FIELD NOTE / INSTAGRAM
Four signals of attention debt.
The complete written thought and the evidence behind it. The video edition will follow its public release.
The written argument is here.
This approved Instagram edition is on the journal now. Its video player and original platform link will appear after each public release is verified.
Four signals of attention debt.
Video caption
Four signals of attention debt.
Many plans may rely on the same unstated assumption.
Pick a sample and trace each plan backward.
Do this because output volume is not organizational memory.
#EricFieldNotes
Full written post / accessibility read
I would not try to measure an AI attention bubble by counting prompts. I would watch whether consequential decisions lose owners, source links and limiting conditions, and whether reversals and exception work grow after AI-generated handoffs scale.
Imagine a fictional organization with a wall of completed AI-created plans. Five depend on an approval that expired, but no decision card carries the expiry. The portfolio looks diverse; the failure mode is common.
For every sampled commitment, record source ID, owner, condition, expiry, dissent and actual result. Compare the last task with the source. Count missing fields and time to reconstruct the chain, not only task completion.
If the gap trend rises, add a source-readback gate before more agent-to-agent automation. If it falls, keep measuring. 'Attention debt' is my proposed risk lens; only your own audit can establish whether it describes your company.
#EricFieldNotes
Four-beat scene transcript
1. Four signals of attention debt.
I would not try to measure an AI attention bubble by counting prompts. I would watch whether consequential decisions lose owners, source links and limiting conditions, and whether reversals and exception work grow after AI-generated handoffs scale.
Visual: Orphaned decisions, lost links, reversals and exception backlog.
2. A green portfolio can hide one shared dependency.
Imagine a fictional organization with a wall of completed AI-created plans. Five depend on an approval that expired, but no decision card carries the expiry. The portfolio looks diverse; the failure mode is common.
Visual: Many plans may rely on the same unstated assumption.
3. Run a portfolio source audit.
For every sampled commitment, record source ID, owner, condition, expiry, dissent and actual result. Compare the last task with the source. Count missing fields and time to reconstruct the chain, not only task completion.
Visual: Pick a sample and trace each plan backward.
4. Treat a rising gap as a control signal.
If the gap trend rises, add a source-readback gate before more agent-to-agent automation. If it falls, keep measuring. 'Attention debt' is my proposed risk lens; only your own audit can establish whether it describes your company.
Visual: Do this because output volume is not organizational memory.
Research and claim limits
- Mohamed et al., LLM as a Broken Telephone, ACL 2025 (S167)
- Perez et al., When LLMs Play the Telephone Game, ICLR 2025 (S168)
- NIST AI RMF 1.0 (S171)
The examples identified as illustrative or simulated are design probes, not reported incidents. Vendor specifications do not establish workload performance.