Turning a sentence into a task without an LLM

Backlogs is a kanban-based to-do list application that lets users create tasks using natural language. Instead of filling out separate fields for a title, due date, and priority, users can type something like “Finish CS assignment tomorrow” or “Buy groceries next Friday high priority.” The parser extracts the deadline and priority automatically while leaving the rest as the task title.

Why I built it

I built Backlogs as my final project for Harvard’s CS50 Web Programming course.¹

On the surface, it’s another to-do list app. But I wasn’t interested in rebuilding CRUD for the hundredth time. I wanted to recreate one of my favorite productivity features without relying on an LLM.

Around that time, it felt like many “AI-powered” features were really just wrappers around an API. I wanted to prove that this particular problem didn’t need one.

A rule-based parser should be enough.

The kanban board was another challenge I wanted to take on. At that point, I had never built drag-and-drop interactions or an interface this dynamic using only Django templates, HTML, CSS, and JavaScript.

How it turned out

Natural language input

The parser understands natural language and extracts structured information while the user types.

For example,

Buy groceries next Friday high priority

becomes:

• Task: Buy groceries
• Due date: Next Friday
• Priority: High

As keywords are recognized, they’re highlighted immediately so users can see exactly what the parser understands before creating the task.

Responsive kanban board

Tasks can be moved between columns using drag and drop.

The board adapts to different screen sizes, switching from a three-column layout on desktop to a tabbed interface on mobile.

Task management

Each task stores its priority and due date, making it easier to organize work and decide what to tackle next.

Authentication

Users can register, log in, and manage their own personal backlog.

How I built it

The parser is entirely rule based.

I built it from scratch using regular expressions and a growing collection of parsing rules instead of calling an LLM. It recognizes relative dates like “tomorrow” and “next Friday,” explicit dates, and priority keywords before converting them into structured task data.

As I kept adding rules, I ran into the same lesson I’d already learned from my NLP projects: language is messy. Every time I thought I had covered all the edge cases, another input proved me wrong.

One part I’m particularly happy with is the highlighted input.

Instead of using a rich text editor, I overlaid a div on top of a normal text input. The div renders highlighted keywords while the input continues handling typing. It’s a bit hacky, but it gave me complete control over the experience without introducing a much heavier editor.

Task creation and updates happen through AJAX, so users can create, edit, and move tasks without refreshing the page.

The kanban board itself uses the native HTML5 drag-and-drop API together with JavaScript event handlers such as dragstart, dragenter, and drop. Managing all those interactions turned into one of the biggest JavaScript modules I’d written at the time, but it ended up working surprisingly well.

On the frontend, I leaned heavily on Django template includes to keep components reusable. Even small pieces like task cards and checkboxes became reusable templates instead of duplicated HTML. That habit probably came from using React. It just made sense.

What I learned

Backlogs reinforced something I’d been noticing across several of my projects.

Not every language feature needs a large language model.

For problems with clear rules and predictable behavior, a deterministic parser can be simpler, faster, easier to debug, and easier to explain. If I rebuilt this today, I’d probably use a hybrid approach, letting rules handle predictable cases while using an LLM only when the input becomes genuinely ambiguous.

The project also pushed my frontend skills much further than I expected. Building a responsive kanban board, implementing drag and drop from scratch, and designing an interactive parser without React forced me to understand how browsers actually work instead of relying on a framework.

Looking back, that’s probably what I’m most proud of: almost every interesting part of this project was built from scratch.