Roadmap is live as of this morning. Every Pro candidate with a target test date now sees a per-day study plan the engine has written for them — which surface to sit in today (diagnostic, mock, practice), what length, which focus, and why this day, not another. We have owed candidates this surface since the Pro 4-month launch in June 2024. Today is the version we are comfortable signing off on.
What it is
Roadmap is a single page at /roadmap. The top of the page shows the candidate's target date and the number of days remaining. Underneath is a chronological list of session blocks — one per day the engine has scheduled work — each with a length, a focus, a surface, and a one-line note describing what the engine wants the session to teach it.
The plan is not a recommendation. It is what the engine thinks is the best next session for this candidate right now, written in advance for as many days as the engine has confidence in. The plan rewrites itself every night, after the calibration pipeline runs and the candidate's mastery vector is up to date.
How the engine writes the plan
Three inputs, in order of weight.
The mastery vector. The same 30-element sub-skill structure that drives session routing now drives day routing. The engine asks, for each upcoming day: which sub-skill is currently most worth a session of attention, given what we know about this candidate's vector and what the calibration pipeline says we can learn from a session against it.
The target date. A candidate sitting in 47 days gets a different sequence than a candidate sitting in 12. The closer the date, the more the plan weights mock rehearsals and the less it weights exploratory practice on sub-skills outside the candidate's highest-leverage three. The engine does not ramp the plan blindly; the ramp follows the variance in the candidate's recent responses, not a calendar function.
The candidate's own schedule pattern. The engine reads which days of the week the candidate has historically sat sessions and at what length, and does not schedule work on days the candidate has never sat one. A candidate who has not opened the product on a Sunday since signing up will not see a Sunday block in their Roadmap. This is the rule we had the longest argument about and shipped most cautiously.
What we cut
Two features that lived in the prototype for months and did not ship in this version.
A daily-progress streak indicator. Same argument as Practice. The Roadmap reads the candidate's history, not the calendar. A candidate who skips Wednesday and sits a full forty-five-minute Practice on Thursday is in better shape than a candidate who logs a token five-minute session every day. The Roadmap will not pretend otherwise.
An "ahead of plan / behind plan" status pill. We built two iterations of this and discarded both. The Roadmap is rewritten nightly against the candidate's actual position; there is no fixed plan to be ahead of or behind. A candidate whose vector improved faster than the engine expected gets a shorter sequence; a candidate whose vector moved less gets a longer one. Status against an outdated plan is a measurement of the wrong thing.
How the rewrites work
Every night, after the calibration pipeline finishes the nightly re-fit, the planner re-runs. It reads the candidate's updated mastery vector, the days remaining, and the previous fourteen days of session density. It generates the next twenty-one days of planned sessions — twenty-one because that is roughly the horizon at which our internal validation says the plan stops being meaningfully better than chance against a candidate's actual trajectory.
Beyond day 21, the candidate still sees scheduled mock rehearsals and the recommended diagnostic re-take cadence — the surfaces whose timing the engine can plan with confidence further out. The per-day Practice blocks past day 21 fill in as the candidate moves through the window.
Who gets it
Roadmap is available today to every Pro 4-month and Pro 6-month subscriber. Pro 1-month subscribers see a seven-day version — a horizon long enough to be useful inside a 30-day window without claiming a confidence the data does not yet support at that commitment length.
A candidate on the free diagnostic does not see a Roadmap. Roadmap requires a target date and at least one re-take of the diagnostic to anchor the mastery vector against; both are inside the Pro surface. The diagnostic itself remains free and the path of least friction into the product.
What we will write about next
Section Analytics goes to beta in April. The Roadmap reads from the same per-section analytics the candidate will see in their own panel, and the next Product post will introduce the panel surface itself — what is visible, what the engine reads from it, and why a candidate who reads it weekly tends to sit a better mock six weeks later.
— Brightroom Product