01
SURVEY
You describe the destination in your own words. It compiles into concrete skills and target levels against a canonical graph — never invented, only selected.
LEARNING PATH PLANNER
EVERY COURSE
IS A STATION.
LEARNING IS
THE LINE BETWEEN.
Catalogue search hands you a pile of stations and wishes you luck. Waypoint builds the line — working out which skills stand between you and where you want to be, then ordering the journey so you never arrive somewhere you are not ready for.
Or pick a destination yourself →No sign-up, plots in a second.
- DATA ANALYSIS
- WEB DEVELOPMENT
- FOUNDATIONS
- INTERCHANGE
0.0% PREREQUISITE VIOLATIONS. 93.0% SKILL GAP CLOSED. 0.392 AGREEMENT WITH EXPERT. 126 RESOURCES SURVEYED. 10 EXPERT-LABELLED ROUTES. 181 TESTS PASSING. Every figure is measured, not claimed.
A ROUTE, PLOTTED
Starting from nothing, aiming at Deep Learning Basics at a working professional level.
- STOPS
- 11
- HOURS
- 114
- VIOLATIONS
- 0
- REACHES GOAL
- YES
3Blue1Brown: Essence of Linear Algebra
5h3Blue1Brown · difficulty 2/5 · Linear Algebra Basics 0▸3
StatQuest with Josh Starmer
15hStatQuest · difficulty 2/5 · Descriptive Statistics 0▸3 · Inferential Statistics 0▸3 · Supervised Learning 0▸2 · Model Evaluation 0▸2
Kaggle Learn: Python
5hKaggle · difficulty 1/5 · Python Basics 0▸3 · Python Data Structures 0▸2 · Python Functions & Modules 0▸2
Kaggle Learn: Intermediate Machine Learning
5hKaggle · difficulty 3/5 · Scikit-learn Basics 2▸4 · Model Evaluation 2▸3 · Supervised Learning 2▸3
freeCodeCamp: Scientific Computing with Python
25hfreeCodeCamp · difficulty 2/5 · Python Functions & Modules 2▸3 · Python Data Structures 2▸3
Practical Deep Learning for Coders
40hINTERCHANGEfast.ai · difficulty 3/5 · Deep Learning Basics 3▸4
BOTH ROUTES ARE REAL PLANNER OUTPUT AGAINST THE LIVE CORPUS
HOW THE LINE IS BUILT
Four mechanisms, none of them a chat wrapper. The language model reads your goal and phrases the reasoning; it never chooses a course, never decides an order, and never sees the catalogue.
02
MEASURE
Your position minus the destination gives a gap, expanded across every prerequisite. Retrieval matches that gap, not your sentence, against what each resource teaches.
03
PLOT
Beam search over the prerequisite graph. Only resources you are ready for are ever considered, so an impossible ordering cannot be produced — not filtered out, never generated.
04
RE-ROUTE
Tell it a course did not work. Mastery moves, the line is re-plotted around it, and the diff names exactly what changed and why it changed.
MEASURED, NOT CLAIMED
Scored against 9 of 10 learning paths a human expert wrote by hand, alongside the semantic search this is meant to beat. Both saw the same catalogue of 126 resources and the same embeddings.
0.0%
similarity search: 49.0%PREREQUISITE VIOLATIONS
93.0%
similarity search: 64.7%SKILL GAP CLOSED
0.392
similarity search: 0.168AGREEMENT WITH EXPERT
0.459
similarity search: 0.111ORDERING CORRELATION
The last figure is the one worth staring at. A correlation of 0.111 means the baseline's ordering is indistinguishable from shuffling the results. It finds relevant courses and then hands them to you in an order nobody could learn from.
Measured 30 August 2026 · the baseline searches with the same embeddings we do