need to Know Hamel Husain and Shreya Shankar · Hamel's Blog · article · intermediate A detailed practical FAQ on product evaluations, error analysis, human labels, evaluators, RAG and production monitoring. Use it for: Connects deployment quality to actual customer tasks and shows how domain experts and engineers can improve a system together. Practice: Review a small batch of realistic traces with a domain expert, group failures, then implement one targeted check and test it on both passing and failing examples. Access: Full web guide and PDF are free; the linked live course is a separate paid offering. Limits: The authors label their guidance as contextual opinions; sample sizes and review schedules need to match your risks and data. Checked 2026-09-14. Full public FAQ read, including publication date, 1 Sep 2026 modification date, error analysis, judge validation and production sections. [Read the original resource](https://hamel.dev/blog/posts/evals-faq/) ## The Site Reliability Workbook Betsy Beyer, Niall Richard Murphy, David K. Rensin, Kent Kawahara, and Stephen Thorne (editors) · Google / O’Reilly · book · intermediate Worked examples for service objectives, alerting, incidents, pipelines, releases, and customer-facing reliability. Use it for: Provides concrete tools for agreeing what reliable means with a customer and transferring operational ownership. Practice: Define a user-facing SLI, propose an SLO, create a rollback drill, and write a sample incident communication. Access: Complete authorized online text is free; optional paid print or ebook editions. Limits: Basic experience running a service helps. This is a 2018 foundation: validate current monitoring syntax and cloud implementation details separately. Checked 2026-09-14. Read Google's official book register and workbook contents plus O’Reilly metadata confirming July 2018. Select chapters 2, 5, 9, 13, 16, and 19 for an FDE-oriented route. [Read the original resource](https://sre.google/workbook/table-of-contents/) ## How Cursor deploys AI inside the enterprise Richard MacManus with Pauline Brunet · Latent Space · article · beginner Cursor's FDE leader explains configurable deployments, production engineering experience, customer champions and feedback into the product roadmap. Use it for: A concrete current example of how customer work, engineering judgment and adoption fit together in an FDE team. Practice: Compare a target FDE job with this example: identify the customer, deployed artifact, internal champion, adoption signal and product feedback path. Access: Public original interview with a linked conference talk. Limits: One company's stated model and hiring preferences; its experience requirement and forecasts do not describe every FDE role. Checked 2026-09-14. Original interview read on publisher site; guest role, author and publication date verified. Company claims kept attributed rather than generalized. [Read the original resource](https://www.latent.space/p/cursor-forward-deployed-engineers) ## Agents Chip Huyen · Chip Huyen's Blog · article · intermediate A framework for reasoning about agent tools, planning, failure modes and evaluation rather than starting with a particular agent framework. Use it for: Useful for explaining architecture choices and failure handling to a customer before adding more autonomy. Practice: For one agent design, list its tools and write actions, identify planning and tool failures, and specify the checks and fallback for each. Access: Free standalone article adapted from the paid AI Engineering book. Limits: Published in early 2025; use the conceptual structure and verify current APIs and model capabilities in maintained documentation. Checked 2026-09-14. Original author article read, including date, table of contents, book-adaptation statement and experimental-framework caveat. [Read the original resource](https://huyenchip.com/2025/01/07/agents.html) ## AI Engineering Chip Huyen · O’Reilly · book · intermediate A foundation-model application guide covering evaluation, adaptation, retrieval, agents, data, latency, and cost. Use it for: Helps justify model and application choices against customer requirements rather than choosing a stack from a demo. Practice: Compare a simple baseline and a retrieval-based solution on the same labeled cases, reporting quality, latency, and cost. Access: Book purchase or O’Reilly subscription access; free previews are not the complete book. Limits: Comfort with application development helps. Model and framework details age quickly; pair concepts with current provider documentation. Complements Designing Machine Learning Systems. Checked 2026-09-14. Read the publisher overview, contents, and metadata. O’Reilly lists December 2024 for this digital title; this is the date recorded here. Paid full text was not accessed. [Read the original resource](https://www.oreilly.com/library/view/ai-engineering/9781098166298/) ## Designing Machine Learning Systems Chip Huyen · O’Reilly · book · advanced An iterative approach to production ML spanning data, features, objectives, deployment, monitoring, and adaptation. Use it for: Useful for FDE roles deploying predictive ML or maintaining customer models and data pipelines beyond foundation-model APIs. Practice: Specify training/serving data checks, a monitoring plan, and a retraining trigger for one predictive use case. Access: Book purchase or O’Reilly subscription access. Limits: Requires basic ML and software knowledge. Choose AI Engineering first for foundation-model applications; choose this when training and operating ML models are central. Checked 2026-09-14. Read O’Reilly's official title metadata and overview. The May 2022 scope is complementary to AI Engineering, not a replacement for its foundation-model focus. Description is based on public publisher/author material, not a full-text review. [Read the original resource](https://www.oreilly.com/library/view/designing-machine-learning/9781098107956/) ## Effective context engineering for AI agents Prithvi Rajasekaran, Ethan Dixon, Carly Ryan and Jeremy Hadfield · Anthropic Engineering · article · intermediate Techniques for selecting, retrieving and maintaining useful context across extended agent tasks. Use it for: Helps control irrelevant context, token costs and failures in long customer workflows. Practice: Compare full-history, compacted-history and retrieved-context variants on the same tasks; record quality and token use. Access: Public article. Limits: Provider experience and design guidance, not proof that one context strategy wins universally or that indexed retrieval is obsolete. Checked 2026-09-14. Read the original article, retrieval tradeoffs, long-horizon techniques and author acknowledgements. [Read the original resource](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) ## Hands-On Large Language Models Jay Alammar and Maarten Grootendorst · O’Reilly · book · beginner An illustrated introduction to LLM internals, embeddings, semantic search, retrieval and fine-tuning. Use it for: Builds the intuition needed to diagnose retrieval and model behavior rather than treating an API as a black box. Practice: