7 Learning Modes: One Lab, Seven Ways to Master It

Here is something I have learned from ten years of training DBAs: watching someone complete a task correctly once does not mean they can do it again under different conditions. The engineer who sets up streaming replication flawlessly in Guided mode might freeze when the same cluster is handed to them already broken. The one who debugs replication issues all day might struggle to set it up from scratch against a timer.

Mastery is not one skill. It is a collection of related skills - knowledge acquisition, problem solving, diagnosis, speed, exploration, collaboration, and deep understanding. Each one needs to be trained separately.

That is why every GritLabs lab runs in seven modes. Same infrastructure, same topic, seven completely different learning experiences. Here is how each mode works, why it matters, and when to use it.

Guided - "Learn"

What it is: Step-by-step instruction with full AI tutor support. Sage walks you through each objective, explains concepts as you go, and proactively offers help when you get stuck.

How Sage behaves: Proactive. Sage watches your terminal output and jumps in with context before you ask. If you run psql and connect successfully, Sage might say: "Good - you're connected to the primary. Before we configure replication, let's check the current WAL level. What command would show you the current setting?" Sage teaches through questions even in guided mode, but the questions are scaffolded to lead you in the right direction.

Scoring: Standard - complete objectives, earn points. No penalties for asking for help. No time pressure beyond the session duration.

When to use it: First time encountering a topic. You have never set up streaming replication before. You need someone (or something) to walk you through the process, explain why each step matters, and catch your mistakes before they compound. This is learning mode, not testing mode.

Who it is for: Beginners to the specific topic. Even experienced DBAs should start in Guided mode when they encounter a lab on a technology they have not used before.

Challenge - "Prove"

What it is: You see the objectives but get zero instructions. Sage only responds when you directly ask for help, and even then, it answers with questions rather than commands. Hints are available but each one reduces your score.

How Sage behaves: Reactive. Sage sits quietly unless you address it. When you do ask, it responds with diagnostic questions: "What have you tried so far?" or "What does the error message tell you about the connection?" It never gives direct commands. The point is to make you think through the problem, not to hand you the answer.

Scoring: Penalty-based. You start with full marks. Each hint request costs points. Each failed objective check costs a small amount. The fastest completions with fewest hints score highest.

When to use it: After you have completed the lab in Guided mode at least once. You understand the concepts but want to test whether you can execute independently. Challenge mode exposes the gaps between recognition ("I remember seeing this step") and recall ("I can reproduce this step from memory").

Who it is for: Intermediate learners who want to solidify their knowledge. Also excellent for experienced engineers who want to verify they can still execute a procedure they have not done recently.

Chaos - "Survive"

What it is: Your infrastructure starts broken. Something is wrong - a misconfigured pg_hba.conf, a replication slot that is not created, a standby that refuses to connect, a WAL directory that is full. You do not know what the problem is. Your job is to diagnose it, fix it, and get the system working.

How Sage behaves: Socratic. Sage role-plays as a junior on-call engineer who does not know the answer either. It asks diagnostic questions: "What does that log line tell you?" and "What would you check next?" and "Why do you think that process is not running?" It never tells you the root cause. It guides your diagnostic process through questions.

Scoring: Standard scoring based on objectives completed, but the objectives are hidden. You discover what needs to be fixed by investigating the system. As you fix each issue, the corresponding objective marks itself complete.

When to use it: When you want to build real troubleshooting instincts. Chaos mode is the closest simulation to a production incident. You are dropped into a broken system with no context, and you need to figure out what is wrong. This is where pattern recognition develops - the kind of intuition that makes the difference between a 12-minute fix and a 2-hour outage.

Who it is for: Anyone preparing for on-call rotations. Anyone who has completed the lab in Guided and Challenge modes and wants to test their diagnostic skills. This mode is hard - expect to struggle, and expect to learn more from the struggle than from any video.

Speedrun - "Compete"

What it is: Same objectives as Challenge mode, but with a prominent countdown timer. No tutor, no hints. Complete all objectives as fast as possible. Your time is recorded on the leaderboard.

How Sage behaves: Minimal. Sage is effectively disabled. It only responds to direct questions with one or two sentence answers. No explanations, no context, no teaching. You are on your own.

Scoring: Time-weighted. Completing all objectives is the baseline. Your score is determined by how fast you finish. Leaderboards track the fastest completions globally and within your cohort.

