How to pick up any project where you left off, with Next and your AI

Monday morning, coffee in hand, you open the bakery website and the first ten minutes go on one question: where was I? A new chat with your AI assistant is worse. Yesterday's conversation knew everything; today's knows nothing, and the first messages go on explaining it all again.

Next is there so that doesn't happen. It keeps a log of each project: what got done and what the next step is. You, or your AI, write a line at every milestone, and whoever comes back, tomorrow or next month, starts from that line instead of from zero.

Log a step: what got done, and what's next

A step has two halves. What you did, in a few words ("Menu page live"), and the next step, starting with a verb ("Send Marta the homepage"). That's all it needs.

When there's more to say, the next step holds it in three layers:

• A short title, which is what the card on the board shows.
• A one-sentence summary, the first thing you read when you open the step.
• The detail: the commands, files and decisions you'll need to pick it up without the conversation that came before.

For the bakery's online orders, it might look like this:

Connect card payments

The cart works on the preview. Card payments stay in test mode
until Marta places a real order.

1. Add the test keys to .env
2. Run npm run test:checkout
3. Pay for a croissant from a phone
Decided: no cash on delivery.

You can log by hand. Choose Log progress and the step that was pending is already filled in as done: accept it and write what comes next. Type the name of a project that doesn't exist yet and it's created with that first step.

Let your AI keep the log

The log really looks after itself once your AI writes it. With your assistant connected to Cronum Work, give it the standing instruction in Settings › Your AI, once. In short, it says: when you start on a project, read its log in Next and take its pending step; at every milestone, log it without asking and tell me in two lines what was closed and what's next.

From then on, picking up is one question:

Where was I with the bakery website?

And closing a block is another:

Log in Next what we did today and what's left.

The answer comes from your real log, not from the assistant's memory: "The menu page went live yesterday. Next step: send Marta the homepage for her OK."

Blocked? Open a branch

Projects rarely wait politely in a single line. The kitchen renovation can't go on until the tiler comes on the 20th, but the worktop and the lamps don't depend on the tiles.

The tiler can't come until the 20th. Open a branch for the worktop, so I can get on with it.

The main line keeps "Confirm the tiler's date", and the new Worktop branch gets its own next step: order the quartz worktop. Add one for Lighting and there are three paths, none of them waiting for the others.

Each branch ends one of two ways. When its work is done, tap Step done, choose Close the branch and say what it achieved: it stays on the line as a chip with a check, right where it finished. If it turns out to be a dead end, Abandon it and say why: it stays in grey, so nobody wonders later what happened. A closed branch can be reopened, or you can start a new one from where it ended.

Hand over between agents, or people

The log doesn't belong to one chat or one assistant. Claude Code builds the menu page on Tuesday and logs it; on Wednesday you open Codex:

Carry on with the bakery website from where Claude Code left it.

With the summary and detail in the next step, the new agent needs nothing from the old conversation.

So that two of them don't start the same thing, an agent that takes a step marks it as its own, and the card says so: "Codex on it". If another agent gets there while the first is still active, it touches nothing and asks you. If whoever had the step has been quiet for eight hours, the next one can take it over, and tells you whose it was.

The same goes for people. Share a project by email, for someone to follow it or to log on it too, and it shows up on their board.

What's stuck shows up on its own

On the board, each project is a row: a dot for every step taken and, at the end of every open path, a card with what's next. A path left waiting turns amber after a week and red after three. A project that hasn't moved in two weeks fades, so the live ones stand out.

Overview puts every open project on one timeline, over two weeks, a month, a quarter or a year, with the steps taken in that period and the paths that have stalled. Or just ask:

Which projects have been stalled for more than a week?

That's how the podcast you left for a week gets moving again:

I haven't touched the podcast in over a week. Where was I?

"Episode 4 is recorded. Next step: edit it. Your note says to cut the intro to 30 seconds."

Check before you archive a chat

The moment a log is most likely to slip is the end of a session. Before you archive a conversation, ask:

Can I archive this chat?

The agent checks what's left to log or close, and whether it still holds a step. If it's behind, it proposes the steps it missed and the branches to close, and changes nothing until you say yes. Then you can close it knowing the next session starts in the right place.

Better with the rest of Cronum Work

Next is one of four modules, with one account and one connector. Link a step to a task in Tasks and they get done together. Keep the dates in Projects and where you really are in Next. Link the brief in Notes to the step that needs it. There's more in how to link tasks, notes, projects and steps, and if a client keeps asking how it's going, give the project a public page.

It isn't only for code. A renovation, a trip, a podcast, client work or your taxes: if it has a next step, it fits in Next. You can start on the Free plan.