# Changelog

> Meaningful improvements to the SWE Interview Prep curriculum, practice loop, and personal learning system.

Canonical: https://learn.significanthobbies.com/changelog

# SWE Interview Prep changelog

Meaningful improvements to the curriculum, practice loop, and personal learning system.

- [Roadmap](https://github.com/Significant-Hobbies/swe-interview-prep/issues)

## 2026-09-03 — Drills grade what you actually wrote

- TypeScript solutions are graded instead of rejected, and a real syntax error now names its line and column instead of reporting an unlocatable "Unexpected token".
- Every auto-graded drill states the function its tests will call and the exact statement the grader runs, so passing no longer depends on guessing an unnamed symbol.
- Five drills that asserted an output no correct solution could produce were repaired, and "Full workspace" carries the drill you were on instead of opening an unrelated question.

## 2026-08-27 — Turn a job description into a learning target

- A new role-fit workflow maps exact requirement phrases from a pasted job description onto canonical curriculum concepts and names unsupported requirements honestly.
- The resulting target separates existing evidence from unverified and weak areas, expands prerequisites, and opens a focused Known, Fuzzy, or New Sweep without retaining the raw posting.

## 2026-08-09 — Trace a Tensor from model math to serving

- A new 30-day synthesis roadmap connects tensor representation and backpropagation to memory, kernels, hardware, inference scheduling, and serving economics.
- Its capstone requires a reproducible workload, a layer map, a measured bottleneck, and before-and-after evidence for one defended optimization.

## 2026-07-25 — A public curriculum anyone can inspect

- The complete curriculum became readable without JavaScript or sign-in across track, roadmap, and concept pages.
- Every generated page carries unique metadata, substantive learning content, internal navigation, and sitemap coverage.

## 2026-07-25 — Eleven more learning domains

- The curriculum expanded to 18 tracks and 222 concepts spanning systems, AI, developer tools, applications, and multimodal work.
- Each added concept includes a sequenced path, drill, review prompt, canonical source, and synthesis artifact.

## 2026-07-25 — A more honest learning loop

- Failed reviews can no longer appear mastered, and roadmap selection now persists through the canonical profile.
- Drills and artifacts are graded by evidence strength, with prerequisites and matching library sections visible.

## 2026-07-13 — Reader and High Signal joined daily learning

- Owner sessions can study current briefing items and private saved reading without copying source bodies into the static catalog.
- External items can be reviewed with FSRS, saved to notes, or opened as a focused Playground exercise.