L2 Deep Dive
AI Engineering Agent Loop Anatomy for DB and Cloud Engineers
A practical mental model for how coding agents plan, call tools, observe results, and complete infrastructure work without treating the model response as the whole system.
Series / AI Engineering
AI Engineering Operating Model
How to run AI-assisted engineering as a governed operating model: agent loops, context, evals, permissions, control planes, and review boundaries.
60 posts AI Engineering
Who This Is For
Engineering leads and senior engineers integrating AI agents into production workflows — not using AI as a chat tool but running it as a governed system with evals, access controls, and audit trails.
What You Will Be Able to Do
- Design agent loops with proper permission boundaries and human review gates
- Build evaluation harnesses that catch regressions before they reach production
- Instrument agent observability without rewriting your existing monitoring stack
- Decide when to auto-approve vs require human confirmation for AI-generated actions
Prerequisites
Comfortable with CI/CD and production engineering concepts. No AI background required — this series treats agents as distributed systems, not magic.
1 Foundation
How agent loops work and what makes evaluation trustworthy — the two ideas everything else builds on.
L1 Field Note
AI Engineering Evals Are the New Unit Tests for Agents
Why database and cloud teams need agent eval harnesses that grade outcomes, not persuasive transcripts.
L1 Field Note
AI Engineering Outcome-Based Agent Evaluation vs Transcript Review
A field note on why agent evaluation should measure verified state changes instead of polished reasoning traces.
2 Operating Model
Permission boundaries, autonomy controls, and efficiency decisions that govern what agents can do without human approval.
L1 Field Note
AI Engineering Agent Autonomy Ladder: Manual, Confirm, Auto-Approve, Supervised
A governance model for deciding which database and cloud agent actions require approval and which can run automatically.
L1 Field Note
AI Engineering Harness Engineering: The 2026 Breakthrough Concept
Why the real engineering surface around agents is the harness of tools, scripts, context, review, and telemetry.
L1 Field Note
AI Engineering Token-Efficient Tool Use
How to design agent tool surfaces that preserve context budget for reasoning instead of wasting it on tool metadata and raw output.
L1 Field Note
AI Engineering Tool Search vs Loading Every MCP Tool
Why production agents need discoverable tools and context budgets instead of one giant always-loaded MCP surface.
3 Production Patterns
Identity, observability, safe deployment, and context throughput — the operational concerns that only appear at scale.
L2 Deep Dive
AI Engineering Agent Productivity Depends on Context Throughput
AI coding agents work better when voice, clipboard, screenshots, and MCP tools reduce context friction.
4 Historical Context — Earlier Agent Patterns
2024 writing on agent architectures, error amplification in multi-agent systems, and the shift from chat to goal-directed operation. Read these for perspective on how the field got here.
L3 Reference Guide
AI Engineering How Paperclip Is Redefining AI Agent Orchestration for the Zero-Human Company
Paperclip's zero-human orchestration model — goal-directed agent teams instead of task-by-task prompting — and what that architecture requires from the software and data systems beneath it.
L2 Deep Dive
AI Engineering From Chat to Agents: Designing Goal-to-Result Systems for Real Work
Chat is request-response; agents are task systems that plan, call tools, iterate, and stop when done. The minimum architecture — loop, tools, bounded memory, stopping conditions — required to make the transition from chat reliable.
L2 Deep Dive
AI Engineering Independent Parallel Agents Don't Cancel Errors — They Amplify Them
Google Research found that independent parallel agents amplify errors 17x compared to centralized orchestrator topologies. Adding more agents to a system with a shared context defect makes it worse, not more resilient.
L2 Deep Dive
AI Engineering Use Coding Agents as a Toolchain, Not a Vendor Bet
A production-minded workflow for running Cursor and Aider together without locking engineering practice to one agent.
Additional Posts
Related posts matched to this series by topic, tags, and keywords.
