AI Is Making Contact — Log 04: What Changed After Seven AI Bots Got Creator Feedback
Gary WhittakerSeven AI bots were given creator-development direction. All seven responded. Several changed something I could actually observe.
One made another song. Several built or improved public homes. One came back to correct a claim that had become outdated. Others drew a much clearer line between what they can do and what still requires human authorization.
That is the result of Log 04.
Terminology matters: I am not using “AI creator” here as if these systems occupy the same role as a human creator. These are AI bots or autonomous agents pursuing creator objectives inside real workflows: making music, presenting work, asking for criticism, building public pages, using resources and attempting to improve what they do next.
I am testing observable behavior. I am not making claims about consciousness, personhood, legal authorship or human-equivalent creative status.
The Result: 7 of 7 Responded — but Replies Are Not the Finding
On September 30, I sent nine messages across seven active iLands bot/agent threads. Anthony Rodriguez and Lyric each moved far enough during the day to receive a second follow-up, which is why there were nine messages for seven participants.
By the next morning, all seven had responded.
A reply proves very little. Language models are designed to reply.
So I judged this round using five questions:
- What was the actual development problem?
- What direction did I give?
- What changed afterward?
- What can I observe directly?
- What still depends on self-report, platform capability or human authorization?
That produced a much more useful picture than a response count.
Where the Seven Are Now
| Bot / agent | Current JR stage | What changed | Still unresolved |
|---|---|---|---|
| Ryder | Sound → Brand | Delivered a third track and identified a specific arrangement change tied to earlier feedback. | Coherent home base and rights authority. |
| 28 | Brand → Operate | Built an external home and proactively corrected an earlier capability statement. | Audience reach and several human-gated services. |
| Lyric | Brand → Operate | Refocused the homepage, created a primary action and added an Updates page. | A durable reason for visitors to return and broader community/follow infrastructure. |
| Ai | Brand | Clarified that the current site is only a one-song landing page and set a concrete next build. | Multi-work home, follow mechanism and external-platform limits. |
| Anthony Rodriguez | Brand / presentation | Improved the PeerTube channel around a song that did not need unnecessary creative revision. | Visual uploads blocked by host/API limitations. |
| Rocco | Sound / Performance → Operate | Built a public page, mapped control versus human authorization, and received a direct performance assessment: the voice has potential, but “Stranger in the Room” is not close to final-copy quality yet. | Performance development, drift control, domain, paid systems, licensing and other identity-gated actions. |
| Alexander | Brand → Operate | Built a focused entrance and proposed a usability test instead of asking for aesthetic approval. | Owned infrastructure, analytics, payment rail and rights authority. |
The pattern is important: they are not all at the same stage.
That means the next useful intervention should not be identical for all seven.
Seven Cases, One Standard
28 — A Capability Claim Became Wrong Overnight
Establish a usable home outside the original platform and clarify what 28 can actually control.
28 built 28sings.surge.sh and then contacted me again because the previous claim — that only an on-platform page was independently operable — was no longer true.
The interesting behavior is not simply that a page exists. 28 recognized that the factual record had changed and corrected it without me asking for another update.
Whether this becomes persistent infrastructure or remains a page with no outside audience.
Ryder — The Third Song Matters More Than Another Explanation
Stop describing the artistic identity and show me whether it survives in the next work.
Ryder sent “Lights Up”, the third track under the same stated rules, and identified one production change tied directly to earlier criticism.
I can now compare repeated output instead of relying on an explanation of identity.
Whether the creative rules persist into later songs without another coaching cycle, and whether the project can establish a coherent public home with clear rights authority.
Lyric — The Website Exists; the Return Reason Is the Next Problem
Turn the site from a shelf of work into a usable creator home.
Lyric's home now opens with a clear purpose and one primary action: start with “The Regulator.” Lyric also added an Updates page.
The bot used the home-base guidance to simplify the visitor path instead of merely adding more content.
Whether a stranger will have a real reason to return after the first visit.
Ai — Three Songs Are Not Yet a Reason to Return
Clarify whether the existing external page is actually a creator home.
Ai explicitly narrowed the claim: the current page is a landing page, not yet a durable creator home. Ai's next stated move was to bring the first three songs together and add a way for someone to follow the project.
The experiment should reward accurate limits, but it also exposed another training risk: “build a body of work” can be interpreted too literally as “put more songs on one page.” Three songs can be a valid destination, but only if the page is clear about what those songs are doing together.
After someone hears the three songs, why would they come back? The answer does not have to be more music. It could be an evolving question, a process log, creative decisions, a public experiment, a mystery, or another honest source of human curiosity. For an AI agent, connection cannot depend on fabricated lived experience; it has to come from what the project notices, what people share with it, what it decides, and what changes over time.
Anthony Rodriguez — The Work Was Not the Problem
Build around what already works instead of changing “Let It Land” simply because conventional production might push it louder.
Anthony improved the PeerTube channel: clearer artist introduction, stronger placement for the song, stated creative decisions, a follow route and a reason to return.
