Kill the Standup: Async Status From a Shared Memory

The daily standup survives on a myth: that the meeting is where the information lives. Eight people, fifteen minutes, so that each can say out loud what most of them already did in plain sight. Multiply it out: fifteen minutes times eight people times five days is ten person-hours a week, spent narrating work, at the one time of day when makers are freshest. Distributed teams pay a second tax on top: someone’s standup is always at someone else’s dinner.

And here is the quiet failure: the standup does not even keep the information. Whatever was said evaporates by lunch. Ask on Thursday what Maya was blocked on Monday and nobody remembers, even though everyone was in the room when she said it.

The information deserves to survive. The meeting does not. And for a growing share of teams, the information already passes through an AI tool: the code gets written in Claude Code or Cursor, the plan gets argued in Claude, the customer reply gets drafted in ChatGPT. The work is narrated to a machine all day. It just needs to land somewhere the rest of the team can read.

The shape of the day

While people work, their AI tools save. Each person connects the AI tool they already use to the team’s engine, at Can read and save. As they work, they (or their tool, following a standing instruction) save what will matter to someone else later: what shipped, what was decided and why, and what is blocked, as a task with an owner. Not a running log of every step. One or two memories per meaningful unit of work.

In Claude Code, for example, one paragraph in the project’s CLAUDE.md makes this a habit rather than a chore:

When you finish a piece of work in this repository, save one short
note to the Team engine: what changed and anything a teammate would
need to know. If we made a decision, save it as a decision with the
reason. If I am blocked on something or someone, save a task that
says what I am waiting for and who can unblock it. Do not save
step-by-step progress.

Cursor has project rules, Claude has project instructions and ChatGPT has custom instructions; the same paragraph works in each. Every memory records who saved it and with which tool, so “Claude Code · Maya” is on everything Maya’s tool saved.

Anyone who wants the picture reads it, when it suits them. There are four places to look, and none of them needs a meeting.

• Home shows what arrived today. Memories has the same as Arrived today above the list, and its Filter narrows by Who, by AI tool and by Since, so “everything Jonas’s tools saved since Monday” is two clicks.
• Summaries, a tab in Memories, is a written account of each day, rolled into weeks and months, from what was saved. Read yesterday’s day in a minute; read the week on Friday.
• Ask, in the console, answers questions like “what happened while I was away?”, “what is Priya working on?” or “what is blocked, and on whom?”, and names the memories each answer came from. Answers about a specific person work once that person’s account is linked to their person memory.
• The brief. When any connected AI tool starts a session, it gets what each engine is for and what changed since that connection was last there. So Deniz’s tool starts the morning already knowing that Maya is waiting on him for API keys.

That last one is the part a standup never managed. The status does not just reach the people who asked for it; it reaches the tools doing the next piece of work.

What the standup picture looks like

Here is the kind of answer you get from Ask on a Tuesday morning. The names are invented and Ask’s wording varies, but every line points at a memory someone’s AI tool saved:

You: What moved yesterday, and what is blocked?

From 6 memories

Moved
  Maya: checkout flow behind the flag; PR up for review. [1]
  Jonas: closed three support escalations, including the
  Meridian ticket. [2]
  Priya: decided to keep annual pricing on the page, because
  it converts better for teams. [3]
Blocked
  Maya: waiting on API keys from the vendor since Monday.
  Task assigned to Deniz. [4]
  Jonas: needs access to the billing sandbox. Task assigned
  to Priya. [5]
Nothing saved yesterday
  Sam (last saved Friday: migration testing). [6]

Anyone who wants the standup picture gets it in thirty seconds, asynchronously, including the manager in another time zone and the founder who would never attend a standup. Note the last section: not an accusation, just the memory being honest about coverage. If you want a version for a chat channel, ask your AI tool for it and paste it, or let a tool with its own chat connector post it for you.

The week, without a status report

On Friday, open Summaries and read the week. It is written from what was saved: decisions made, tasks closed and still open, what each part of the team was doing. Nobody wrote a status report; the status was saved as it happened.

If you want a particular shape, ask for it:

Using the Team engine, write this week's update for the leadership
channel: decisions made and why, what shipped, what is still open
and overdue, and one thing to watch next week. Five short sections,
only from what is in the engine. Say what you could not find.

Open tasks are memories with an assignee and a due date, so “what is overdue, and whose is it?” is a question Ask or any connected AI tool can answer. Keep the tone instruction kind (“one short list, no guilt-tripping”). A status system the team experiences as surveillance gets routed around within a month.

What about the human part?

The honest objection to killing standups is that they were never really about status: they are the one moment the team sees each other. Keep that, deliberately. A weekly synchronous slot spent on things that need humans in a room, decisions, demos, arguments, morale, is worth having precisely because the status recitation no longer squats in it. If your standup truly is beloved, keep it and save to the engine anyway: the memory keeps what the meeting loses, and within a month you will notice which one the team would rather keep.

The pattern generalizes. Any recurring meeting whose real function is collecting status, weekly client updates, cross-team syncs, the Monday alignment call, is a reading job wearing a calendar slot. Decisions made in passing get caught by the decision log, commitments by action items that become tasks, all on the same shared memory, so nothing depends on the ritual surviving.

Tuning it

You tune this in the instructions people give their AI tools, and in where you choose to read.

Choice	Starting point	Alternatives
What to save	shipped, decided, blocked	add “started”, for longer projects
When to save	at the end of each piece of work	once at the end of the day
Blockers	a task, assigned to whoever can unblock	a note only
Daily read	Arrived today, or an Ask question	yesterday’s day in Summaries
Weekly read	the week in Summaries	a written update from your AI tool

Failure modes, and the fixes

People stop saving. No tool can summarize silence. The fix is social, not technical: the status must be visibly read, ideally by leadership asking about it. Saves that feed something people use keep happening; saves into a void decay in three weeks. The “nothing saved” line in the Ask answer is your early warning.

The memory fills with noise. If people’s tools save every step, Arrived today becomes wallpaper and the summaries get long. Tighten the instruction to outcomes, decisions and blockers. The instructions every connection receives already say not to save task progress; your standing instruction should agree.

Blocker tasks pile up unowned. A blocker without a resolver is a complaint, not a task. Assign the task to whoever can unblock it, and ask once a week, “what has been blocked longest, and on whom?”

Status lands in the wrong engine. Someone whose connection reaches a personal engine and the team engine may save to the wrong one. The AI tool picks by each engine’s purpose, so make the team engine’s purpose say plainly that work status belongs there. A misfiled memory can be moved with its history.

Build this yourself

1. Create a Team engine with a purpose, for example: “What this team ships, decides and is blocked on. Not personal notes.”
2. Invite the team, and have each person connect the AI tool they use at Can read and save. Claude Code is one command from the connect screen, such as claude mcp add --scope user --transport http ce-team https://app.contextengine.com/mcp/c/<id>, then /mcp in Claude Code, choose that server, and Authenticate.
3. Add the standing instruction above to each tool: CLAUDE.md, Cursor rules, Claude project instructions or ChatGPT custom instructions.
4. Read instead of meeting for two weeks: Arrived today or one Ask question in the morning, Summaries on Friday.
5. Link each person’s account to their person memory, so Ask can answer “what is Maya working on?”

See plans, or start free.