I built this in October, after a Sunday I lost to planning. Not to studying — to planning. I had the syllabus, a spreadsheet, three colours of highlighter, and no idea what to do on Monday at 19:30. The spreadsheet said “Fixed Income”. That is not a task. That is a topic.
So I wrote a script that prints the day. It started as one command, cfa today. Then I missed a week. The plan kept pretending I had not, which was the real bug: a plan that does not read your log is fiction. The log became the source of truth, and the weekly plan became something that reacts to it instead of performing at it.
That is the whole idea. The rest is plumbing.
What it does
cfa today prints the day’s tasks with minutes, the target, what you have logged, the next deadline and the next mock. cfa log records a session against a learning module. cfa due is the spaced-review queue. cfa errors is the error log, with a why-line. cfa analytics turns the log into a readiness index and a next action. cfa pace tells you whether the calendar is still honest.
$ cfa today
Wed 04 Nov 2026 | 0 - Setup | 558 days to exam
Target 195 min · logged 0 min · English 0/60 min · 0%
Next deadline: Registration opens (May 2028) + Access Scholarship window opens - 284 days (2027-08-15)
Focus: Toolkit: error log, formula sheet, flashcard deck, calculator drills, baseline diagnostic
[ ] English: 30 min reading + 30 min explaining yesterday's topic aloud (60 min)
[ ] Build the formula sheet skeleton for all 10 topics (80 min)
[ ] Practice the same material: 15 questions, immediate correction (55 min)
The hour model is per module, not per topic. plan/lessons.json gives each of the 102 learning modules a learn / practice / review budget. The total is 471 hours. Add mocks, timed sets and remediation and you get roughly 730 hours across 79 weeks, about 9.3 CFA hours a week. The 30-hour ceiling in profile.json is insurance, not a target. A plan that assumes 30 hours a week for 18 months dies in month four.
The week
On Sunday night, cfa week --next reads the master plan, the current dated block, the exact modules due that week, the review queue, the open errors, last week’s plan and the recent log. It writes a Monday-to-Sunday plan under hard rules: stay inside the budget, use only this week’s module ids, put reviews before new material, keep the English block where it is, never invent a resource. The Sunday 19:00 timer runs the review and then the plan.
The planner can call a model. I gave it a key, and I use it, but I do not depend on it. If the call fails, the script builds the week from curriculum.json: a seven-day arc of read, practise, read, practise, review, long block, review, with the budget allocated by day weights. The fallback is not a degraded mode. On a bad night it is the mode I trust most.
The log is the system
Everything derives from progress.csv: coverage, hours by topic, streak, the spaced-review queue. The intervals are 1, 3, 7, 21 and 45 days. Logging a review against the same module id is what advances the interval. The queue cannot be gamed by opening a flashcard app and closing it.
The error log is deliberately blunt:
cfa error "Duration sign flipped on a putable bond" --lesson LM58 --why "confused modified duration with Macaulay"
cfa close 3
The why-line is the useful part. “I got it wrong” is not a cause. “I used modified duration where the question asked for the price change per 100 bp” is a cause you can fix.
Excel, because sometimes I want to see it
CFA_Level_I_Tracker.xlsx is generated from the same JSON files: Dashboard, Sessions, Modules, Errors, Mocks, Plan, Setup. Yellow cells are inputs, the rest are formulas. It is the same data in Bloomberg clothes, and it is useful when I want to look at the whole programme without reading a terminal. python make_workbook.py rebuilds it; finalize_workbook.ps1 recalculates it with Excel on Windows.
What it does not do
- It is not a question bank. It tells you what to do and records what you did. The learning happens in the questions and in the correction.
- It is not official CFA material. The 102 module names and the 471-hour budget are my own planning notes. No curriculum text, question bank or mock is included. CFA Institute is not affiliated with this project and does not endorse it.
- The exam date is an estimate. 15 May 2028 is a placeholder for the May 2028 window. Change it in
profile.jsonwhen the calendar is published. - The model can be wrong. The prompt constrains the budget and the module ids, but I still read the week before I follow it.
- No mobile app, no cloud, no sync. CSV files in a folder. That is the point, and it is also a limit.
- Timers are platform-specific. systemd on Linux, Task Scheduler on Windows. On macOS the installer stops after the
cfacommand; you run it by hand or wire your own scheduler. - Not investment advice.
Install it
Linux, macOS, WSL or Git Bash:
git clone https://github.com/yirvisom/cfa-l1-autopilot.git
cd cfa-l1-autopilot
bash install.sh
cfa doctor
cfa today
Windows PowerShell:
git clone https://github.com/yirvisom/cfa-l1-autopilot.git
cd cfa-l1-autopilot
powershell -ExecutionPolicy Bypass -File .\install.ps1
cfa today
Add -Timers to the Windows installer if you want the 07:30, 21:30, 22:30 and Sunday 19:00 tasks. The key is optional. If you want model-written weeks, put one in ~/.config/cfa/env (or %USERPROFILE%\.config\cfa\env). Without it, cfa week --offline is the honest default.
Why I published it
I published it because the useful part is not the code. It is the rule: if the plan does not read the log, it is decoration. If you are sitting CFA Level I, steal the rule and the hour table. Change the dates. Delete the English block if it is not yours. Keep the why-line.