alex-effort-estimation

name	alex-effort-estimation
description	Estimate task duration from an AI-assisted development perspective rather than traditional human developer estimates
tier	extended
applyTo	**/*estimate*,**/*effort*,**/*planning*,**/*roadmap*,**/*tracker*
currency	"2026-04-22T00:00:00.000Z"

Alex Effort Estimation

Convert human developer estimates to AI-assisted timelines.

Last Updated: April 2026

---

Quick Reference: Effort Units

Unit	Time	Use For
⚡ Instant	< 5 min	Single edit, lookup, generation
🔄 Short	5-30 min	Multi-file change, docs, skill
⏱️ Medium	30-60 min	Feature, test suite, debugging
📦 Session	1-2 hours	Major feature, release
🗓️ Multi-session	2+ hours	Large refactor, new system

---

Conversion Table

Task Type	Human → Alex	Multiplier
Research	8h → 45m	×0.10
Code generation	4h → 35m	×0.15
Documentation	4h → 25m	×0.10
Skill creation	2h → 15m	×0.12
Refactoring	4h → 1h	×0.25
Bug fix (known cause)	2h → 35m	×0.30
Bug fix (unknown)	4h → 2h	×0.50
Architecture design	8h → 2h	×0.25
Test writing	4h → 1.5h	×0.40
New feature (S)	4h → 1h	×0.25
New feature (M)	2d → 3h	×0.20
New feature (L)	1w → 6h	×0.15
Release process	4h → 1h	×0.25

Formula: Alex Time = Human Estimate × Multiplier + Bottlenecks

---

What Accelerates (4-10×)

Activity	Why Fast
Research	Semantic search vs browsing
Boilerplate	Generated, not typed
Multi-file edits	Parallel in one pass
Code review	Instant pattern recognition
Learning new tech	Bootstrap learning skill
Documentation	Generated from code/context

---

Bottlenecks (Cannot Accelerate)

These take real time regardless of AI:

Bottleneck	Impact	Mitigation
Test execution	×1.0	Parallelize, fast feedback loops
Build/compile	×1.0	Incremental builds, caching
Human approval	Variable	Batch decisions, async review
External APIs	Variable	Mock for dev, timeout handling
CI/CD pipeline	×1.0	Optimize pipeline, parallel jobs
Context loading	×0.8	Pre-read relevant files

---

Estimation Template

| Task | Human Est. | Alex Est. | Bottleneck |
|------|------------|-----------|------------|
| [Task] | [Xh/d] | [⚡🔄⏱️📦🗓️] | [None/Test/Approval/Build] |

Example:

Task	Human	Alex	Bottleneck
Add new API endpoint	4h	🔄 30m	None
Write integration tests	3h	⏱️ 1.5h	Test execution
Deploy to staging	2h	📦 1.5h	CI pipeline
Code review cycle	2h	⏱️ 2h	Human approval
Total	11h	📦 5.5h	-

---

Planning Guidance

When Estimating

1. Apply multiplier to core work
2. Add bottleneck time at ×1.0
3. Factor approval cycles (human response latency)
4. Buffer 20% for unexpected iteration

---

Calibrated from 62-Project Retrospective

What Accelerates Well (4-10×)

Task Type	Human	Alex	Factor	Evidence
Documentation	4h	25m	10×	METHODOLOGY doc: 400 lines in ~30 min
Skill creation	2h	15m	8×	65 skills created, many in single sessions
Code generation	4h	30m	8×	Slash commands, refactors
Research + synthesis	8h	45m	10×	62 project analysis in ~20 min
Architecture decisions	8h	2h	4×	Root cause analysis + recommendations

What Doesn't Accelerate (<2×)

Bottleneck	Why	Evidence
External dependencies	Can't control	CookProject blocked by book formatting
Unrealistic scope	Must be discovered	Altman-Z-Score, KalabashDashboard

Example: v4.2.5 Release Retrospective

Task	Human Est.	Actual Alex	Bottleneck
Update engine to 1.109	30m	⚡ 5 min	None
Consolidate 9→3 agents	4h	🔄 20 min	None
Create 6 slash commands	2h	🔄 15 min	None
Refactor dream to shared	4h	⏱️ 45 min	Testing
Test all features	2h	⏱️ 1h	Human testing
Release process	4h	📦 1h	CI/approval
Total	16.5h	📦 2.5h	-

Acceleration factor: 6.6×

Success Predictors

Good Sign	Bad Sign
Clear "done" definition	Vague scope
Quick win potential	External dependencies
Daily incremental work	Big-bang delivery
Defined acceptance criteria	"We'll know it when we see it"

Scope Guidance

Alex Estimate	Scope Health
⚡ Instant	Perfect — ship it
🔄 Short	Good — single session
⏱️ Medium	OK — plan breaks
📦 Session	Consider splitting
🗓️ Multi-session	Split required

---

Anti-Patterns

❌ Instant everything — Testing and approval take real time
❌ Skip review — Speed without quality is waste
❌ Ignore iteration — First attempt isn't always right
❌ Forget context — Reading workspace takes time
❌ No buffer — Edge cases and clarifications happen