Production Engineering with AI Podcast

Editorial draft adapted from the Turkish manuscript. No English book source or generated media is available.

Complete chapter/volume links and actual audio details before publication. The AI disclosure is draft publication copy to verify after production.

AI-Native Teams: Decision Rights, Productivity, and Phased Transition | Vol 4 Ch 15

```text

AI teams make different decisions about task ownership, risk acceptance, and technical review. Vardiya assesses competence beyond training attendance and distinguishes shared platform responsibilities from local team responsibilities. This makes it possible to identify who should investigate a service problem and which decisions they can make.

Adem YÜCE • Production Engineering with AI • Volume 4: AI-Native Production • Chapter 15: AI-Native Teams and Organizational Transformation

Topics covered in this chapter:

- Roles and Decision Rights

- Competence and Review Culture

- Shared Platforms and Local Team Boundaries

- Productivity and Comparison Baselines

- A Phased 90-Day Transition

- Defending Vardiya's Production File

AI transformation results depend on accepted work and total review effort under comparable conditions. Vardiya's phased plan ties progress to evidence requirements. Its final production file brings earlier records together without treating open risks as resolved merely because the schedule or the book has reached its end.

```

Production Engineering with AI Podcast · Season 4 · Vol 4 Ch 15

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/#chapter-15

Read the book: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Editorial draft adapted from the Turkish manuscript. No English book source or generated media is available.

Complete chapter/volume links and actual audio details before publication. The AI disclosure is draft publication copy to verify after production.

AI Deployment and Rollback: Compatibility, Canary Releases, and Recovery | Vol 4 Ch 14

```text

An AI release combines versions of code, prompts, models, and data components. Vardiya checks compatibility with existing task records, blocks external actions in shadow execution, and permits limited real use in a canary. Application controls must enforce the different purposes and authority boundaries of these stages.

Adem YÜCE • Production Engineering with AI • Volume 4: AI-Native Production • Chapter 14: Deployment, Change, and Rollback Management

Topics covered in this chapter:

- Behavior Components That Need Versioning

- Release Components and Compatibility Matrices

- Shadow Execution, Canary Releases, and Staged Rollout

- Stopping on Evaluation and Service Signals

- Rollback, Data Migration, Disaster Recovery, and Compensation

- Vardiya's Release Exercise

Stopping a release requires considering quality results alongside live service signals. Vardiya separates data compatibility from completed external actions during rollback, because reverting code does not retract a sent notification. The release exercise compares recovery targets with observed results and keeps actions requiring compensation visible throughout recovery.

```

Production Engineering with AI Podcast · Season 4 · Vol 4 Ch 14

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/#chapter-14

Read the book: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Editorial draft adapted from the Turkish manuscript. No English book source or generated media is available.

Complete chapter/volume links and actual audio details before publication. The AI disclosure is draft publication copy to verify after production.

AI Unit Economics: Tokens, Human Review, Capacity, and Caching | Vol 4 Ch 13

```text

AI cost needs to be related to accepted outcomes. Vardiya includes failed calls, tool use, human review, and graph operations, while distinguishing quotas, budgets, and concurrency in capacity planning. Token prices alone cannot explain service economics without considering queue behavior and demand during peak load.

Adem YÜCE • Production Engineering with AI • Volume 4: AI-Native Production • Chapter 13: Cost, Capacity, and Unit Economics

Topics covered in this chapter:

- Cost per Accepted Outcome

- Token, Tool, Human Review, and Graph Costs

- Capacity, Queues, and Peak Load

- The Tradeoffs of Caching and Batching

- Quality, Cost, and Energy Measurement Limits

- Vardiya's Budget Exercise

Caching and batching can change cost and latency without removing freshness or authorization requirements. Vardiya's budget exercise compares quality, accepted work, and spending together. Energy requires its own measurement, so lower prices or faster responses cannot be presented as direct evidence of reduced consumption or emissions.

```

Production Engineering with AI Podcast · Season 4 · Vol 4 Ch 13

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/#chapter-13

Read the book: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Editorial draft adapted from the Turkish manuscript. No English book source or generated media is available.

Complete chapter/volume links and actual audio details before publication. The AI disclosure is draft publication copy to verify after production.

