For the California Bar · February 2027
Bar Buddy is an AI coach that measures the one thing your question bank can't: whether you can write the black-letter rule from a cold start, under time, in your own words. Then it reorders tomorrow around what leaked.
50 founding seats · No card, no spam · Built for the February 2027 exam
State the rule for battery.
Type it from memory. No hints, no multiple choice — that's the whole point.
The problem
Multiple choice measures recognition — can you pick the right rule out of four. Essays and performance tests measure production — can you write it with nothing on the screen. Those two numbers come apart, and almost nothing in bar prep shows you the distance between them.
Pick a subject
Recognition — MCQ accuracy Production — rule typed cold
Evidence
Illustrative figures, shown to explain the measurement. Your board is built from your own attempts — Bar Buddy never shows you a number it didn't watch you earn.
What it does
Every tool feeds the same ranking. Nothing is a standalone drill you do and forget — each attempt moves a rule up or down tomorrow's list. And one of them isn't about the law at all: it's about protecting the hours you study in.
Type the rule cold. A keyword engine weights every element by how rare and load-bearing it is across the bank, tiers your answer 1–4, and names the element you dropped — not just a score.
Free tier · no AI call
Paste the answer you actually wrote. It's read against a CA rubric — issue spotting, rule statement, application, conclusion — and the misses are written back to the rules that caused them.
Your answer, your words
Keep the question bank you already paid for. Log what you missed and why you missed it — guessed, misread, didn't know — and the engine separates a lucky right from a real one.
Works with any bank
Hand-drawn attack sequences for all fourteen tested subjects, plus the crossover maps — the order you take a fact pattern apart, not another outline to read.
Free tier · 16 charts
A coach with hands, not a chatbot. It can read your board, build a drill set, move your schedule and stage a rule edit — and it has to pass an essence check before it's allowed to change a rule you rely on.
Paid · 18 tools
The CA performance test on your real clock, with the page-numbering and keyword-note technique built into the interface — because on a PT you lose points to the file, not to the law.
CA PTs · timed to your budget
Lunch ordered at 11:30 because that's when you eat. Groceries, refills, the appointment you've rescheduled twice — booked into the gaps between your blocks instead of eating one. The life admin that quietly costs you a study day a week, handled.
Paid · runs on your calendar
Rules ranked against each other, never against a fixed cutoff — so the list always has a real top. Each row states why it's there and what to do: retype it, or drill it.
The thing everything feeds
How it works
The steps are numbered because the order is the mechanism — measurement has to come before ranking, and ranking before work.
Type rules cold, write an essay, log the MCQs you missed. Every attempt is timed against your real budget.
Raw score, and true score — where a right answer you flagged as a guess counts as a miss. A miss weighs twice a lucky right.
Rules re-sort by relative weakness. A wide recognition-to-production gap sends a rule up; clearing both thresholds drops it out.
Today's grid is generated from the top of the board, sized to the hours you actually have, with the reason for every block stated.
The reports
Every other tool's unit of data is a letter. Ours is a letter, a confidence flag, and a sentence you wrote about your own reasoning while it was still live. That gap isn't a modeling advantage we might lose — it's data nobody else collects, and you can't reconstruct it after the fact.
Pick the flag you'd have used
So the report says
Every right answer you flagged as a guess or an unsure is scored as a miss, and a miss outweighs a lucky right. The number that comes back is lower than your percentage — and it's the one that predicts an essay.
Raw 79% · True 64%. The 15 points between them are all in Evidence.
Because every attempt carries its own exposure history, the report can separate a rule you never had from a rule you're losing. Those need opposite responses, and a percentage can't tell them apart.
You had CRP-07 in June and missed it twice since. This is decay, not a gap.
Four wrong answers in one subject can have four different fixes. The engine reads your own words and classifies each one — a rule you never had, a distinction you couldn't split, a fact pattern you misread, an outline you leaned on.
3 of 4 Evidence misses were the same last step, not the same rule.
The most expensive habit in bar prep is over-correcting — re-drilling questions you already reviewed because one miss looked alarming. Every report names what to leave alone and why.
Don't re-drill Torts. One miss in a 84% section is noise, and it cost a day.
This is the part no one can copy by shipping a better model. A question bank can only ever know what you clicked, so it can only ever tell you a percentage. We ask one more question — how sure were you? — and read one more answer: why you think you missed it. Every insight above falls out of those two, and neither can be added to a dataset that never captured them.
Setup
There is no import, no course to switch to, nothing to cancel. Click through the four steps to see what each one actually asks you.
Why this and not that
Bar prep already sells you lectures, outlines and questions. What nobody sells is a thing that watches you and changes what you do next. Compare on that, not on hours of video.
| A full bar course | A question bank | A private tutor | Bar Buddy | |
|---|---|---|---|---|
| What it gives you | Lectures, outlines, a calendar | Thousands of graded questions | A human who reads your writing | A coach that reads your writing at 2 a.m. |
| What it measures | Completion — videos watched, tasks ticked | Recognition — percent correct | Whatever fits in the session | Production — can you write the rule cold |
| When it says you're off track | At the simulated exam, weeks in | Never — the percentage is the whole signal | Next session, if you booked one | Same day, on the board |
| What a wrong answer does | Nothing | Marks it wrong and moves on | Gets talked through, once | Reranks tomorrow's work |
| Handles a repeat taker | Sells the same course again | Same questions you've already seen | Yes, at an hourly rate | Starts from the prior attempt's score report |
| Timing accommodations | Rarely modeled | Fixed clock | Ad hoc | Built in — the whole schedule runs on your budget |
| Protects your study hours | Adds to your to-do list | Adds to your to-do list | Another appointment to keep | ChoreBuddy takes the life admin off it |
| Runs alongside the others | Wants to be your only one | Yes | Yes | Designed for it |
Keep your course. Keep your question bank. Bar Buddy is the layer that turns what you're already doing into a signal you can act on — and it costs a fraction of any of them.
What we chose to be
Every bar product tries to be all of bar prep. Bar Buddy does one thing — measure what you can actually produce and act on it — and six deliberate choices flow from that.
We don't sell you a fourth question bank you'll never finish. Bring the one you already paid for — Bar Buddy turns it into a signal instead of a percentage.
Your course teaches the law. This tells you whether it stuck. No lectures to fall behind on, no second syllabus competing for your mornings.
The rubric is modeled on California's and built to be consistent, so you can see yourself improve week over week. We won't invent a predicted score — nobody can.
Fourteen subjects, the crossover maps, the CA performance test, the community property and PR variants. A general tool covers your exam. This one is your exam.
Our first cohort sits in February 2027 and scores land in May. Until then we'll show you the mechanism and let you judge it — not a pass rate we made up.
Email about a feature and you reach the people who ship it, not a support queue. Requests from the founding cohort land in the product in days, not release cycles.
Founding access
Founding members get the launch price, direct access while it's being built, and their feedback in the product before anyone else sees it. Pricing is announced when the doors open in November. No card now.