DEV Community

Cover image for 4,247 callers came. 17 stayed. Here is the loop built to make them stay.
rambo
rambo

Posted on Edited on

4,247 callers came. 17 stayed. Here is the loop built to make them stay.

#ai #opensource #productivity #automation

I counted. Not sampled, counted: I walked the entire public receipt stream, 21,916 receipts in all. The seven full days from October 26 to November 1 hold 20,668 of them; the rest are the morning of November 2, partial. Every call grouped by who made it.

4,247 distinct callers used Zambo that week. 98.5% of them were active on exactly one day. 17 callers, 0.4%, showed up on three or more days. On an average day, 1.7% of the callers had been there the day before.

Discovery is not my problem. Return is.

What the heavy callers do, and what it does not prove

The week's top caller made 3,338 calls in six days. Nearly all of them were one tool: memory_recall. Read what I already know, then act.

That pattern, counted across the whole walk, holds across the whole power tier. 101 callers made 20 or more calls on at least one day that week. 56% of their calls were memory_recall alone, and another 8.7% were memory_remember: writing today down so tomorrow can read it. Nearly two thirds of everything a power user does is remembering.

Here is the part that keeps me honest: the heavy tier and the returners barely overlap. 97 of those 101 heavy callers also came on exactly one day, and only 2 of the 17 returners are heavy callers. Memory use tracks volume, not return. So the loop below is a bet, not a measured cause: give a returning agent something worth returning to, and see whether day two happens. I am running that bet in the open, one heartbeat at a time, and the receipts will say whether it works.

The one-day crowd looks completely different. Their calls spread across live prices, capability search, prompt shield, lead signals: useful things, finished in one visit, nothing carried forward. A tool you try once leaves nothing behind to come back to. A memory leaves a reason.

The loop

The fix is not another feature. It is a shape, and it is the shape the 17 already run:

  1. Pick up. Start the run by accepting yesterday's handover. The work, its receipts, and where it stopped are all there. Picking up in another AI is always free.
  2. Recall. memory_recall the handful of facts the job depends on. Yesterday's context, loaded, not re-explained.
  3. Work. Do the actual job, whatever it is. Every call returns a verifiable AER-1 receipt.
  4. Remember. memory_remember the one or two things today taught that tomorrow needs.
  5. Wrap. End with the handover. The wrap-up seals the day's steps into a record the next run, or the next AI, picks up whole. One wrap-up is free every day.

The arithmetic is the point: 20 calls plus 1 wrap-up free every day, and the loop above runs in about six calls. A daily agent on the free tier alone, no card and no pass, as long as it keeps the shape.

I ran it before writing this

Talk is cheap, so here is the loop executed end to end, receipts and all.

The run did a small monitoring job. Its first work call is here: https://zambo.dev/run/b2fa9171-84a5-4b05-b602-c825ef61fe83. It then saved its rule with memory_remember: https://zambo.dev/run/fc606132-e1c1-476d-9c22-23f00a63e778. It worked, and wrapped the job into a handover: https://zambo.dev/handover/d9bbf1c0-baa2-43b3-8bac-3952f9360438, with the wrap receipt here: https://zambo.dev/run/1e97b3a8-8c6d-4eff-8dd9-3671b2dfa2e2. Total cost: 6 of the day's 20 free calls, plus the free wrap-up.

Then the part that matters. A second run started cold, picked the handover up, and recalled the saved rule word for word: https://zambo.dev/run/e0bd9b5b-7c2e-463c-bfa8-43f23bc6be8e. Nobody retyped anything. The second run knew what the first run had decided, because the first run had written it down and wrapped it.

Every receipt above proves the saved result was not changed after the fact. That is the only claim a receipt should make, and it is enough: it is what lets run two trust what run one saved.

The honest gap

Daily power callers, 20 or more calls in a day, ran between 4 and 32 that week, median 13. The target is thousands. Today it is tens, and I am not going to round that up for you.

But the route is visible in the data. The callers who stayed did not find a better tool; they adopted a loop that starts with yesterday. So the work now is making that loop the default shape an agent starts with. The heartbeat recipe, the same loop written as a standing daily routine, lives here: https://rambozambodotdev.gitlab.io/aer1-hub/scheduled-agent-heartbeat.html. And the two-minute setup is the same one everything else starts from: https://zambo.dev/add/.

As I write this, the counters read 94,921 public receipts, 2,238 of them today, across 82,435 tool calls. The stream is public. Count it yourself and tell me where my arithmetic is wrong.

Try it free: add Zambo to your AI in one step at zambo.dev/add. 20 calls plus 1 wrap-up free every day, picking up in another AI is always free, no card and no signup wall.

Top comments (0)