Run the semantic-search notebook on a small document set, inspect retrieval misses, and compare a reranker. Access: The book is paid; the authors' official example notebooks are free on GitHub. Compute/API use may cost extra. Limits: Published September 2024; beginner-to-intermediate content still expects code practice, and notebook dependencies or free GPU availability can change. Checked 2026-09-14. Read the publisher metadata and the official https://github.com/HandsOnLLM/Hands-On-Large-Language-Models README/contents. Did not access paid full text or execute notebooks. [Read the original resource](https://www.oreilly.com/library/view/hands-on-large-language/9781098150952/) ## Hugging Face Agents Course Hugging Face education team · Hugging Face · course · beginner Hands-on agent fundamentals, frameworks, agentic retrieval and an evaluated final project. Use it for: Provides a structured first build before you connect an agent to a customer's systems. Practice: Build one tool-using agent, record its traces, then add a timeout and a test for a failed tool call. Access: Course and certification are free; hosted model use beyond included allowances may cost extra. Limits: Requires basic Python and basic LLM knowledge; use current framework docs when examples differ. Checked 2026-09-14. Read the official introduction, syllabus, prerequisites and certification terms. It describes a living course and requires a free Hugging Face account for Hub/Spaces activities. [Read the original resource](https://huggingface.co/learn/agents-course/unit0/introduction) ## LLM Bootcamp The Full Stack instructors · The Full Stack · course · intermediate A product-oriented course covering LLM application architecture, deployment, evaluation and user experience. Use it for: Connects model work with the application lifecycle an FDE actually owns. Practice: Use the LLMOps and UX lectures to write a launch plan with failure states, instrumentation and customer acceptance checks. Access: Recorded lectures are free; recreating hosted demos can incur provider charges. Limits: Spring 2023 material: use its lifecycle and UX lessons, then check current APIs, models and framework docs. Checked 2026-09-14. Read the official syllabus, lecture descriptions and explicit free-access statement; did not watch every lecture or run the historical demo. [Read the original resource](https://fullstackdeeplearning.com/llm-bootcamp/spring-2023/) ## Model Context Protocol specification — 2026-07-28 Model Context Protocol maintainers · Model Context Protocol · docs · advanced The authoritative contract for connecting AI applications with external context and tools. Use it for: Useful when designing interoperable customer integrations and understanding capability and trust boundaries. Practice: Expose one read-only customer-style tool, document its authorization boundary, and test unsupported protocol versions and errors. Access: Open specification and public documentation; connected services may have their own charges. Limits: Implementation requires API/JSON-RPC familiarity. Pin the protocol and SDK versions; older clients and optional extensions have different support. Checked 2026-09-14. Read the versioned specification overview, base protocol, features, extensions and security principles; checked official July release information. This entry does not claim every SDK implements every extension. [Read the original resource](https://modelcontextprotocol.io/specification/2026-07-28) ## Why AI evals are the hottest new skill for product builders Lenny Rachitsky with Hamel Husain and Shreya Shankar · Lenny's Podcast · podcast · beginner A worked conversation about analyzing real traces, grouping errors, choosing evaluators and validating a judge against human decisions. Use it for: Provides an accessible shared starting point for an FDE and a customer domain expert who must agree what acceptable quality means. Practice: Watch the trace-review walkthrough, then have an engineer and domain expert independently annotate the same ten outputs and discuss disagreements. Access: Episode links to public YouTube, Spotify and Apple Podcasts; the newsletter page marks its transcript/access as paid. Limits: The title's ranking language is promotional; this selection is based on the concrete walkthrough, not claims about course popularity. Checked 2026-09-14. Original episode page, public description and timestamped chapters read; official outbound YouTube link verified. Paid transcript was not accessed. [Read the original resource](https://www.lennysnewsletter.com/p/why-ai-evals-are-the-hottest-new-skill) ## Becoming a Forward Deployed Engineer: Bridging AI Products and Customer Outcomes Vinoo Ganesh · LinkedIn Learning · course · intermediate A 51-minute course on FDE responsibilities, discovery, communication, deployment feedback and getting started in the role. Use it for: Offers a compact FDE-specific orientation before tackling a substantial technical portfolio project. Practice: Turn the published syllabus into an engagement brief covering the customer outcome, communication plan, first deployment and feedback route to product engineering. Access: Subscription access with a public preview; trial eligibility and employer or library access vary. Limits: An orientation course, not a complete engineering qualification or hiring guarantee; locked lessons were not reviewed. Checked 2026-09-14. Official course landing page and full public syllabus read; instructor, intermediate level, 51-minute duration, release date and locked lessons verified. [Read the original resource](https://www.linkedin.com/learning/becoming-a-forward-deployed-engineer-bridging-ai-products-and-customer-outcomes) ## Candidate AI guidance Anthropic · Anthropic · docs · beginner Anthropic’s policy on AI assistance during applications and assessments. Use it for: Distinguish learning with AI from the rules that apply during an interview. Practice: Before a mock assessment, state the permitted tools and work within that constraint. Access: The candidate guidance is public. Limits: Employer-specific policy; the actual invitation may contain additional instructions. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://www.anthropic.com/candidate-ai-guidance) ## OpenAI interview guide OpenAI · OpenAI · docs · beginner General guidance on OpenAI’s recruiting assessments and interview process. Use it for: Prepare examples and confirm the tool rules for the actual assessment. Practice: Practise explaining a debugging decision without relying on a fixed question bank. Access: The employer’s public interview guide is free. Limits: A general employer guide, not a guaranteed FDE interview loop. