Skip to main content
The Hard Parts.dev
FM-13 ai FM Failure Modes
Severity high Freq increasing

Synthetic Velocity

Output volume rises sharply while true understanding, maintainability, and durable progress do not.

Severity
high
Frequency
increasing
trend
Lifecycle
build · delivery
Recovery
medium
Confidence
high
At a glanceFM-13
Also known as

output theateractivity inflationbusy but not movingthroughput without progress

First noticed by

staff engineerengineering managerai engineer

Mistaken for
10x productivity
Often mistaken as
modern engineering leverage

Why it looks healthy

Concrete external tells that make the pattern read as responsible behavior.

  • PR count, ticket count, and lines of code climb noticeably
  • Engineers report feeling fast and productive
  • Leadership sees visible tool adoption and modern tooling
  • Docs and code comments are more plentiful than before

Definition

What it is

Blast radius code team delivery operations

Teams produce more code, tickets, documents, and visible activity - often assisted by AI - without the expected improvement in outcomes.

How it unfolds

The arc of the pattern

  1. Starts

    New tooling makes many kinds of work faster to produce.

  2. Feels reasonable because

    The increase in visible activity is real, and some of it is genuinely useful.

  3. Escalates

    Review becomes shallower than generation. Ownership weakens. Rework and confusion rise slowly rather than dramatically.

  4. Ends

    The organization confuses artifact volume with actual capability and pays later in defects, drift, and cognitive load.

Recognition

Warning signs by stage

Observable signals as the pattern progresses.

EARLY

Early

  • PR volume spikes faster than review capacity.
  • Generated docs and code proliferate quickly.
  • People cite speed more often than understanding.

MID

Mid

  • Rework rises quietly.
  • Code authorship becomes blurry.
  • The team closes more work while feeling less certain.

LATE

Late

  • Incident rate, regression rate, or maintenance burden climbs.
  • Nobody fully owns some generated areas.
  • The system looks productive but feels brittle.

Root causes

Why it happens

  • Organizations overvalue visible output
  • Review and validation do not scale with generation
  • Productivity claims arrive before operational evidence
  • AI lowers production cost faster than decision cost

Response

What to do

Immediate triage first, then structural fixes.

First move

Pick one AI-heavy area and measure a durable outcome (rework rate, incident rate, rollback rate) over the same window as the volume gain.

Hard trade-off

Accept that visible output will slow once review, ownership, and testing are held to the same bar as the generation rate.

Recovery trap

Adding more AI tools to review the output of the AI tools.

Immediate actions

  • Measure rework, incidents, rollback rate, and review burden
  • Require clear ownership for generated artifacts
  • Slow down in high-risk areas

Structural fixes

  • Treat AI-assisted generation as input, not completion
  • Upgrade review practices for generated code and docs
  • Measure durable outcomes instead of raw output

What not to do

  • Do not ban useful tools reactively
  • Do not keep celebrating volume while ignoring maintenance cost

AI impact

How AI distorts this pattern

Where AI-assisted workflows accelerate, hide, or help with this failure mode.

AI can help with

  • AI can still reduce toil and accelerate well-bounded implementation if paired with strong review, ownership, and testing.

AI can make worse by

  • This failure mode is itself AI-amplified: generation becomes cheap, so weak ideas and half-understood changes travel further and faster.

Relationships

Connected patterns

Causal flows inside Failure Modes, and related entries across the site.

Easy to confuse with

Nearby patterns and how this one differs.

  • Adjacent concept Genuine productivity gains

    Real gains show up in outcomes - fewer incidents, faster learning loops, more durable changes. Synthetic velocity shows up only in output volume.

  • Ticket theater performs work that was never there. Synthetic velocity produces real artifacts - but of shallower quality than their volume suggests.

  • Autocomplete architecture is about the shape of generated systems. Synthetic velocity is about the pace and volume of generated work, whatever its shape.

Heard in the wild

What it sounds like

The phrase that signals the pattern is about to start, and who tends to say it.

Heard in the wild

We're shipping way faster now.

Said byengineering manager or team lead

Notes from practice

What experienced people notice

Annotations from engineers who have worked this pattern before.

Best momentWhen intervention actually changes the trajectory.
When output accelerates before quality controls mature
Counter moveThe specific action that breaks the pattern.
Ask what got easier to produce, and what got easier to trust.
False positiveWhen this pattern is actually the correct call.
Real productivity gains exist. Synthetic velocity begins when output and understanding separate.