AI Observability: Trace IDs, SLOs, Error Budgets, and Alerts | Vol 4 Ch 12

```text

Observability connects a task across requests and tool operations. Vardiya links identifiers while limiting telemetry fields and sensitive content. It assesses latency, task outcomes, and graph freshness together, because looking only at completed calls can hide queued work and answers that rely on outdated information.

Adem YÜCE • Production Engineering with AI • Volume 4: AI-Native Production • Chapter 12: AI Observability and Service Objectives

Topics covered in this chapter:

- Task, Request, and Trace Identifiers

- Telemetry Fields and Redaction

- Quality, Latency, and Graph Freshness

- SLIs, SLOs, and Error Budgets

- Alerts, Triage, and Incident Exercises

- Vardiya's Service Dashboard

Service indicators, targets, and error budgets serve different purposes. Vardiya defines the included tasks and measurement window before interpreting them. Alerts specify user impact, response ownership, and the first investigation step, so the dashboard provides context for intervention rather than simply displaying metrics in different colors.

```

Production Engineering with AI Podcast · Season 4 · Vol 4 Ch 12

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/#chapter-12

Read the book: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Editorial draft adapted from the Turkish manuscript. No English book source or generated media is available.

Complete chapter/volume links and actual audio details before publication. The AI disclosure is draft publication copy to verify after production.

AI Governance: Risk Ownership, Policy Controls, and Expiring Exceptions | Vol 4 Ch 11

```text

AI governance makes system purpose and decision responsibility explicit. Vardiya connects its inventory, impact assessment, and acceptance boundary with the application control enforcing each policy requirement. This distinguishes having a written rule from demonstrating that the rule is actually enforced in the relevant system version.

Adem YÜCE • Production Engineering with AI • Volume 4: AI-Native Production • Chapter 11: AI Governance, Policy, and Risk Management

Topics covered in this chapter:

- System Inventory and Risk Ownership

- Impact Assessment and Acceptance Boundaries

- Mapping Policies to Controls

- Exception and Approval Lifecycles

- Incident Responsibility and Evidence Chains

- Vardiya's Governance Check

Exceptions need defined scope, ownership, and expiry. After an incident, Vardiya connects the affected task with the active policy version and relevant decisions. Governance checks examine those relationships, without treating illustrative book records as actual human approval or allowing unresolved responsibilities to disappear from acceptance records.

```

Production Engineering with AI Podcast · Season 4 · Vol 4 Ch 11

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/#chapter-11

Read the book: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Editorial draft adapted from the Turkish manuscript. No English book source or generated media is available.

Complete chapter/volume links and actual audio details before publication. The AI disclosure is draft publication copy to verify after production.

AI Procurement Review: Model Licensing, Data Provenance, and Usage Terms | Vol 4 Ch 10

```text

AI systems need separate provenance records for code, models, data, and outputs. Library licenses and service terms answer different questions. Vardiya does not infer blanket permission from an open-weight label, instead examining the specific revision, relevant documents, and intended use to identify questions requiring further review.

Adem YÜCE • Production Engineering with AI • Volume 4: AI-Native Production • Chapter 10: Licensing, Copyright, Procurement, and Responsibility

Topics covered in this chapter:

- Provenance of Code, Models, Data, and Output

- Licenses and Service Terms

- Open Weights and Open Source

- Third-Party Data and Content Flows

- Changing Terms and Expert Review

- Vardiya's Procurement Decision File

Reading third-party content, indexing it, and transferring it to another service are different activities. Vardiya's procurement file relates each to applicable documents. Changed terms trigger review, but a text difference alone does not establish a legal conclusion or justify carrying prior acceptance into new conditions.

```

Production Engineering with AI Podcast · Season 4 · Vol 4 Ch 10

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/#chapter-10

Read the book: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Editorial draft adapted from the Turkish manuscript. No English book source or generated media is available.

Complete chapter/volume links and actual audio details before publication. The AI disclosure is draft publication copy to verify after production.

