How skill practice works
Writing is split into small skills you can improve one by one. A skill might be clear cause-and-effect, better scene flow, or cleaner sentences.
Each assignment focuses on exactly three active skills. Those skills become your target for that draft, so feedback stays specific and useful.
When you show steady control of a skill, the app marks it as stronger and unlocks the next skill in the path. If an older skill starts slipping, it can pause new unlocks until that foundation is steady again.
What you get
- Assignments based on the skills you need most right now.
- Feedback tied to clear goals.
- Revision briefs when a draft needs another pass.
- A progress view that shows what is improving.
- A clear path to the next skill.
One clear target at a time
Most writers feel stuck when every draft is judged on everything at once. This app narrows the target so you can see what changed and repeat that improvement later.
Feedback you can act on
Reviews point to strengths to keep, problems to fix, and the next step to take.
Progress with memory
Old assignments, drafts, reviews, and revisions stay linked. That history helps the app choose better next assignments.
What makes this different
Each assignment is part of a training path. The app tracks what you practiced and what changed.
Each review is tied to that assignment's goals. Later, you can look back and still understand what each draft was training.
A clear, step-by-step review process
The app uses the same robust set of checks each time so scoring stays consistent. Those checks are focused on your skill goals for this assignment, then passed to AI for plain-language explanation.
The app tracks grammar flags, sentence structure, readability, style warnings, repeated phrasing, unclear references (like vague "it" or "they"), and whether key claims are backed by support.
The app picks a scoring guide based on your writing type and track. A technical memo and a story scene are judged in different ways because their goals are different.
Each assignment has a few skill goals. Scoring gives extra weight to those goals. Older skills are still checked lightly to catch backsliding.
Read the pedagogy paper
Want the academic grounding and design rationale behind this coaching loop? Read the pedagogy paper describing the hybrid objective model, research basis, implementation details, and limitations.
Check tools we use
These tools are part of our fixed-check layer. Most are open-source projects. We also use one built-in rules engine.
This is our built-in rule checker. It catches direct writing problems fast, like very long sentences or repeated weak patterns.
Vale is a writing linter. It scans your text for style problems, like unclear wording or tone problems, based on rule files.
Open projectLicense: MIT
LanguageTool checks grammar, spelling, and punctuation. It helps find sentence-level mistakes quickly.
Open projectLicense: LGPL-2.1-or-later
spaCy is an NLP library. It breaks text into useful parts so we can measure things like sentence length, passive voice, and clarity signals.
Open projectLicense: MIT
TextDescriptives adds extra text metrics on top of spaCy. It helps us track readability and other writing patterns over time.
Open projectLicense: Apache-2.0
CoreNLP is an optional helper service. When enabled, it gives stronger signal on unclear references like vague "it" or "they" chains.
Open projectLicense: GPL-3.0-or-later
How AI is used
AI is used in limited ways in Writing Coach, mainly to make the interface and feedback easier to read.
We collect your writing style, active skill goals, and recent assignment history to prepare the prompt context.
AI uses that prepared context to draft assignment text in plain language.
We run all fixed checks to collect evidence about your writing on an assignment draft or revision.
We apply the right scoring guide for your writing style and focus the review on your chosen skill goals for this assignment.
We pass the scoring breakdown and evidence to the AI layer so it can format feedback in plain language with clear, contextual notes.
Open-source and license attributions
These are key third-party tools and assets used in Writing Coach, with project links and license terms.
- React - Open project - License: MIT
- Next.js - Open project - License: MIT
- Tailwind CSS - Open project - License: MIT
- Headless UI - Open project - License: MIT
- Heroicons (icon pack) - Open project - License: MIT
- Inter (font) - Open project - License: SIL Open Font License 1.1
- Vale - Open project - License: MIT
- LanguageTool - Open project - License: LGPL-2.1-or-later
- spaCy - Open project - License: MIT
- TextDescriptives - Open project - License: Apache-2.0
- Stanford CoreNLP (optional) - Open project - License: GPL-3.0-or-later
- Ory Kratos - Open project - License: Apache-2.0
- SQLite - Open project - License: Public Domain
- Tailwind Plus / Catalyst UI materials - Open project - License: Tailwind Plus commercial license (see web/LICENSE.md)
- Writing Coach (this project) - Open project - License: GPL-3.0-or-later