When to use it: When you have mastered a lab and want to build fluency. The first time you set up streaming replication, it might take 45 minutes. After three Speedrun attempts, you can do it in 12. When production is down and every minute costs money, fluency matters. Speedrun mode develops the muscle memory that turns a procedure into a reflex.

Who it is for: Advanced learners who want to compete with themselves or others. Also excellent for teams preparing for incident response drills where speed of execution is critical.

Sandbox - "Explore"

What it is: No objectives, no grading, no timer. You get the infrastructure and complete freedom to experiment. Want to test what happens when you set wal_level = minimal on a primary with an active replica? Try it. Want to see how Patroni handles a split-brain scenario? Create one.

How Sage behaves: Full access. Sage is your conversational partner. Ask anything - explain concepts, suggest experiments, discuss trade-offs, help debug. There are no constraints on the conversation. Sage is curious and encourages exploration.

Scoring: None. Sandbox mode is not graded. There is no timer and no objectives. The only measure of success is what you learn.

When to use it: When you have a specific question you want to answer through experimentation. When you want to test a theory before applying it in production. When you just want to play with the infrastructure and see what happens. Some of the deepest learning happens in Sandbox mode because you are following your own curiosity instead of someone else's curriculum.

Who it is for: Curious engineers at any level. Senior DBAs testing edge cases. Students who finished the structured labs and want to go deeper. Anyone who learns best by breaking things.

Pair - "Collaborate"

What it is: Two students share the same lab infrastructure. Each person gets their own terminal, but the servers are shared. Objectives can be assigned to specific students or shared. You need to coordinate - one person configures the primary while the other prepares the replica.

How Sage behaves: Collaborative. Sage can see both students' terminals and acts as a mediator. It suggests task division, flags when one student's actions might conflict with the other's, and ensures both students are learning rather than one doing all the work.

Scoring: Shared. Both students receive the same score. This incentivizes collaboration over competition. If one student carries the other, Sage notices and prompts the less-active student to take on tasks.

When to use it: Team exercises, study groups, or pair programming sessions. Pair mode mirrors real-world operations where teams coordinate across multiple servers during deployments or incidents. It also develops the communication skills that are critical during incident response - explaining what you see, coordinating actions, and avoiding stepping on each other.

Who it is for: Any two students who want to learn together. Particularly valuable for teams that will be working together on-call. The collaboration patterns you build in Pair mode transfer directly to incident response.

AI Lab - "Understand"

What it is: A teaching-first mode where Sage leads a structured learning session with distinct phases: concept, practice, checkpoint, and reflect. Before you touch the terminal, Sage teaches the core concept through dialogue. It asks questions, builds mental models, uses analogies. Only after you demonstrate conceptual understanding does Sage unlock the terminal for hands-on practice.

How Sage behaves: Teaching mode. Sage drives the session. In the concept phase, it asks questions like: "If the primary writes data and the replica has not received it yet, what happens if the primary crashes?" It builds your mental model before you start executing. During practice, it guides your hands-on work. At checkpoints, it quizzes you to verify understanding. In the reflect phase, it asks you to explain what you learned in your own words.

Scoring: Mastery-based. You are scored on both execution (did you complete the tasks?) and understanding (did you answer the checkpoint questions correctly?). High scores require both.

When to use it: When you want to deeply understand a topic, not just complete the tasks. AI Lab mode is slower than other modes on purpose. The goal is comprehension, not completion. If you rush through a Guided lab and complete the objectives without understanding why each step matters, AI Lab mode forces the understanding.

Who it is for: Students who want depth over breadth. Beginners who benefit from concept-first teaching. Anyone who has completed a lab before but realized they were following steps mechanically without understanding the underlying system.

Which mode should you start with?

If you are completely new to the topic: start with Guided, then move to Challenge, then Chaos.

If you are experienced but rusty: start with Challenge. If you struggle, drop back to Guided. If it is too easy, jump to Speedrun.

If you want to build diagnostic skills: go straight to Chaos mode. Expect to fail. That is the point.

If you learn best through conversation: AI Lab mode will suit you. Sage teaches before you practice.

If you want to compete: Speedrun. Check the leaderboard. Beat your own time.

If you want to explore: Sandbox. No rules, no objectives, just infrastructure and curiosity.

The modes are not a linear progression. They are complementary skills. The strongest DBAs I know can execute in all seven modes - they can follow a procedure, improvise when things break, race the clock, explore edge cases, collaborate under pressure, and teach others what they know.

One lab. Seven modes. Seven different kinds of mastery.