99 — Pre-Mortem: How This Plan Fails

Warning: this directory is honest. Do not read on a bad day. Return to it every quarter for a check.

What a Pre-Mortem Is

Gary Klein invented the pre-mortem. The setup is simple: imagine the project has failed. It is August 2027. Thirteen months are gone. The plan did not work. You are back where you started, or worse — burned out, disillusioned, still not writing Java by hand.

Now answer one question: why did it fail?

That is the pre-mortem. You do it before you start, not after. You surface the failure modes while you still have time to defend against them. A regular post-mortem asks “what happened?” A pre-mortem asks “what will happen unless we act?”

This directory is that exercise, applied to your specific plan, your specific life, and the specific ways engineers like you fail at long-horizon technical rebuilds.

Why It Matters More Than the Optimistic Files

The rest of this roadmap (Phase 01 through Phase 12 plus the discipline directory) tells you what to do when things go right. This directory tells you what to do when things go wrong — and things will go wrong. Every 13-month plan has a bad month. Every rebuild has a week where you seriously consider quitting. Every engineer with a full-time job and family has a stretch where the plan collapses to zero for a while.

The question is not whether those weeks happen. It is what happens after them. Engineers who plan for failure recover in days. Engineers who assume success collapse for months.

The Eight Failure Modes (Ranked by Probability)

You have one plan and roughly eight ways it can fail. They are not equal in likelihood. The table below is calibrated to your specific profile: applied ML engineer at Zoho, AI-agent-dependent, 10-15 hrs/week, based in Chennai, learning Java again after ~2 years away from hand-coding.

#	Failure Mode	Probability (no countermeasures)	When it strikes	File
F1	AI-crutch relapse	70%	M1-M3	01_fm_ai_crutch_relapse.md
F2	Work-load erosion	60%	Any month, especially quarter-end	02_fm_work_load_erosion.md
F3	Tutorial hell	60%	M1-M4	05_fm_tutorial_hell.md
F4	Isolation & no feedback	55%	M3-M9	06_fm_isolation_and_no_feedback.md
F5	DSA black hole	50%	M2-M6	03_fm_dsa_black_hole.md
F6	Scope creep & shiny objects	45%	M3-M10	07_fm_scope_creep_and_shiny_objects.md
F7	Framework worship	40%	M4-M8	04_fm_framework_worship.md
F8	Life & health	100% (guaranteed to visit; severity varies)	Any month	08_fm_life_and_health.md

F1 is the highest-probability failure mode because it is the reason you are running this plan at all — you have been outsourcing your Java thinking to AI agents for two years, and habits of that depth do not break easily. F8 is not the highest-probability but it is the only one marked 100%: at some point in 13 months, life will interrupt the plan. Wedding season, a parent’s health, a Zoho crunch that eats a month, a personal grief. Plan for it.

How to Use This Directory

• Read the whole thing once, at the start of Sprint 1 (July 2026). This calibrates your expectations.

• Skim the table above at the start of every sprint. Ask: “which of these is closest to me right now?”

• Read the specific file when you feel the failure mode approaching. Do not wait until you are inside it.

• Re-read 08_fm_life_and_health.md any time you feel guilty for pausing. That file exists to give you explicit permission to be a person.

• Return quarterly (every 3 sprints) for a full re-read. Failure modes evolve as you progress; the ones that felt distant in M1 will feel close in M6.

Contents

File	Purpose
01_fm_ai_crutch_relapse.md	You keep using Cursor/Claude to write your Java for you.
02_fm_work_load_erosion.md	Zoho crunch eats your evenings and weekends.
03_fm_dsa_black_hole.md	LeetCode swallows M2-M6; you never get to real projects.
04_fm_framework_worship.md	You jump to Spring Boot before Java fluency returns.
05_fm_tutorial_hell.md	Endless Baeldung tabs, videos on 1.5x, no shipped code.
06_fm_isolation_and_no_feedback.md	The plan is a private cathedral. No one sees it.
07_fm_scope_creep_and_shiny_objects.md	Every new JEP and library pulls you off course.
08_fm_life_and_health.md	Life interrupts the plan. This is the emotional file.
09_summary_and_reset_protocol.md	You fell off. Here is how to come back.

The Governing Principle

Every failure mode in this directory can be recovered from if you catch it early. None of them can be recovered from if you catch them six months late. Early warning matters more than perfect execution. Read this directory as a set of tripwires, not a set of accusations.

The plan is a tool. You are a person. This directory is what keeps the plan honest to that fact.

---

Next: 01_fm_ai_crutch_relapse.md — the highest-probability failure mode.