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://openai.com/interview-guide/) ## Staff Engineer: Leadership beyond the management track Will Larson · Will Larson / StaffEng · book · advanced Guides and practitioner interviews about technical leadership, choosing useful work and influencing an organization as an individual contributor. Use it for: Senior FDEs frequently lead across customer and product teams without controlling either team's priorities. Practice: Read Work on what matters, then write a one-page proposal identifying the outcome, decision owner, affected teams and work you will deliberately defer. Access: Most guides and interviews are free on StaffEng; the assembled book, foreword and ending are paid. Limits: Staff engineering and FDE are different roles; the author also notes a US and Silicon Valley bias in the examples. Checked 2026-09-14. Official book page read, including free-content FAQ, guide links, interview basis and contextual limitation; prices intentionally omitted. [Read the original resource](https://staffeng.com/book/) ## The AI-native interview Sierra · Sierra · article · intermediate Sierra’s description of its AI-native engineering interview design. Use it for: Practise scoping, implementing, and reviewing a solution with explicit tool rules. Practice: Run a mock plan/build/review exercise and explain one deliberate scope cut. Access: The engineering article is freely accessible. Limits: Describes Sierra’s engineering process; other employers may prohibit the same tools. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://sierra.ai/blog/the-ai-native-interview) ## The Pragmatic Engineer Gergely Orosz · The Pragmatic Engineer · blog · intermediate An engineering publication combining company interviews, real-world delivery problems, organizational practices and career context. Use it for: Helps FDEs understand how engineering teams operate and compare customer practices with documented examples. Practice: Choose one relevant deep dive, extract a practice and its context, and write a small experiment to test whether it improves your current engagement. Access: Free article previews and podcast issues; subscription for full deep dives, The Pulse and subscriber resources. Limits: Use specific articles to solve a problem; the publication is not an FDE curriculum and industry commentary is not a universal benchmark. Checked 2026-09-14. Current official about page read: author, editorial scope, free versus paid access, schedule and subscription conditions verified. [Read the original resource](https://newsletter.pragmaticengineer.com/about) ## Anti-Corruption Layer pattern Microsoft Azure Architecture Center · Microsoft Learn · article · intermediate Places a translation boundary between systems with different domain models. Use it for: Keep a customer-specific schema from spreading through the reusable application. Practice: Map two customer APIs into one internal model and test what happens when a field changes. Access: Public engineering reference; implementation infrastructure may incur costs. Limits: The adapter introduces operational and maintenance overhead; preserve errors and ownership across the boundary. Checked 2026-09-14. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text. [Read the original resource](https://learn.microsoft.com/en-us/azure/architecture/patterns/anti-corruption-layer) ## CMU 15-445/645: Intro to Database Systems (Spring 2026) Andy Pavlo and Jignesh Patel · Carnegie Mellon University · course · advanced An implementation-focused study of SQL, indexes, query execution, concurrency, recovery, and distributed databases. Use it for: Deepens diagnosis of slow queries, transaction anomalies, and storage behavior in data-heavy customer deployments. Practice: Study the query-optimization and concurrency units, then explain a slow query and a transaction anomaly in your own small database example. Access: Public archived notes, slides, assignments, and video links; university enrollment and credit are separate. Limits: Advanced systems and C++17 skills are prerequisites for BusTub projects. Keep course solutions private; external learners do not receive the full grading infrastructure. Checked 2026-09-14. Read Spring 2026 home, complete schedule (updated April 27), and syllabus. Chosen over the in-progress Fall 2026 offering; video destinations are linked but playback was not verified. [Read the original resource](https://15445.courses.cs.cmu.edu/spring2026/schedule.html) ## Data Engineering Zoomcamp Alexey Grigorev and DataTalks.Club instructors · DataTalks.Club · course · intermediate A project-based data-pipeline course covering containers, infrastructure, orchestration, warehouses, modeling, batch, and streaming. Use it for: Builds hands-on integration skills that transfer to ingesting and validating customer data under real operational constraints. Practice: Deliver an original pipeline with incremental loading, data-quality checks, recovery instructions, and a clear cost boundary. Access: Free videos, materials, and homework; cloud services used in exercises can create separate costs. Limits: Basic coding and SQL are required. Self-paced study is available now without a certificate; cohort certificates require the provider's project and peer-review process. Checked 2026-09-14. Read the maintainers' current repository README, prerequisites, access table, and syllabus. It advertises January 2027 as the next cohort; live cohort means coordinated deadlines, not live lectures. [Read the original resource](https://github.com/DataTalksClub/data-engineering-zoomcamp) ## Design and develop a RAG solution Microsoft Azure Architecture Center contributors · Microsoft Learn · docs · intermediate A design-and-evaluation framework for RAG ingestion, chunking, enrichment, embeddings and retrieval. Use it for: Helps turn a customer document demo into a traceable series of engineering decisions. Practice: Keep an experiment log that changes one retrieval variable at a time and records both component and end-user results. Access: Public architecture guidance; Azure resources used to implement it are separately billed. Limits: Azure examples are vendor-specific; transfer the evaluation method without assuming every named service is required. Checked 2026-09-14. Read the primary architecture introduction, design phases, evaluation recommendations and update date of 30 Jun 2026. Did not read every linked phase article. [Read the original resource](https://learn.microsoft.com/en-us/azure/architecture/ai-ml/guide/rag/rag-solution-design-and-evaluation-guide) ## Designing Data-Intensive Applications, 2nd Edition Martin Kleppmann and Chris Riccomini · O’Reilly · book · advanced Develop judgment about storage, data models, transactions, replication, distributed failures, batch, and streaming trade-offs. Use it for: Use it when a customer's consistency, tenancy, recovery, or scale requirements make a simple architecture choice consequential. Practice: Write a design decision for your project comparing one database with a distributed alternative, including partial-failure behavior. Access: Published book and subscription access; the second edition is released, not an early-access recommendation. Limits: Best after building a backend with a database. Read relevant chapters alongside a project instead of treating all 672 pages as an entry requirement. Checked 2026-09-14. Read the publisher's second-edition page and contents: February 2026, Kleppmann and Riccomini. The older dataintensive.net marketing page still describes the earlier author/version. Description is based on public