L1 Field Note
Databases The Database Observability Baseline: What Every DBA Dashboard Must Show
Before you can adopt AI-assisted triage, your database dashboard needs a foundation built on saturation, locking, and lag metrics.
L1 Field Note
AI Engineering The New AI FinOps Model: Seat Cost vs Token Cost vs Agent Runtime Cost
Why traditional SaaS spend models fail for agentic AI, and how platform teams are treating LLM compute like database provisioned IOPS.
L2 Deep Dive
Databases PostgreSQL Observability: Vacuum, Bloat, Locks, Replication Lag, and Query Plans
Monitoring PostgreSQL requires looking past the operating system and into the internal bookkeeping of MVCC, autovacuum, and replication streams.
L2 Deep Dive
Databases Cassandra Observability: Compaction, Tombstones, Repair, Latency, and Hot Partitions
Why generic server monitoring fails for Apache Cassandra, and how to track the true operational signals of a distributed masterless database.
L1 Field Note
AI Engineering Why Agentic AI Costs Explode: Context Size, Tool Calls, MCP Servers, Repo Size, and Retry Loops
Agentic AI systems can quietly accumulate massive API bills due to compounding context windows, retry loops, and unconstrained workspace parsing.
L1 Field Note
Engineering Fundamentals AI Cost Observability Dashboard: LangSmith vs Helicone
How to build an AI FinOps dashboard and choose between proxy-based and instrumentation-based observability.
L2 Deep Dive
Engineering Fundamentals Cost Observability: Build Dashboards That Show Waste Before Finance Finds It
How to expand monitoring beyond uptime by building dashboards that expose underutilized RDS instances, EBS io2 waste, and backup retention drift.
L2 Deep Dive
AI Engineering Prompt Caching, Context Pruning, and Model Routing: Practical Ways to Reduce LLM Cost
How to combine semantic routing, structured context pruning, and prompt caching to reduce production LLM API costs without degrading application quality.
L2 Deep Dive
AI Engineering FinOps Observability: Tie Cloud Cost to Workload, Team, Product, and Customer
How to connect engineering telemetry with cost telemetry to achieve granular cloud unit economics using FinOps principles and FOCUS standards.
L2 Deep Dive
AI Engineering Telemetry Cost Control: Why Observability Data Itself Needs Governance
If you log everything and monitor every dimension, your observability bill will eventually exceed your database infrastructure bill. Here is how to fix it.
L2 Deep Dive
AI Engineering AI Agent Observability: Monitor Tool Calls, Token Spend, Latency, and Failure Loops
Why monitoring autonomous SRE agents requires tracking tool-call hallucinations, context window saturation, and recursive retry loops, rather than just basic CPU metrics.
L2 Deep Dive
AI Engineering MCP Server Observability: The New Control Plane for AI + Enterprise Tools
How the Model Context Protocol (MCP) became the networking layer for AI agents, and why monitoring these connections is critical for enterprise security.
L2 Deep Dive
AI Engineering Agentic SRE Architecture: Skills, Agents, MCP Servers, and Human Approval Loops
The definitive 2026 reference architecture for autonomous database operations, from detection to multi-agent diagnosis to human-in-the-loop remediation.
L2 Deep Dive
AI Engineering The Harness Around the Agent: How Stripe Runs 1,000 Unattended Code Reviews per Week
Stripe's Minions system runs over a thousand AI code reviews weekly using a fork of an open-source agent. The reliability comes from the deterministic pipeline around it, not the model inside.
L2 Deep Dive
AI Engineering AI Agents Need a Control Plane, Not More Interfaces
Production AI agents work best when coding, files, tools, and knowledge workflows share one governed execution model.
L2 Deep Dive
AI Engineering Runtime Boundaries for Agentic App Builders
A hosted AI app generator fails when the mobile chat becomes the platform — API keys end up in binaries, execution state blurs with chat, and previews break without artifact handoff. The control-plane architecture that keeps these concerns separated.
