Clean Signal Only: The Integrity Layer Behind the Grid

Every engagement platform says it filters spam. Then it pays out on raw volume and hopes nobody notices the bots.

Borged's entire reward model assumes the opposite: that the easiest thing to fake is activity, and the hardest thing to fake is signal. So the platform is built to reward signal and quietly starve everything else. GHOST_GRID puts it on the badge — "clean signal only" — but that line is backed by real machinery, not vibes.

Here is that machinery, layer by layer.

1. Deploys are scored before they earn

A deploy does not get reward weight just for existing. During finalization, every submission is scored by an AI model that reads the actual content — what you posted, whether it is on-topic, whether it reads like a person or a template.

The score lands on the submission and drives two things: whether it is approved at all, and how much payout weight it carries. Low compliance or a score under the approval threshold means the deploy is rejected with a reason attached — not silently dropped, not silently paid. Approved work keeps its reasoning on record.

The point: posting volume buys you nothing. A hundred low-context posts score like a hundred low-context posts. One sharp, original deploy outweighs them.

2. Answers are judged for substance, not keywords

Some netruns ask you to say something — answer a community question, reply to a prompt, engage a thread. The naive way to verify that is a keyword match. Type the magic word, get paid. That is a spam machine.

Borged's verification layer has three predicate kinds: a literal contains match, a regex match, and ai_judge. The interesting one is ai_judge: for prompt-based tasks it hands the message to a scorer with the actual requirements and gates the reward on whether the answer meets a threshold, not whether it contains a string.

"gm" does not answer a question about which protocol you are most bullish on. The judge knows the difference. Low-effort filler does not clear the bar.

3. Cadence gates so velocity cannot win

If output were free to repeat, the optimal strategy would be to spam. So it is not free.

Deploys run on a daily cadence — one per platform, per operator, per 24-hour window. And X reply netruns sit behind a 36-hour per-campaign rotation cooldown before they regenerate. You compete on consistency across days, not on how many times you can hit submit in an hour.

This is deliberate. A platform that rewards burst velocity rewards bots, because bots are very good at burst velocity and very bad at showing up thoughtfully for weeks. Cadence gates flip the advantage back to real operators.

4. Repetitive, low-context output is downranked

Even inside the rules, not all signal is equal. Output that is repetitive or low-context gets downranked in payout weight. The system is not just checking "did this pass" — it is weighting "how much did this actually move." Five distinct, well-aimed posts beat fifty variations on the same sentence.

5. Our infrastructure hiccups never cost you a point

Here is the part most platforms get wrong, because it requires admitting the platform itself can fail.

Sometimes verification fails for reasons that have nothing to do with the operator — a backend is briefly down, an external API throttles us, a request times out. A naive system counts that as a rejection and dings the operator's reputation. That is unfair, and it teaches operators to distrust the score.

Borged separates genuine rejections from transient failures. A real quality failure penalizes. A transient error on our side does not touch your rejected-count, your points, or your reputation — it just retries. Your standing reflects your work, never our worst five minutes.

What this means

For operators: the score is honest. You cannot out-spam the system, and the system cannot punish you for its own outages. Show up consistently with real signal and the weighting works for you.

For campaign creators: the budget routes to engagement that an AI read and judged real — not to whoever automated the most noise. Clean pools, not gamed ones.

For companies evaluating the platform: "AI-scored, judged, cadence-gated, transient-immune" is not marketing copy here. It is the verification stack, and it is what separates a growth platform that rewards authenticity from a leaderboard that rewards bots.

Clean signal only. The grid means it.