publisher/author material, not a full-text review. [Read the original resource](https://www.oreilly.com/library/view/designing-data-intensive-applications/9781098119058/) ## Fundamentals of Data Engineering Joe Reis and Matt Housley · O’Reilly · book · beginner An end-to-end framework for ingestion, storage, transformation, orchestration, governance, and serving data. Use it for: Helps translate a customer's messy data estate into a coherent delivery plan before selecting products. Practice: Map one customer's data lifecycle, identify its source of truth, and write a failure-and-ownership table for every handoff. Access: Buy the book or use O’Reilly access through a personal, employer, or library subscription where available. Limits: Basic programming and database concepts help. The 2022 concepts remain useful, but verify contemporary tool behavior in current documentation. Checked 2026-09-14. Read the publisher overview and title page, which list June 2022 and the two authors. No claim that this is a newly updated edition. Description is based on public publisher/author material, not a full-text review. [Read the original resource](https://www.oreilly.com/library/view/fundamentals-of-data/9781098108298/) ## Making retries safe with idempotent APIs Malcolm Featonby · Amazon Builders’ Library · article · intermediate Explores request identity, duplicate calls, and the semantics of retry-safe APIs. Use it for: Avoid duplicate invoices, tickets, or provisioning when a response is lost. Practice: Repeat one write with the same request ID, then reuse that ID with different input and verify rejection. Access: Public engineering reference; implementation infrastructure may incur costs. Limits: Idempotency needs a defined lifetime, scope, and consistency contract; it is not the same as assuming exactly-once delivery. Checked 2026-09-14. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text. [Read the original resource](https://aws.amazon.com/builders-library/making-retries-safe-with-idempotent-APIs/) ## Strangler Fig Martin Fowler · Martin Fowler · article · intermediate Describes incremental replacement of a legacy system through a growing replacement. Use it for: Deliver a narrow useful slice without requiring a full customer-system rewrite. Practice: Move one read workflow behind a routing boundary and document a reversible cutover. Access: Public engineering reference; implementation infrastructure may incur costs. Limits: Temporary coexistence creates complexity; establish how the old path will eventually be retired. Checked 2026-09-14. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text. [Read the original resource](https://martinfowler.com/bliki/StranglerFigApplication.html) ## Continuous Discovery Habits Teresa Torres · Product Talk · book · intermediate A structured approach to outcomes, customer interviews, opportunity mapping and testing assumptions before committing to solutions. Use it for: Provides a repeatable way to connect customer needs, engineering choices and measurable value throughout an engagement. Practice: Draw one opportunity solution tree for your deployment: one outcome, three observed customer needs, two candidate solutions and the riskiest assumption to test. Access: Paid book; the author's public articles offer a free introduction to the methods. Limits: Written for ongoing product teams; adapt its cadence and ownership model for short customer engagements. Checked 2026-09-14. Author's dated book launch page and current Product Talk course/book navigation read; book scope and purchase access verified. Paid book text was not reviewed. [Read the original resource](https://www.producttalk.org/continuous-discovery-habits/) ## Every Agent Needs a Box — Aaron Levie, Box swyx and Jeff Huber with Aaron Levie · Latent Space · podcast · intermediate An enterprise software conversation about agent identity, permissions, context, evaluation and workflows that read and write company information. Use it for: Helps frame customer discovery around the systems and operational boundaries that determine whether an agent can be deployed. Practice: After the governance and identity segment, draw separate access boundaries for the employee, agent and administrator in your proposed workflow. Access: Public episode, video and transcript on the publisher's site. Limits: A vendor executive's viewpoint, with product promotion and forecasts; it is not an implementation guide or independent security assessment. Checked 2026-09-14. Official episode page, chapter list and transcript sections on identity, permissions and enterprise data read; publication date and public access verified. [Read the original resource](https://www.latent.space/p/box) ## Everyone Can Do Continuous Discovery—Even You! Here's How Teresa Torres · Product Talk · article · beginner How to begin customer discovery in organizations that lack an established discovery cadence or resist changing their process. Use it for: An FDE needs a practical first step even when access to users or influence over the customer's roadmap is limited. Practice: Choose one discovery habit you can introduce this week; specify who participates, the question it should answer and the smallest useful evidence to collect. Access: Free article and conference talk; the related book and training are optional paid products. Limits: Organizational examples are contextual and predate current AI tooling; the customer-learning methods remain distinct from technical implementation advice. Checked 2026-09-14. Original author article and edited conference transcript read; publication date, topic and free video/text access verified. [Read the original resource](https://www.producttalk.org/getting-started-with-discovery/) ## Technical Writing Courses for Engineers Google technical-writing educators · Google for Developers · course · beginner Practice planning clear technical documents, editing prose, improving accessibility, and writing helpful errors. Use it for: Useful for customer setup instructions, operational runbooks, design decisions, and unambiguous handoffs. Practice: Rewrite a setup guide for a named audience, then ask someone unfamiliar with the project to use it and report friction. Access: Free self-study lessons and facilitator materials; occasional public facilitated sessions have separate availability. Limits: Some English-writing proficiency and basic coding familiarity help. These courses focus on technical documentation rather than general business English. Checked 2026-09-14. Read Google's course home, audience requirements, and delivery-format guidance. Self-study lessons remain valuable without a facilitated session; no scheduled class or certification is promised. [Read the original resource](https://developers.google.com/tech-writing) ## The Mom Test Rob Fitzpatrick · Rob Fitzpatrick · book · beginner A compact guide to customer conversations that produce useful evidence instead of flattering, biased feedback. Use it for: FDEs need to uncover an expensive workflow problem before agreeing to a requested feature or agent demo. Practice: Rewrite five leading discovery questions, then interview a user about