navigateenter openesc close
Every Tuesday

Failures

In most of these cases the technology worked. That is the finding, not the caveat. Fifty-two case files, each answering one question: what structural condition made this fail, and how would a buyer have detected it before spending.

The six modes

One primary mode per case
ModeNameWhat it means
1No workflowThe system functioned and was never wired to the moment a human decides. Output arrives beside the decision rather than inside it.
2No data pathThe system cannot reach the data it needs from inside the institution, so humans carry the data to it by hand.
3No ownerNo named person is accountable for the decision the system was built to change, so no one can authorise the change in practice.
4Wrong bottleneckVisibility, prestige or ambition selected the use case rather than a measured constraint. The system solves something that was not costing anything.
5No baselineNobody measured what the humans did before, so improvement can never be demonstrated and the programme is defended on anecdote.
6No stop conditionNo agreed circumstance under which the programme would be halted, so it is extended instead of judged.

The modes describe structure rather than technology, which is why they survive changes in the technology. The primary mode is the one whose removal would have changed the outcome.

The calendar

51 scheduled
Failures15 Sep 2026

The Epic sepsis prediction model

ScheduledTuesday
Failures22 Sep 2026

Air Canada and the Moffatt decision

ScheduledTuesday
Failures29 Sep 2026

The Optum care management algorithm

ScheduledTuesday
Failures6 Oct 2026

McDonald's and IBM at the drive-thru

ScheduledTuesday
Failures13 Oct 2026

Zillow Offers

ScheduledTuesday
Failures20 Oct 2026

Australia's Robodebt

ScheduledTuesday
Failures27 Oct 2026

Amazon's recruiting engine

ScheduledTuesday
Failures3 Nov 2026

The Replit agent and the production database

ScheduledTuesday
Failures10 Nov 2026

Klarna's customer service reversal

ScheduledTuesday
Failures17 Nov 2026

Mata v Avianca

ScheduledTuesday
Failures24 Nov 2026

Knight Capital

ScheduledTuesday
Failures1 Dec 2026

Chevrolet of Watsonville

ScheduledTuesday
Failures8 Dec 2026

The Dutch childcare benefits scandal

ScheduledTuesday
Failures15 Dec 2026

COMPAS recidivism scoring

ScheduledTuesday
Failures22 Dec 2026

CNET's AI-written finance articles

ScheduledTuesday
Failures29 Dec 2026

Microsoft Tay

ScheduledTuesday
Failures5 Jan 2027

Michigan MiDAS

ScheduledTuesday
Failures12 Jan 2027

DPD's customer service chatbot

ScheduledTuesday
Failures19 Jan 2027

Google Gemini image generation

ScheduledTuesday
Failures26 Jan 2027

Mobley v Workday

ScheduledTuesday
Failures2 Feb 2027

Sports Illustrated and fabricated authors

ScheduledTuesday
Failures9 Feb 2027

Detroit facial recognition wrongful arrests

ScheduledTuesday
Failures16 Feb 2027

The GPT-4o sycophancy rollback

ScheduledTuesday
Failures23 Feb 2027

Rite Aid and the FTC

ScheduledTuesday
The weekly

Six pieces of research a week. One email on Sunday.

An organisation mapped, a failure documented, a category defined, an obligation explained, a company running AI in production, and an investor mapped by position. Free, and the archive is public.