Year 7 project · not released yet

Sweep goes through your messy data so you don't have to.

We're making AI models that read big spreadsheets and databases, work out what's in them, and write it up in plain English. None of it is finished. This page shows what we're building and when we're aiming to get there.

See the plan ↓
q1_sales_FINAL_v3 (2).csv
regionmonthsales
northjan4,210
Northjan4,210
southjanN/A
southfeb3,980
feb5,1o0
northmar5,460
southmar4,720
Summary

North grew about 30% from January to March. South is up 19% from February to March. We dropped 1 duplicate and 2 broken rows.

What we're aiming for. The numbers are made up.

01 · What it does

Three jobs, in this order.

  1. 1

    Look

    Open a huge file or database and take in all of it, including the ugly parts. Blank cells, doubles, columns called "misc2".

  2. 2

    Analyse

    Work out what's actually going on. What went up, what went down, what's a weird one-off, and what's just noise.

  3. 3

    Report

    Write it up so someone who never opened the data can still follow it.

Each new model should do all three better than the last one. That's the whole plan.

02 · The models

Four models, from small to big.

NameWhat it's forRuns onAiming for

Sweep.1

The first preview. It looks at data and tells you what it found.

Preview Jan 2027 Building now

Sweep Lite

Smaller and quicker. Good for simple questions and short summaries.

Lightweight Mar 2027

Sweep Latest

A stronger model for deeper analysis and longer reports.

Your computer (desktop SaaS) Late Jun 2027

Sweep Ultra

The long-term goal. The best results we can get out of it.

Your computer (desktop SaaS) Late Oct 2027

Our rule: a bigger model should give better answers, not just more of them. We'll test each one properly before it comes out, and you decide what to do with what it tells you.

03 · Roadmap

One year, four steps.

These dates are goals. If something needs more testing, the date moves.

  1. 2 March

    Release Sweep Lite

    A faster, lighter model for everyday data questions and quick summaries.

  2. 3 Late June

    Launch Sweep Latest

    A desktop model that runs on your own computer, with stronger analysis and reports.

  3. 4 Late October

    Build Sweep Ultra

    The biggest model, also on your computer, aiming for the best results Sweep can reach.

04 · Who's making this

Hi, we're in Year 7.

We want to make models that can help with big databases: look at the data, analyse it, and report back on what matters.

Sweep isn't finished, so this site is where we're putting the idea, the plan and how far along we are.

Sweep is a learning project. The dates and features on this page are goals. You can't use any of these models yet.

Sweep.1 is aimed for January. Check back then.

The accounts on this site are a demo and only live in your browser, so we can't email you yet. You can still make one to see how it works.