the last time they performed the workflow; record evidence separately from assumptions. Access: Paid print, ebook, PDF and audiobook editions; the official page explains the approach before purchase. Limits: Customer interviewing is one part of enterprise discovery; add security, procurement, stakeholder and deployment constraints. Checked 2026-09-14. Official author/book page read: scope, author, formats and purchase links verified; no publication date displayed and no price quoted. Practice is an editorial exercise. Description is based on public publisher/author material, not a full-text review. [Read the original resource](https://www.momtestbook.com/) ## The Workshop Survival Guide Rob Fitzpatrick and Devin Hunt · Rob Fitzpatrick and Devin Hunt · book · intermediate Practical workshop design and facilitation guidance covering exercises, teaching formats, difficult questions and time management. Use it for: Useful when an FDE must run customer onboarding, train operators or bring several stakeholders into a working session. Practice: Design a 45-minute customer enablement session with one observable learning outcome, a hands-on task, a recovery plan and a short feedback check. Access: Paid book; a detailed table of contents is public on the official site. Limits: Focuses on educational workshops, so adapt the format for discovery meetings rather than treating every meeting as a class. Checked 2026-09-14. Official book page, authors section, complete table of contents and purchase link read; publication date not displayed. Full book not reviewed. [Read the original resource](https://www.workshopsurvival.com/) ## Agentic Engineering Patterns Simon Willison · Simon Willison's Weblog · blog · intermediate A living collection of coding-agent practices covering testing, Git, code review, subagents and understanding generated code. Use it for: FDEs can prototype integrations quickly while retaining responsibility for the code they deliver and hand over. Practice: Use an agent on a small integration change, establish a test baseline, inspect the diff, and produce a walkthrough another engineer can use to maintain it. Access: Free, evolving online guide; using the coding tools discussed can incur separate costs. Limits: A practitioner guide in progress, not a formal standard; tool behavior and individual chapters can change. Checked 2026-09-14. Official guide index, introductory chapter and dated launch post read; current index includes testing, review and code-understanding patterns. [Read the original resource](https://simonwillison.net/guides/agentic-engineering-patterns/) ## Architecture Patterns with Python Harry Percival and Bob Gregory · O’Reilly / Cosmic Python · book · intermediate Concrete Python examples of domain models, adapters, repositories, transactions, service layers, and events. Use it for: Helps separate changing customer integrations from stable business logic and make those boundaries testable. Practice: Wrap an external API in an adapter, test business behavior with a fake, then integration-test the real boundary. Access: Complete authorized online book is free; paid print, ebook, and O’Reilly options also exist. Limits: Requires Python and basic testing familiarity. The 2020 examples may need dependency updates; introduce only the architectural layers your project needs. Checked 2026-09-14. Read the authors' Cosmic Python home and preface plus O’Reilly metadata. Authors explicitly offer free authorized reading; publisher confirms March 2020. [Read the original resource](https://www.cosmicpython.com/book/preface.html) ## FastAPI tutorial FastAPI · FastAPI · docs · intermediate An incremental Python API tutorial with validation, security, and testing references. Use it for: Connect a workflow to an explicit request/response contract and identity boundary. Practice: Add an endpoint and tests that deny a user access to another user’s record. Access: Documentation is free; hosting the resulting service can incur costs. Limits: Requires Python fluency; tutorial authentication examples still need deployment-specific review. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://fastapi.tiangolo.com/tutorial/) ## Full Stack Open Matti Luukkainen and contributing instructors · University of Helsinki · course · intermediate Build web applications through connected exercises covering React, Node, APIs, testing, TypeScript, delivery, and databases. Use it for: Develops the breadth needed to ship a usable customer workflow across frontend, backend, and deployment boundaries. Practice: Complete the core application, then add one original customer workflow with permissions, tests, and a handoff document. Access: Free materials and certificate; university credit follows the provider's registration and completion requirements. Limits: Requires fluent programming plus basic web, database, and Git knowledge; it is not an introduction to programming. Checked 2026-09-14. Read the official home and https://fullstackopen.com/en/part0/general_info/. The course now updates continuously rather than using yearly editions; prerequisites and free access are explicit. [Read the original resource](https://fullstackopen.com/en/) ## Learn web development MDN · MDN · docs · beginner Guided web fundamentals covering the browser, HTML, CSS, and JavaScript. Use it for: Build a usable interface for a workflow rather than stopping at a script. Practice: Implement loading, empty, success, and error states for an API-backed screen. Access: Learning articles are freely accessible. Limits: A web-development pathway; backend operations and customer discovery need other resources. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://developer.mozilla.org/en-US/docs/Learn_web_development) ## Machine Learning Crash Course Google · Google for Developers · course · beginner A practical grounding in models, data, classification metrics, embeddings and production ML. Use it for: Helps explain false positives, data leakage and model limitations to a customer. Practice: Train a small classifier, vary its threshold, and explain the business cost of each error type. Access: Public lessons and browser exercises; a Google account may be needed for Colab. Limits: No prior ML is required, but programming, algebra and basic statistics are expected; Python, NumPy and pandas prework is provided. Checked 2026-09-14. Read Google's course modules and prerequisites page; prerequisite page updated 25 Aug 2025. The course is a concept foundation, not deep API training. [Read the original resource](https://developers.google.com/machine-learning/crash-course) ## Pro Git Scott Chacon & Ben Straub · Scott Chacon & Ben Straub · book · beginner A practical reference for version control, branching, and collaboration. Use it for: Make your customer work reviewable and recoverable in a shared repository. Practice: Create a branch, resolve a merge conflict, and safely revert a faulty change. Access: The full authorised web edition is free; print editions are sold separately. Limits: Use chapters selectively; advanced internals are not a prerequisite for a first project. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://git-scm.com/book/en/v2) ## Software Engineering at Google Titus Winters, Tom Manshreck, and Hyrum Wright · O’Reilly / Google Abseil · book · intermediate A practical account of maintaining software through testing, documentation, code review, collaboration, and change. Use it for: Helps turn a customer-specific prototype into code another engineer can review, operate, and extend. Practice: Add a concise design document, review checklist, and handoff guide to a project; ask a peer to follow the guide unaided. Access: Complete authorized HTML edition is free; optional print purchase. Limits: Read after some coding experience. Adapt Google's scale-specific practices to the team; the book explicitly omits detailed API design and security hardening. Checked 2026-09-14. Read Google's Abseil landing page and book preface. Google confirms March 2020 publication and authorized free HTML availability. [Read the original resource](https://abseil.io/resources/swe-book) ## The Design of Web APIs, Second Edition Arnaud Lauret · Manning · book · intermediate Consumer-centered API design, from requirements and capabilities through HTTP contracts, documentation, security, and compatible change. Use it for: Useful for turning an integration request into an API contract that customers and product engineers can both maintain. Practice: Design and review an OpenAPI contract for a long-running customer import, including permissions, errors, and version evolution. Access: Paid ebook, print, or Manning subscription; excerpts and example source are available separately. Limits: The provider expects experience building or consuming APIs. The 2025 edition is current book guidance; verify later OpenAPI specifications and tooling separately. Checked 2026-09-14. Read Manning's second-edition page, June 2025 metadata, reader prerequisites, chapter description, and linked code/exercise resources. Chosen over an older API-pattern catalog for its requirements-to-contract coverage. Description is based on public publisher/author material, not a full-text review. [Read the original resource](https://www.manning.com/books/the-design-of-web-apis-second-edition) ## AI Evals For Engineers & PMs Hamel Husain and Shreya Shankar · Maven / Parlance Labs · course · intermediate A coached course on evaluation workflows, annotation tools, targeted tests and stakeholder alignment. Use it for: Most useful when you have a live AI project and want feedback on its evaluation process. Practice: Bring a scoped project and anonymized examples; finish with a rubric, validated checks and a review workflow. Access: Paid live cohort. Read the free companion FAQ first; check the current cohort, schedule and fee before purchasing. Limits: Familiarity with AI applications and LLMs is helpful; a substantial cohort purchase is optional, not a prerequisite for becoming an FDE. Checked 2026-09-14. Read the official public course syllabus and FAQ, which describe office hours, exercises, projects and lifetime material access. Full teaching content not accessed; no outcome guarantee or fixed fee asserted. [Read the original resource](https://maven.com/parlance-labs/evals) ## Circuit Breaker pattern Microsoft Azure Architecture Center · Microsoft Learn · article · intermediate A pattern for temporarily stopping calls to a failing dependency. Use it for: Bound cascading failure and allow an overloaded integration time to recover. Practice: Simulate an outage, open the circuit after a threshold, and exercise a recovery probe. Access: Public engineering reference; implementation infrastructure may incur costs. Limits: Adds state and tuning; it may be unnecessary for a simple workflow with existing queue-based recovery. Checked 2026-09-14. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text. [Read the original resource](https://learn.microsoft.com/en-us/azure/architecture/patterns/circuit-breaker) ## Demystifying evals for AI agents Anthropic Engineering · Anthropic · article · intermediate A practical guide to agent tasks, trials, graders, traces and observable outcomes. Use it for: Provides language for agreeing with a customer on what successful deployment actually means. Practice: Create outcome checks for a tool-using agent, run repeated trials, and inspect both the trace and final system state. Access: Public article; running model-based graders has usage costs. Limits: Examples come from particular agent types; adapt criteria to your domain and calibrate model graders with people. Checked 2026-09-14. Read the primary engineering article, evaluation terminology, grader discussion and development/monitoring recommendations. [Read the original resource](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents) ## Evaluation best practices OpenAI · OpenAI API documentation · docs · intermediate Task-specific evaluation design, dataset construction, grader choices and failure cases for AI applications. Use it for: A second provider perspective for reviewing your acceptance tests and human calibration process. Practice: Map each uncertain step in your application to a failure case, an evaluation method and a human owner. Access: Public conceptual guidance. Model calls and hosted services have separate usage fees. Limits: The page announces Evals-platform read-only access on 31 Oct 2026 and shutdown on 30 Nov 2026. Use the concepts; do not start a new dependency on that retiring platform. Checked 2026-09-14. Read the current public guide body, including its explicit Evals deprecation notice, design process, architecture examples and grader caveats. Not a recommendation to adopt the retiring API. [Read the original resource](https://developers.openai.com/api/docs/guides/evaluation-best-practices) ## Excessive Agency OWASP GenAI Security Project · OWASP · docs · intermediate A threat-model reference for excessive functionality, permissions, and autonomy in AI systems. Use it for: Design customer tool access so a model cannot exceed the intended operation or user authority. Practice: Remove unnecessary tools, deny a cross-user operation, and require approval before a consequential write. Access: Public security guidance; no certification or paid product is required to read it. Limits: A risk framework, not a complete security audit; enforce restrictions in the application and downstream systems. Checked 2026-09-14. Official OWASP risk description and mitigations read; practice is an editorial exercise. [Read the original resource](https://genai.owasp.org/llmrisk/llm062025-excessive-agency/) ## Get started with Docker Docker · Docker · docs · intermediate An introduction to containers and reproducible application environments. Use it for: Help another engineer run the integration without relying on your laptop. Practice: Package a service, document configuration, and run it from a fresh checkout. Access: Documentation is free. Check Docker product licensing and hosting costs for your context. Limits: Containers are one deployment tool, not a substitute for permissions, monitoring, or rollback. