Skip to content

Topic

AI-Assisted Debugging Contributions

Open-source bug investigations conducted with an AI coding assistant under human direction, with disclosed workflow and human re-verification of measurements.

Current stories

buildOne report1 publisher

Django's icontains steered Postgres around a trigram index by wrapping the column in UPPER()

Django's icontains made Postgres filter UPPER(product_name) while a team's pg_trgm index covered the raw column, so product searches ran for several seconds. Having the indexes proved nothing until the team read the generated SQL beside the query plan.

Publishers:dev.to

Reality

Evidence58
Adoption
Not scored
Hype gap−5
Incentives
Not scored
Confidence60
buildConfirmed2 publishers

The AI told Torvalds his bug was unsolvable. Twenty-four patches later it was one character.

A model that quits is harder to audit than a model that lies, because a refusal leaves no artifact to check. The Xe driver fix shows the quitting verdict can be flatly wrong.

Perspective Coverage

3 publishers
Builder
Builder 62%
Operator
Operator 33%
Investor
Investor 5%

Reality

Evidence62
Adoption15
Hype gap+30
Incentives
Not scored
Confidence55
buildOne report1 publisher

A decorative checkmark crashes pytest on a runner that boots with LANG=C

The crash happens inside the reporter while it renders the failure message, so the message you need is the one that fails to print. Four conditions have to hold for it to reach your CI. A UTF-8 laptop hides all of them.

Publishers:dev.to

Reality

Evidence58
Adoption
Not scored
Hype gap−10
Incentives65
Confidence55
buildOne report1 publisher

A month of side-by-side use sorts Cursor, Copilot and Claude Code into three different jobs

One developer ran all three through implementation and debugging in an existing project and would not name a winner. The split tracks how much context each tool gathers before it writes, which is a claim about your repo as much as theirs.

Publishers:dev.to

Reality

Evidence32
Adoption
Not scored
Hype gap+12
Incentives22
Confidence45
buildOne report1 publisher

A 58% flake in a shared disk cache was a real race, found only when Linux-only CI met Windows

Seven failures in twelve runs of one Vesuvius Challenge test exposed three PermissionError surfaces in a chunk cache, and one of them under-evicts on every operating system.

Publishers:dev.to

Reality

Evidence58
Adoption22
Hype gap+12
Incentives55
Confidence52