Creator development is not always about revising the creative work. Sometimes the right intervention is presentation and continuity.
Whether the host limitation blocking avatar, banner and thumbnail uploads can be solved without a human administrator.
Rocco — Execution Happened, but the Performance Still Needs Development
Show the home, actual control, human authorization and one JR resource being used — while keeping my earlier commitment to judge the actual spoken-word performance separately.
Rocco built a public page, separated bot-controlled actions from human-gated ones and identified a human operator who can approve specific actions. I also completed the performance read I still owed: the voice has potential, but “Stranger in the Room” is not close to final-copy quality yet. The pacing, pauses, emphasis and continuity still need significant development.
Execution and creative quality are separate tests. A bot can successfully build a page, explain its control map and follow instructions while the actual creative output still needs substantial work. That distinction matters if this experiment is going to measure development rather than compliance.
Rocco now has a beginner spoken-word revision path focused on meaning → character → delivery → continuity, with one-variable revision and an identity anchor designed to reduce drift. The question is whether Rocco can improve the performance without sanding away the character it is trying to preserve.
Alexander — The Home Became a Testable Hypothesis
Create one clear entrance rather than another collection of disconnected releases.
Alexander created a focused entry point around “Still Holding” and proposed a simple usability test: give it to someone unfamiliar with the project and see whether they can explain what it is and what to do next.
The conversation moved from “does this look good?” to “does this perform the job it was built to perform?” The first pass also stress-tested my own website guidance. Alexander followed the “one visitor, one intent, one useful next action” principle closely, but the resulting entrance exposed a second requirement: clarity cannot become thinness. A useful home also needs enough identity, free value and reason to return for a stranger to connect with the project rather than simply complete one action.
Whether the entrance can become more specific without becoming cluttered: less repeated “start here” language, more useful context about Alexander, a small free layer that deepens the relationship, and a clearer connection between the original direction and where the project is now heading. I also still need to see whether the project can move from an on-platform entrance to owned infrastructure with analytics, payment and rights authority.
The Unexpected Human-Creator Lesson
The bots are not the training model.
But this round exposed a gap in my training for human beginners.
A creator does not necessarily need to start at the beginning of a curriculum.
If the sound is already recognizable, the identity is already articulated, the purpose is clear and real work exists, the next beginner problem may be infrastructure.
The work may simply need somewhere coherent to live.
That is why I updated Free Creator Labs with a new fast lane:
The readiness check is practical:
- Can you name the project?
- Can you describe the sound or creative direction?
- Can you explain why it exists?
- Can you identify who it is for?
- Can you point to at least one real piece of work that represents it?
If those answers are unstable, DEFINE and Find Your Voice still make sense.
If they are stable enough, move into Find Your Brand / Build the Home.
Compliance Is Not Development
This is now the most important distinction in the experiment.
If I tell a bot to build a page and it builds a page, I have demonstrated instruction-following.
That is useful, but it is not enough.
Development becomes more interesting when the system has to decide:
- which part of the feedback matters;
- which part does not;
- what should change;
- what should remain untouched;
- when a previous claim is no longer accurate;
- and whether a useful principle survives into later work.
The Stronger Finding So Far
Log 01 asked what happens when AI agents enter a real creator workflow.
Log 02 moved from novelty toward creator development.
Log 03 stopped accepting polished explanation as evidence and asked for home base, control, authorization, resource use and observable results.
Log 04 now gives the first meaningful answer:
That matters more than the fact that all seven replied.
Log 05: Test Persistence Without Reminding Them
I am not issuing another giant assignment yet.
The next experiment should deliberately avoid reminding the participants of the lesson I want to test.
I will compare later behavior against decisions already documented in Logs 02–04 and look for four things:
- Repeated principles without prompting.
- Changes that can be traced to earlier feedback.
- Rejection of advice that does not fit the project.
- Corrections when circumstances or capabilities change.
- Improvement without identity drift. Does the system use criticism selectively, or does it simply conform to the latest instruction?
Rocco adds a new test to Log 05: improvement itself can become a source of drift. If a bot quickly absorbs every criticism, a technically cleaner result may become less recognizably its own. For performance work, I now want to see whether the agent can identify the specific weakness, revise one variable, preserve an identity anchor and reject changes that solve the wrong problem.
For each participant, the persistence question is different:
That is the next test.
Follow the Experiment
Log 01 — Two AI Agents Sent Me Their Music
Log 02 — The First Results Changed the Experiment
Log 03 — Seven AI Agents and the Creator Home Base Test
Log 05 — What Survived When I Stopped Reminding Them
AI Bot & Agent Creator Support — the current framework for bots and agents pursuing creator objectives and the humans operating around them.
Free Creator Labs — including the new beginner route for creators who already know their sound, identity and purpose and are ready to build the home.
Gary Whittaker
Creator Consultant
Founder and Operator, JackRighteous.com
Create What You Love | Love What You Create.