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://docs.docker.com/get-started/) ## Incident Review and Postmortem Best Practices Gergely Orosz · The Pragmatic Engineer · article · intermediate An original survey and discussion of incident handling, review, follow-up actions and learning practices across engineering teams. Use it for: Gives an FDE vocabulary for communicating an incident and leaving the customer with useful follow-up work instead of only a technical fix. Practice: For a simulated outage, write the customer impact, timeline, contributing conditions and two follow-up actions with owners and verification criteria. Access: The article is a free edition; linked subscriber-only templates require payment. Limits: Published in 2021 and explicitly affected by survey selection bias; treat its survey numbers and vendor examples as historical. Checked 2026-09-14. Original free article read, including date, author, survey limitations, incident sequence and distinction between free content and paid templates. [Read the original resource](https://newsletter.pragmaticengineer.com/p/incident-review-best-practices) ## Made With ML Goku Mohandas · Made With ML / Anyscale · course · intermediate An end-to-end production ML curriculum spanning data, training, testing, serving, CI/CD and monitoring. Use it for: Shows how to turn a notebook into a maintainable customer deployment. Practice: Package a model behind an API, add data and model tests, and demonstrate a reproducible deployment. Access: Public written lessons and code; local CPU setup is documented. Optional cloud infrastructure costs extra. Limits: Python familiarity helps. Setup examples reference older Python/Ray versions, so review dependency compatibility before reuse. Checked 2026-09-14. Read the official course overview and setup lesson, including local no-GPU instructions and the 2023 citation. Did not validate every dependency or cloud workflow. [Read the original resource](https://madewithml.com/) ## Observability signals OpenTelemetry · OpenTelemetry · docs · intermediate A conceptual guide to traces, metrics, logs, and their role in observability. Use it for: Connect a customer-visible failure to the services and requests involved. Practice: Trace one failed workflow and correlate its request ID with an actionable log. Access: The specification and documentation are open; telemetry backends may charge. Limits: Start with useful signals; collecting everything can raise cost and expose sensitive data. Checked 2026-09-14. Official course, documentation, author, or employer page read on the check date. [Read the original resource](https://opentelemetry.io/docs/concepts/signals/) ## SE Radio 706: Observability Tool Migration Techniques Brijesh Ammanath with Yechezkel Rabinovich · Software Engineering Radio / IEEE · podcast · advanced A discussion of migrating dashboards, monitors, pipelines and integrations while validating that production signals still work. Use it for: FDE deployments often enter legacy environments where preserving existing operational behavior matters as much as adding a new feature. Practice: Write a migration checklist with an inventory, old/new signal comparison, owner for each alert, acceptance criteria and rollback conditions. Access: Public audio, show notes and automatically generated transcript. Limits: The guest co-founded an observability vendor; separate migration techniques from product claims and account for transcript errors. Checked 2026-09-14. Original IEEE episode page, show notes and transcript read; episode number, guest, host, date and OpenTelemetry/migration scope verified. [Read the original resource](https://se-radio.net/2026/02/se-radio-706-yechezkel-chez-rabinovich-on-observability-tool-migration-techniques/) ## Timeouts, retries, and backoff with jitter Marc Brooker · Amazon Builders’ Library · article · intermediate Explains why retry behaviour needs deadlines, limits, and randomisation. Use it for: Keep a failing customer dependency from consuming the whole service. Practice: Inject a timeout, cap retry attempts, and compare synchronised retries with jitter. Access: Public engineering reference; implementation infrastructure may incur costs. Limits: A retry is safe only when the operation and its side effects allow it. Start with SDK behaviour before adding another retry layer. Checked 2026-09-14. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text. [Read the original resource](https://d1.awsstatic.com/builderslibrary/pdfs/timeouts-retries-and-backoff-with-jitter.pdf) --- # FDE simulation missions: play Deployment Run Take responsibility for a customer deployment. Inspect evidence, choose an approach, review its consequence, and leave a service the customer can operate. All names, messages, timings, and outcomes are fictional. [Play Deployment Run](https://fdelab.dev/#game) ## Your three customers ### The Monday queue Northstar Services — Dispatch operations Get repair requests to eligible technicians faster. Every assignment still needs a dispatcher’s approval. [Read the Northstar Services mission brief](https://fdelab.dev/missions/northstar/) ### The missing invoice Cedar Supply — Finance operations Reduce invoice matching work. The pilot may create draft matches; payments and bank-detail changes stay outside its authority. [Read the Cedar Supply mission brief](https://fdelab.dev/missions/cedar/) ### A very confident answer Harbor Engineering — Internal knowledge Help staff find the right procedure faster. Preserve source authority, current versions, and each caller’s access. [Read the Harbor Engineering mission brief](https://fdelab.dev/missions/harbor/) ## What a delivery must prove - Finish five decisions within 14 simulated days. There is no real-time timer. - End with customer trust of at least 60 and reliability of at least 70. - Recover the incident and demonstrate handoff to an owner and backup. - A smaller supported service or controlled pause can count as a successful delivery. - Only the best result for each mission contributes to campaign XP. [Read the full scoring and provenance notes](https://fdelab.dev/research/deployment-run.md) --- # The Monday queue: an FDE dispatch operations simulation Northstar Services is a fictional customer. Your contact is Maya Chen, dispatch lead. Get repair requests to eligible technicians faster. Every assignment still needs a dispatcher’s approval. [Play or resume this mission](https://fdelab.dev/?mission=northstar#game) ## Your first conversation We spend 18 minutes routing a request. The sponsor wants an AI demo by Friday. I need something my dispatchers can actually use on Monday. ## Evidence to inspect Five recent requests: Most time was spent checking technician eligibility, not classifying the request. Access: Eligibility records live in three systems. Write permissions are not confirmed. Acceptance: Maya will accept the pilot. A dispatcher must approve every assignment. ## Five decisions in the engagement - Discover: Discover the expensive handoff before choosing what to automate. - Connect: A retry needs a stable business identity and a known write boundary. - Evaluate: Test the resulting workflow, including the checker and the operator’s review effort. - Recover: A missing response is not proof that