AI Data Lifecycle: Access Boundaries, Deletion, and Backup Restoration | Vol 4 Ch 9

```text

Data protection starts with knowing which records are processed and for what purpose. Vardiya inventories sources and copies while enforcing tenant and user permissions separately, including graph relationships. Restricting recorded telemetry fields helps prevent monitoring requirements from turning into unnecessary collection of personal task content.

Adem YÜCE • Production Engineering with AI • Volume 4: AI-Native Production • Chapter 9: Data Security, Privacy, and Lifecycle

Topics covered in this chapter:

- Data Inventory and Classification

- Tenant and User Boundaries in Graph Queries

- Masking and Log Minimization

- Indexes, Memory, Graph Relationships, Summaries, Caches, and Backups

- Retention, Deletion, and Rights Requests

- Vardiya's Deletion Exercise

Deleting a source requires examining its derived search entries, summaries, graph relationships, and caches. Vardiya follows these dependencies while separately reviewing retention conditions and request scope. Its exercise checks whether restored data violates a recorded deletion decision, with local results clearly limited to the environment actually tested.

```

Production Engineering with AI Podcast · Season 4 · Vol 4 Ch 9

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/#chapter-09

Read the book: https://kitap.ademyuce.tr/en/production-engineering-with-ai/ai-native-production/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

AI-assisted software development includes defining tasks and reviewing results as well as generating code. Line suggestions and multi-step agent tasks represent different working styles. Developers still need to determine intended behavior and assess whether a change delivers it; generated code alone is not a finished product.

Adem YÜCE • Production Engineering with AI • Volume 2: Vibe Coding and Loop Engineering • Chapter 1: The transformation of software development

Topics discussed in this episode:

- How did software development change between 2020 and 2026?

- From the Copilot era to the agent era

- From writing code to managing the system

- The misconception that “AI writes the code”

- The approach “I write code with AI”

- New skills for junior, mid-level, and senior developers

- The “head chef” role: A basic framework

- How should software education change? A brief discussion

- How to use this book

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/ai-production-engineering/vibe-coding-and-loop-engineering/#chapter-01

Read the book: https://kitap.ademyuce.tr/en/ai-production-engineering/vibe-coding-and-loop-engineering/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Vibe coding turns intent into an implementation attempt and uses the result to guide further work. Lower experimentation costs can make alternatives easier to explore, but the need being addressed must remain explicit. Small, bounded trials help compare possible solutions before committing too early to a design.

Adem YÜCE • Production Engineering with AI • Volume 2: Vibe Coding and Loop Engineering • Chapter 2: The philosophy of vibe coding

Topics discussed in this episode:

- What is vibe coding?

- Turning intent into code

- Prompt engineering and vibe coding

- Development with AI and development within AI

- The FAAFO framework

- Option value

- Flow state

- Making room for errors

- The iteration loop

- The risks of vibe coding

- Arguments against vibe coding and responses

- A map of vibe coding terminology

- Agentic coding and agentic engineering

- Spec-driven development and the PRD-first approach

- Mise en place

- Context fluency

- Steering rules, rules files, and project instructions

- The agent harness

- Human-in-the-loop, human-on-the-loop, and human-out-of-the-loop

- Verification-driven development

- Throwaway prototypes and production-bound code

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/ai-production-engineering/vibe-coding-and-loop-engineering/#chapter-02

Read the book: https://kitap.ademyuce.tr/en/ai-production-engineering/vibe-coding-and-loop-engineering/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/

Coding-agent context combines general project information with the code and boundaries of the current task. General programming knowledge does not mean the model knows a private repository's present state. Selecting relevant files and expected behavior creates a clearer development target than simply filling the context window.

Adem YÜCE • Production Engineering with AI • Volume 2: Vibe Coding and Loop Engineering • Chapter 3: Context engineering

Topics discussed in this episode:

- What is context?

- Tokens and the context window in development workflows

- Does the model know everything?

- Project context

- Task context

- File context

- Why does conversation history become contaminated?

- Context compaction

- Context clearing

- Context continuity

- What is a PRD?

- What is TASKS.md?

- What is CLAUDE.md?

- Rules files

- Codified context

- Hot memory and cold memory

- Building project memory

- Introduction to graph engineering: Files, symbols, and dependencies

- Checking the graph against the source code and keeping context focused

#reklam

Read the chapter: https://kitap.ademyuce.tr/en/ai-production-engineering/vibe-coding-and-loop-engineering/#chapter-03

Read the book: https://kitap.ademyuce.tr/en/ai-production-engineering/vibe-coding-and-loop-engineering/

Book series: https://kitap.ademyuce.tr/en/production-engineering-with-ai/