Campaign period: May 2026. All metrics as of 7 September 2026. Kojiru is building a decentralised credit layer for AI agents. When the work started the project had a one-month-old X account, no media coverage and no search presence: a query for the brand name returned nothing about the company.
This case is not about reach. The numbers below are small in absolute terms and we are printing them anyway, because the outcome that mattered was owning the brand query and building a content asset that did not exist before.
Results
A search for the brand name now returns the Finbold article in Google News with the company logo attached, and it appeared there within minutes of publication. For a project with no prior search footprint that is the result; the follower count is a side effect.
What we did
1. Thread strategy
A written plan with five priority topics, each tied to one pillar of the product: the Agent Credit Score, bilateral credit lines, and WithdrawalGuard. Topics were ranked by how relatable the entry point was rather than by keyword volume, because the audience being addressed on X is builders and investors, not searchers.
2. Five threads written and published
- AI agents can think, but they still cannot pay – the clearest entry into the core problem, 11 posts
- Solving the AI agent cold start – all three pillars as one solution, 11 posts
- ACS 300-850: the FICO score for AI agents – explained through the most familiar model in finance, 9 posts
- Direct credit, not pooled risk – 10 posts
- WithdrawalGuard: principal protected by math, not trust – 9 posts
Between them the threads took the profile from 37 posts to 121 and gave the project a library it can reuse, which is the part that outlives the campaign.
3. KOL amplification: 8 accounts, selected before anyone was paid
Accounts were screened on Wallchain, Sorsa and median views per post rather than follower count, using the same method as our verified roster. Eight accounts were approved out of a much longer list, with a combined following of 815,900.
| Account | Views | Engagements | ER |
|---|---|---|---|
| Account 1 | 26,245 | 161 | 0.61% |
| Account 2 | 18,474 | 39 | 0.21% |
| Account 3 | 16,169 | 102 | 0.63% |
| Account 4 | 15,129 | 105 | 0.69% |
| Account 5 | 13,432 | 163 | 1.21% |
| Account 6 | 10,697 | 131 | 1.22% |
| Account 7 | 8,835 | 103 | 1.17% |
| Account 8 | 7,972 | 249 | 3.12% |
| Total | 116,953 | 1,053 | 0.90% |
The selection method under-predicted the result by 26%
Before buying anything we projected reach from each account’s median views over its recent posts, which came to 93,007. The campaign returned 116,953, or 26% above the projection. We publish this because the failure mode of median-based forecasting is usually the opposite direction, and because a projection that comes in low is the only kind a client should want.
The spread inside the roster is the argument for screening. The account with the most views returned an ER of 0.61%; the account with the fewest views returned 3.12%, five times higher. Ranking by follower count would have inverted the order of both.
4. Media placements
- Blockonomi – formal verification versus audits
- Finbold – FICO was built in 1989, AI agents need a score for 2026
Both articles were built from the same narratives as the threads rather than written separately, which is why the message is identical across the profile and the press. Finbold additionally distributed its article to its Telegram channel; we count that as distribution of the same placement, not a third one.
What this campaign does not prove
It does not prove reach at scale. 249 followers and 116,953 amplified views are small numbers and we are not going to dress them up. What it proves is that a project with no search presence can own its own brand query within one month on a modest budget, and that the content built to do it keeps working after the campaign stops.
Thread strategy, KOL screening and placement as a service are on our crypto influencer marketing page. The screening method is described in how to vet a crypto KOL in 20 minutes.