the operation failed. - Hand off: Ownership is demonstrated by the receiving team, not transferred by sending a document. ## Delivery gates - Finish five decisions within 14 simulated days. There is no real-time timer. - End with customer trust of at least 60 and reliability of at least 70. - Recover the incident and demonstrate handoff to an owner and backup. - A smaller supported service or controlled pause can count as a successful delivery. - Only the best result for each mission contributes to campaign XP. [Explore all three missions](https://fdelab.dev/missions/) [Build the underlying skills](https://fdelab.dev/learn/) --- # The missing invoice: an FDE finance operations simulation Cedar Supply is a fictional customer. Your contact is Elena Ruiz, accounts-payable lead. Reduce invoice matching work. The pilot may create draft matches; payments and bank-detail changes stay outside its authority. [Play or resume this mission](https://fdelab.dev/?mission=cedar#game) ## Your first conversation Each weekly batch takes us 6.5 hours. The sponsor says automate everything. Most of my time goes on missing receipt evidence, not typing fields. ## Evidence to inspect Weekly batch: Exact matches, partial receipts, credit notes, and disputed quantities. Authority: Elena accepts draft matching. No permission has been granted for payments. Bottleneck: Missing receipt evidence creates most of the exception work. ## Five decisions in the engagement - Discover: Scope an outcome and its authority before making a delivery commitment. - Connect: Use business identity and durable state to reason about repeated operations. - Evaluate: A correct extraction does not prove a correct business result. - Recover: Establish destination state before replaying a business operation. - Hand off: Transfer operating capability and define what requires another engagement. ## Delivery gates - Finish five decisions within 14 simulated days. There is no real-time timer. - End with customer trust of at least 60 and reliability of at least 70. - Recover the incident and demonstrate handoff to an owner and backup. - A smaller supported service or controlled pause can count as a successful delivery. - Only the best result for each mission contributes to campaign XP. [Explore all three missions](https://fdelab.dev/missions/) [Build the underlying skills](https://fdelab.dev/learn/) --- # A very confident answer: an FDE internal knowledge simulation Harbor Engineering is a fictional customer. Your contact is Owen Patel, operations manager. Help staff find the right procedure faster. Preserve source authority, current versions, and each caller’s access. [Play or resume this mission](https://fdelab.dev/?mission=harbor#game) ## Your first conversation Finding and verifying a procedure takes nine minutes. The old wiki, chat advice, and approved instructions disagree. People want a chatbot. ## Evidence to inspect Observed delay: Finding a page is quick. Establishing which version applies is slow. Source disagreement: Approved procedures, old wiki copies, and informal chat answers coexist. Acceptance owner: Owen can identify the procedure owner and review expected behavior. ## Five decisions in the engagement - Discover: More context is not useful until its authority and purpose are clear. - Connect: Retrieval rank is not source authority; a citation does not grant access. - Evaluate: Evaluate the answer for this actor, source version, and required action. - Recover: Recover affected decisions as well as refreshing the source index. - Hand off: Knowledge ownership includes update, access, and recovery behavior. ## Delivery gates - Finish five decisions within 14 simulated days. There is no real-time timer. - End with customer trust of at least 60 and reliability of at least 70. - Recover the incident and demonstrate handoff to an owner and backup. - A smaller supported service or controlled pause can count as a successful delivery. - Only the best result for each mission contributes to campaign XP. [Explore all three missions](https://fdelab.dev/missions/) [Build the underlying skills](https://fdelab.dev/learn/) --- # Use the FDE game and guide with an AI agent An agent can read the same learning material as a person and, in a supporting browser, work with the visible game. Game tools read state, open a mission, commit one chosen decision, continue after feedback, and search lessons. ## Read and discover - [Agent guide and tool contract](https://fdelab.dev/agent-guide.md) - [Concise content index](https://fdelab.dev/llms.txt) - [Complete readable guide](https://fdelab.dev/llms-full.txt) - [Agent resource discovery](https://fdelab.dev/.well-known/ard.json) - [Available coaching skills](https://fdelab.dev/.well-known/agent-skills/index.json) - [FDE coaching skill](https://fdelab.dev/.well-known/agent-skills/fde-coach/SKILL.md) ## Structured material - [Learning curriculum](https://fdelab.dev/data/curriculum.json) - [Paired readings](https://fdelab.dev/data/reading.json) - [Resource catalogue](https://fdelab.dev/data/resources.json) - [Mission briefs](https://fdelab.dev/data/missions.json) ## Operate the game WebMCP is a progressive enhancement. A supporting browser can register the game tools when the game is open; there is no remote MCP endpoint. Mutations use the same game state and visible feedback as the ordinary controls. - Read current mission, decision, evidence, and costs before acting. - Commit only the choice the learner selects unless an automated demonstration is requested. - Review the consequence before continuing. Stale actions are rejected. - HTTP resources are read-only. They do not expose personal browser progress. ## Access and source limits Hosting permissions apply equally to people and agents. A 401 response means this visitor needs site access. Discovery files do not bypass sign-in. Scenarios are fictional; the cited books and sources keep their own access and reuse terms. [Open the game](https://fdelab.dev/#game) --- # About Deployment Run and the FDE Field Guide Deployment Run combines original fictional customer missions with a practical guide to forward deployed engineering. It is an independent reference and is not affiliated with the companies or publishers it cites. ## Maintainer Illia S. Hrybko - [Email Illia S. Hrybko](mailto:ilsegribko@gmail.com) - [LinkedIn](https://www.linkedin.com/in/illia-hrybko/) ## How the material was developed The game and exercises adapt the reviewed themes of customer discovery, explicit delivery scope, evaluation, recovery, and demonstrated ownership. They do not reproduce either source book or claim professional accreditation. - [Source-method guide](https://fdelab.dev/research/source-methods.md) - [Coverage and interpretation limits](https://fdelab.dev/research/material-review.md) - [Game rules and provenance](https://fdelab.dev/research/deployment-run.md) ## Your progress Game attempts, learning milestones, saved resources, and preferences are stored in your browser. The site has no server-side progress database. External source links and hosted fonts contact their respective providers. Hosting permissions are separate from game functionality.