Let me clarify.

Federated Social Credit Management System (FSCMS) in COBOL is, in fact, a joke.

The code, however, is very real. Which is precisely why, if you write software, you immediately know it is a joke. The moment you see ISO-COBOL-9000 and ACTIVITY-LOG living beside CITIZENS.DAT, your brain goes, ah, yes, this is satire dressed up as enterprise compliance.

But if you are not a coder, let me walk through why you absolutely do not need to worry about this doing anything in the real world. Even if it compiles, it is hilariously impractical to the point of uselessness. I wrote it that way on purpose, because I thought it would be funny to build a “moral stability” framework on top of a language last used in anger to batch print mortgage statements.

Here is what this thing actually does.

First, it does not talk to the Fediverse. At all. The comments confidently say:

COMPREHENSIVE REPUTATION TRACKING FOR ACTIVITYPUB NETWORKS

In reality, there is zero HTTP, zero TLS, zero JSON, zero ActivityPub. No WebFinger. No OAuth. Nothing. The program reads and writes three local files:

• CITIZENS.DAT for your “citizen” records
• ACTIVITY.LOG for a text log
• REPORT.TXT for a quarterly compliance report

If you want it to “monitor” Mastodon, you do it manually. You sit at a green screen and type in what somebody did, then hit enter. This is not a global panopticon. It is a nerd with a terminal and too much time.

Second, it runs like a 1970s back office utility, not a modern service. The heart of the program is a main menu that looks like this in spirit:

1. Register new citizen
2. Process activity
3. View citizen
4. Generate quarterly report
5. Exit

You pick menu options one at a time. There is no daemon. No background worker. No scheduler. No scaling. No concurrency. If a million people post at once, nothing happens unless a human operator spends the rest of their natural life tabbing through “PROCESS ACTIVITY TRANSACTION” for each person.

Third, the “social credit” model is absurd on its face.

The system pretends to be serious, then assigns reputation based on a handful of hardcoded activity types like:

• POST-APPROVED-HOURS +15.00
• POST-AFTER-HOURS −25.00
• EMOJI-EXCESSIVE −08.00
• HASHTAG-PROPER +05.00

That is not AI. That is not behavioral analytics. That is a glorified punch card that says you used too many emojis, therefore your “COMPLIANCE-STATUS” is now MONITORING. This is closer to a Dungeons & Dragons homebrew alignment chart than a functioning social control system.

Fourth, it only knows what you type into it.

The core record looks like this conceptually:

• CITIZEN-ID
• CITIZEN-USERNAME
• CITIZEN-INSTANCE
• REPUTATION-SCORE
• COMPLIANCE-STATUS
• LAST-ACTIVITY-DATE and TIME
• TOTAL-ACTIVITIES

There is no identity verification. No signature. No cryptography. Anyone at the terminal can type anything they want and the program will happily accept that as truth. In a real surveillance system, this would be a catastrophic security flaw. In this parody, it is just another layer of ridiculousness.

Fifth, everything is stored in flat files.

Real systems with this kind of ambition use databases, queues, replication, backups, and all the boring infrastructure you need to not lose data when someone sneezes on a power bar.

Here, we have CITIZENS.DAT as an indexed file and logs accumulating forever in ACTIVITY.LOG. There is no rotation. No archival. No privacy model. If you delete the file, the “state” of the entire global reputational order disappears in a puff of smoke. One accidental rm and the regime collapses.

Sixth, the “quarterly compliance report” is just a text file with counters.

When you generate a report, the program walks through CITIZENS.DAT, tallies how many people are EXEMPLARY, COMPLIANT, MONITORING, or CONCERNING, computes a basic average, and dumps lines into REPORT.TXT like:

• TOTAL MONITORED CITIZENS
• AVERAGE REPUTATION INDEX
• A few lines of distribution

No charts. No dashboards. No alerts. No BI tooling. The whole thing could be replaced with a spreadsheet and a coffee. The only thing “enterprise” about it is the word ENTERPRISE in the comment header.

Seventh, the certification is made up.

CERTIFIED: ISO-COBOL-9000 COMPLIANT is not a real standard. It is a joke. It sounds vaguely like the kind of thing you might see in a vendor brochure and that is exactly the point. It is parody of corporate theater, not an actual compliance regime.

Eighth, the “moral stability of COBOL” is the punchline, not the foundation. There is a running gag that newer languages are “chaotic” and COBOL is stable, serious, and trustworthy. So I leaned into that and built an entire fake governance stack on top of it. The contrast is the joke. You do not actually get ethical guarantees by compiling on an IBM Z mainframe. You just get higher electricity bills.

So yes:

• The code compiles
• The menu runs
• You can register fake citizens and adjust their scores based on how seriously they treat hashtags

What you cannot do is run a real social credit system off this. There is no network, no integration, no security, no scale, and no realistic data model. It is a toy dressed up like a compliance department.

That said, I am very tempted to open source the whole thing purely for the comedy value, with very loud disclaimers that this is parody. If nothing else, it might be a useful artifact for future historians trying to understand why, at some point in the 2020s, someone decided that the correct response to AI panic was to build a federated social credit system in COBOL and then deliberately make it useless.

@atomicpoet Obviously a joke, but oh, to live in a world where COBOL was viable for crazy projects.