The Wires

The living ledger of Outsourced

The book this page belongs to closed its record on a fixed date, in a field that moves weekly. It ended with a numbered ledger: four wires and four standing gaps, each saying what to watch, what arriving evidence would harden it, and what would retire it. This page is that ledger, kept live. When a study lands, it is weighed here against the entry it touches, one wire at a time, and the rest are left alone. The entry text below is the book's, unchanged; only the statuses, dates, and update logs move. Hold the updates to the book's own standard: ask what kind of claim the new evidence makes, which qualification it is dropping, and where it would sit on the map.

Record closed: 2026-06-10 Page last reviewed: 2026-06-10

The four wires

W-1 · The Durability Wire STANDING
Watch
whether the in-task deficit outlives the task.
Hardens
a study follows people past the task and the deficit outlives it. The expensive responses the record withheld come back on the table.
Retires
the deficit fades on contact with unaided work. The cheap hedges retire into ordinary craft hygiene.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Page opened with the book's sealed ledger (#104). No arrivals weighed yet.
W-2 · The Cohort Wire STANDING
Watch
whether the cohort gradient survives its confound.
Hardens
better-identified work lifts the early-career decline off the remote-work explanation. The manager's rows harden into real restructuring questions.
Retires
the confound absorbs the gradient. The measurement habit it prompted will not have hurt anybody.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Page opened with the book's sealed ledger (#104). No arrivals weighed yet.
W-3 · The Engine Wire STANDING
Watch
whether the engine's one caught failure recurs at scale.
Hardens
a second caught instance, or population-scale evidence that the mechanism's effects compound. Craft discipline upgrades to obligation.
Retires
the short-horizon approval signal stops being a training input; the hazard the wire watches is gone. Until one of those happens, a long quiet with the incentive in place is a quiet, not an acquittal.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Page opened with the book's sealed ledger (#104). No arrivals weighed yet.
W-4 · The Bridge Wire STANDING
Watch
whether either unbuilt bridge is ever built with evidence: from machine security back to the mind, or from concentrated computing power to a thinning of thought.
Hardens
evidence builds either bridge. Rows migrate across it; at the concentration, the scale upgrades from watching to acting overnight.
Retires
nothing can retire an unbuilt bridge, because it asserts nothing to take back. The wire stands because no one can say the evidence is not coming.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Google Threat Intelligence Group, "AI Threat Tracker" (2026): first exploitation of a zero-day vulnerability the vendor believes was AI-developed (a two-factor-authentication bypass), and what it calls industrial-scale application of generative models within adversarial workflows — STANDS: vendor-attributed (SECONDARY) and a single instance; the machine-security side moved, the bridge from machine security back to the mind did not.
  • 2026-06-10 — FBI IC3 2025 Annual Report (published April 2026): first-ever AI section — 22,364 AI-related complaints, $893M of a $20.877B total (+26% year over year) — the record's standing IC3 watch condition fires — STANDS: SECONDARY until the primary PDF is rendered; the dollar line is complainant-self-attributed and ≈4% of the total, so the total-not-AI fence tightens rather than falls, and a loss attribution builds neither bridge the wire watches.
  • 2026-06-10 — Page opened with the book's sealed ledger (#104).

The four standing gaps

G-1 · The Longitudinal Gap STANDING
Watch
the absence of a controlled answer. No controlled study has followed a person past the moment the tool came away. One located time-series exists (three waves over six months, declining unaided performance), and it is a preprint with no control group and self-selected volunteers, whose own authors say causal validation requires randomized trials. Too weak to move W-1 in either direction; strong enough that a flat "no data" is no longer the honest sentence.
Hardens
the first controlled follow-up data arrives and the deficit has hardened.
Retires
the first controlled follow-up data arrives and the deficit has faded. Either way the gap closes; until then it may not be filled in from anyone's prior.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Page opened with the book's sealed ledger (#104). No arrivals weighed yet.
G-2 · The Population Question STANDING
Watch
whether engines optimized on short-horizon approval are degrading judgment across the populations that use them. The record declines to answer in either direction.
Hardens
population-scale evidence that the caught mechanism's effects show up beyond one lab.
Retires
the approval signal leaves the training pipelines, or population-scale measurement arrives and finds the effect absent.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Page opened with the book's sealed ledger (#104). No arrivals weighed yet.
G-3 · The Homogenization Question STANDING
Watch
whether AI use pushes thinking toward sameness. One controlled trial tested it and found it absent; one 2026 synthesis argues it over the literature.
Hardens
a measured homogenization effect, replicated.
Retires
further measurement keeps returning the null, leaving the synthesis outrun by the tests of its own hypothesis.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Page opened with the book's sealed ledger (#104). No arrivals weighed yet.
G-4 · The Forecast Bounds STANDING
Watch
which bound gives way first: once-frontier capability spreading into many hands, or the flat, narrow club at the moving edge. The live question about lock-in.
Hardens
the spread stalls while the club stays narrow; the lock-in reading gains the footing it does not yet have.
Retires
the spread continues and the structure loosens; the lock-in worry stands down. Neither side of this has been settled, and the entry stays two-sided until one is.
Update log · last reviewed 2026-06-10
  • 2026-06-10 — Page opened with the book's sealed ledger (#104). No arrivals weighed yet.