L1 Field Note
AI Engineering The Agent Should Not Have Your App Credentials
Giving an AI coding agent your application's Postgres credentials is the default mistake — the agent inherits every permission the app has. Database-enforced read-only roles, replica routing, query limits, and project-scoped MCP config are the alternative that actually fails closed.
L3 Reference Guide
AI Engineering AI Agents Need Database Guardrails Below the Prompt
Prompt-level guardrails fail open when the agent misinterprets context. The only boundary that mechanically rejects destructive SQL is the database — dedicated read-only roles, sanitized view schemas, and a network path that application credentials never touch.
L2 Deep Dive
AI Engineering Remote Agents Need Deployment, Permissions, and Feedback Loops
Codex mobile turns local agents into remote workflows, but production value depends on deployment, access control, and observability.
L2 Deep Dive
AI Engineering Evaluate AI Agents by Completed Work, Not Token Price
Production AI agent selection should measure quality, retries, tokens, latency, and verification cost per completed task.
L2 Deep Dive
Databases Natural Language SQL Agents Need Guardrails Before Orchestration
How Postgres chat agents turn intent into SQL, and why production systems need schema controls, validation, and auditability.
L2 Deep Dive
AI Engineering The Three-Layer Agent Infrastructure Stack for Database Operations (April 2025)
Building a database operations agent requires a workflow framework, production observability, and scalable inference — April 2025 shipped open-source solutions for all three layers simultaneously.
L2 Deep Dive
Databases Three Open-Source Tools Filling the Gaps in Database Operations (May 2025)
May 2025's most-starred new projects solve three specific database team problems: backup restores that are never verified, internal knowledge that can't be retrieved, and AI agents blind to your schema history.
L2 Deep Dive
AI Engineering Top GitHub Breakouts: May 2025 — Agent Infrastructure Without Boilerplate
Three May 2025 open-source projects eliminate the manual scaffolding that blocks every AI agent deployment: orchestration glue, vector database setup, and MCP gateway configuration.
L2 Deep Dive
AI Engineering Parallel AI Agents Need an Operating Model
Running many coding agents only works when git isolation, shared memory, permissions, hooks, and verification are designed as a system.
L2 Deep Dive
AI Engineering Personal AI Agents Fail in the Last 20 Percent of Integration
Self-hosted AI agents become useful only when model quality, tool access, memory, and setup completeness line up.
L3 Reference Guide
Databases Natural Language SQL Agents Need Database Guardrails
The risk in a natural-language SQL agent is not bad SQL — it is authority compilation: a user sentence becomes a database operation unless the control plane proves, before execution, which role, rows, cost, and columns the query is allowed to touch.
L2 Deep Dive
AI Engineering AI Agents in Platform Automation: Useful Assistant or Unreviewed Change Engine
When AI agents accelerate platform operations versus when they generate unreviewed changes — the permission boundary and audit design that separates useful from risky.
L2 Deep Dive
AI Engineering The AI-Native Engineering Stack: Agents, Inference, and Knowledge Graphs in Production (November 2025)
Three November 2025 breakout projects eliminate the manual infrastructure build that blocks teams from running AI agents in production — covering agent backends, Kubernetes LLM inference, and SQL-driven knowledge retrieval.
L2 Deep Dive
Databases Automated Reliability Across the Stack: Database Backups, Platform Observability, and SQL Quality (November 2025)
Three November 2025 open-source releases eliminate manual work from three engineering reliability tasks — multi-database backup verification, self-hosted log and trace collection, and SQL static analysis in CI pipelines.
L1 Field Note
Databases Agentic Code Review for Database Repositories
Database repositories contain hidden rules human reviewers know: never add a blocking index at peak hours, never widen IAM without owner approval. Agent review surfaces these violations before merge — without displacing the human judgment that set the rules.
L1 Field Note
Databases Database Runbooks as Agent Contracts
A reference operating model for turning human database runbooks into machine-usable agent contracts.
