99 — Pre-Mortem

The plan you write while calm is worth more than the plan you write while panicking.

What This Directory Is

A pre-mortem is the opposite of a post-mortem. Instead of asking “why did this fail?” after the fact, you ask “where will this fail?” now — while you still have the leverage to change something.

This directory contains 8 named failure modes for a 13-month, part-time, moonlighting, returning-to-C++ roadmap targeting applied-ML engineering roles. Each one has:

  • Probability estimate (subjective, but calibrated against real 2026 India-tech-market data and typical returning-dev patterns)

  • Early warning signals (things you should notice within 1–2 weeks of drift, not 3 months later)

  • Concrete mitigation (what to do the moment a signal fires)

  • Escalation trigger (the specific threshold that forces a scope-cut or reset)

Plus a reset protocol (09_summary_and_reset_protocol.md) that tells you what to do when a failure mode does fire — because it will, at least once, probably twice, and if you’re honest about the base rates, maybe three times over 13 months.

Why This Matters

Failure at M6 with no plan feels catastrophic. Failure at M6 with a written protocol feels like turning a page.

The difference is not that the second person is stronger. The difference is that the second person already decided, in a calm moment, what “failure” looks like and what the response is. When it fires, they execute. They don’t spiral. They don’t quit. They cut scope, ship something tiny within 7 days, and return to cadence.

That is the entire game. Not avoiding failure — rehearsing the response to failure so it stops being a crisis.

The 8 Failure Modes

#

Mode

Probability

File

01

M1 wall — no rung 1 shipped by W4

25%

01_the_M1_wall.md

02

AI rebound — Cursor creeps back in

40%

02_the_AI_rebound.md

03

DSA rabbit hole — LeetCode eats months

15%

03_the_DSA_rabbit_hole.md

04

Zoho creep — day job devours evenings

60%

04_the_zoho_creep.md

05

M9 hard gate miss — Eigen MLP not shipped

30%

05_the_hard_gate_M9_miss.md

06

Capstone scope creep — MiniServe never ships

45%

06_the_capstone_scope_creep.md

07

study fail spiral — first 3 rejections crush morale

40%

07_the_interview_fail_spiral.md

08

Life risk — family/health/relationship crisis

~ (not %)

08_the_life_risk.md

Notice: Zoho creep is 60% — the highest single-source risk. AI rebound is 40% and is the risk most specific to Raghul’s background (two years of AI-assisted work). These two together account for the majority of the delta between “shipped M13” and “still trying at M18”.

Life risk (08) has no probability number. It is not a “failure mode” in the same sense — it is the ground on which the whole roadmap sits. It has its own register.

How To Use This Directory

  1. Read all 10 files once, on Day 1, before you start Phase 0. This is not optional. Understanding the failure modes in advance is what makes the mitigations work.

  2. Re-read the relevant file the moment a signal fires. Don’t wait for a scheduled review.

  3. Run the Sunday audit weekly (see ../13_discipline/07_motivation_sustainment.md). It is the primary early-detection mechanism.

  4. When a failure mode fires, run the reset protocol in 09_summary_and_reset_protocol.md — do not improvise.

The Reset Protocol, in One Paragraph

If a sprint slips: yellow flag, cut next sprint’s stretch target. If two sprints slip in a row: red flag — Sunday audit, identify the failure mode, read the file, cut scope by 50%, ship something tiny within 7 days, resume. If three sprints slip: pause for 1–2 weeks, run the full reset (file 09), then restart the last completed sprint with reduced scope. Escalate to full roadmap re-plan only if the pause exceeds 4 weeks. The roadmap serves you. You do not serve the roadmap.

Nav: ← 13_discipline · 00_command_center · Roadmap root