The ZCode invite window closes next Friday, November 13. After that, the campaign ends and the entry door becomes a normal subscription. But what matters here is not the window: it is the product underneath it, which I use every day.
What is verifiable
What the campaign grants, per Z.AI''s own official announcement: credits for whoever joins via invite and for whoever invites. Without published fine print, the exact values circulating are third-party reports, and I prefer to give what is verifiable.
What is verifiable is the product underneath the invite. ZCode is Z.AI''s coding agent with the GLM family inside: Flash at $0.15 per million input tokens and $0.50 on output, off-peak at 50% off, and the thesis I defended here two weeks ago, the consumption rule that yields more as the price falls.
The test that matters: daily use on a real project
My use is every day, on a real project, under heavy load. It is the tool I use to write, review and audit code, and this week''s posts on PHP2Shell and the kernel came out of it.
It is the same criterion I apply to any agent tool: a benchmark does not run a project. Daily use with real load is what reveals latency, cost and behavior, and what separates a demo from an operation.
The consumption rule is what remains when the campaign ends
A campaign is a window; a consumption rule is permanent operation. Cap, credit, cache and off-peak measured from day one are what turn the tool into an operation with known cost, in the same mechanics I documented in the 30-day commercial cycle of a model and in the off-peak lever.
And an agent that writes and runs code carries the same responsibility as any agent: human review of the diff, permission scope per task, and attention to configuration files, which were already the vector of the TrapDoor campaign.
Conclusion
The invite is in the first comment of the original post, and the channel I use is this one: https://z.ai/subscribe?ic=BGIRP61T2P (10% off the first order). If your stack runs code, the window closes on Friday. But the adoption criterion is not the campaign: it is the tool on a real project, with a documented consumption rule and agent governance. That is how the tool stops being test expense and becomes an operation with known cost.