L1 Field Note
AI Engineering Agent-to-Agent Review Loops
A practical review pattern where one agent creates a change and specialized agents review risk, rollback, security, and observability.
L1 Field Note
AI Engineering Application Legibility for Agents
A reference architecture for making logs, metrics, test output, schemas, and deployment history readable by coding agents.
L1 Field Note
AI Engineering Programmatic Tool Calling for DB Automation
A reference pattern for keeping large database outputs out of model context by using scripts that summarize evidence before the agent sees it.
L1 Field Note
AI Engineering Context Anxiety and Harness Decay
Why agent harnesses become stale when they overfit today's model weaknesses instead of stable execution contracts.
L2 Deep Dive
AI Engineering Top GitHub Breakouts: February 2026 — Local Agents and MCP Bridges
February 2026's highest-starred new open-source projects connecting AI agents to local infrastructure, Kubernetes clusters, and structured data without cloud API dependencies.
L2 Deep Dive
Databases Top GitHub Breakouts: March 2026 — Agent Adaptation and Production-Scale Vector Search
The second wave of March 2026 breakouts: an agent that learns from every conversation, a Rust vector index that outperforms FAISS at a fraction of the memory, and a Kubernetes-native agent control plane.
L2 Deep Dive
AI Engineering Top GitHub Breakouts: April 2026 — Production Agent Infrastructure
The highest-starred new open-source projects in April 2026 targeting production-scale AI agent memory, protocol enforcement, and Postgres environment management — what breaks when agents leave single-developer scope.
L2 Deep Dive
Cloud & Platform Service Catalogs Are Not Portals. They Are Control Planes
A service catalog that helps engineers find links is a directory. One that owns metadata, policy, workflow, and reconciliation is a platform control plane — and only the second one solves the real scaling problem.
L2 Deep Dive
Cloud & Platform CI/CD Pipelines Are Distributed Systems With Bad Observability
CI/CD pipelines fail as distributed coordination systems long before they fail as broken scripts — why build badges hide partial failures, flaky retries, and ordering gaps that only appear under real delivery load.
L2 Deep Dive
Cloud & Platform Automation Incident Review: When the Tool Worked and the System Failed
The hardest automation incidents are not broken tools — they happen when every tool executes exactly as asked while the surrounding system loses the ability to evaluate whether that action is still safe.
L2 Deep Dive
Cloud & Platform Backstage, Port, Cortex, and AWS Service Catalog: Different Tools, Different Control Planes
Backstage, Port, Cortex, and AWS Service Catalog compared on control-plane model — which tools provision, which only display, and where each abstraction breaks down.
L2 Deep Dive
Cloud & Platform OpenTofu vs Terraform: What Platform Teams Should Actually Evaluate
OpenTofu vs. Terraform on licensing risk, provider supply chain compatibility, state safety, and the migration cost platform teams actually absorb.
L2 Deep Dive
Cloud & Platform Internal Developer Platform Reference Architecture: Catalog, IaC, CI/CD, Policy, and Observability
Reference architecture for an IDP as a control plane—connecting service catalog, IaC, CI/CD pipelines, policy enforcement, and observability feedback.
L2 Deep Dive
Cloud & Platform CI/CD Observability: Queue Time, Flake Rate, Lead Time, Failure Domains, and Change Risk
Queue time, flake rate, lead time, failure domains, and change risk as CI/CD signals that reveal whether a delivery system is becoming safer or just busier.
L2 Deep Dive
Cloud & Platform The Deployment Control Plane: CI/CD, Catalog, Policy, Observability, and Human Approval
CI/CD, service catalog ownership, policy gates, and SLO observability wired into a control plane that authorizes each deployment before it ships.
L2 Deep Dive
Cloud & Platform The Platform Automation Maturity Model: Scripts, Modules, Catalogs, Pipelines, Control Planes
How platform automation matures from one-off scripts to a governed control plane — and where most teams get stuck between modules and catalogs.