September 25, 2026
It Took Me 55 Versions to Learn One Make Lesson: Stop Waiting
God Mode Intel made the finals of Make's MCP community challenge. The folder behind it has 55 private scenario versions built in three days, with names like minimal, barebones, and ultra-simple. The one that worked had two scenarios and a callback, and nothing in it waited for anything.
Make put me in the finals. The folder knows what it cost.
In February, Make named God Mode Intel one of six finalists in its community challenge, Scaling the Future with Make and MCP. Their writeup describes a custom MCP server with 59-plus B2B intelligence tools (it is past 79 now) running lead enrichment across two Make scenarios, one to trigger and one to handle results, so the system never times out, even on multi-hour data runs.
That last part reads like a design decision. It was a surrender, and I have the receipts.
v11 through v60, January 28 to January 30
The God Mode Intel repo has a blueprints folder. In it are 55 private scenario versions, v11 through v60, all dated between January 28 and January 30. I did not rename a single one for this post. They did not need it.
v12-polling. v13-get-wait. v14-tools-sleep. v15-sync-dataset. That is four different ways to wait, in four consecutive versions. Then v36-robust, which was not. Then v42-minimal, followed immediately by v43-barebones, because minimal turned out to have too much in it. v45-null-fix. v50-minimal-debug. v51-ultra-simple. Read in order, the filenames are the five stages of grief, formatted as JSON.
Then, near the end: v46A-campaign-trigger and v46B-callback-handler. v57-make-first. v59A-async-trigger and v59B-callback-handler. v60-combined-trigger-callback. The answer showed up around version 46 and it took another fourteen versions to believe it.
Everything has a clock. The pipeline was longer than all of them.
Make stops a scenario at 40 minutes. The error is refreshingly honest about it: MAXIMUM EXECUTION TIMEOUT [40 minutes]. Make's own MCP server docs carry the same number: when a tool call times out, the scenario keeps running for up to 40 minutes and then it is done whether you are or not.
The enrichment itself runs on Apify actors, and Apify's run API will wait for an actor to finish for at most 60 seconds. My own README from that week still asks it to wait 120. That is the whole lesson in one parameter.
So every version that waited, whether it polled, slept, or asked nicely, was trying to fit a multi-hour job into a 40-minute room with a 60-second doorbell. You can make the room tidier. You can make the doorbell louder. The job is still bigger than the room.
Two scenarios. One fires, one listens.
Scenario one takes the request, starts the Apify run without waiting for it, hands the actor the address of scenario two's webhook, and ends. It lives for seconds. Scenario two starts with a Make custom webhook. When the work is done, the actor rings that doorbell, and scenario two pulls the results, writes the records, and posts the summary to Slack. Nothing in the system stands around. The 40 minutes stop mattering because no scenario lives long enough to meet them.
This is not my invention. It is the answer the Make community gives every time someone hits the wall: end the first scenario with a call to a webhook that starts the second. I had read that answer. I just needed a few dozen versions of personal experience to agree with it.
It is also why Make's writeup could say the pipeline scales from 1 lead to 10,000. Ten thousand leads is not a longer scenario. It is more callbacks. The same writeup credits it with a real shift off ZoomInfo, a 12 percent lift in meeting attendance, and a quarter that beat Q1 by 30 percent, and none of that would exist if version 12 had been allowed to keep polling.
Automation is not doing things faster. It is not standing there while they happen.
If any step can take longer than a minute, do not wait for it. Start it, give it a number to call, and leave. Design the callback scenario first (v57 was literally named make-first, which in hindsight was the only good idea in the folder that week). And name your versions honestly. Future you deserves to know that barebones came after minimal.
The actors on the other end of that callback, and what people actually run, are in the Apify piece. The Zapier version of this lesson, where the answer was fewer steps instead of more scenarios, is here.
If you want to build the two-scenario version yourself, start a free Make account and put the webhook scenario in first. Skip versions 12 through 45. I already ran them for you.