Comment by bhawks | Hacker News Reader

Comment by bhawks | Hacker News Reader

Parent Full thread

The system's external name is IssueTracker[1]. Internally it's name is buganizer. Buganizer has been around forever (or it already had a ton of bugs logged when I started in 08).

It gets (ab)used to store tons of semi structured data with a pretty good siloing and acl capability (including allowing Google 3rd party partners to work on issues). This data includes bugs but also tasks, workflows, approval processes etc. Buganizer is very api-first so many different tool frontends use it as their backend.

Generally every huge leap in bug numbers was associated with buganizer assimilating another tracker's bug database (from acquisitions and similar). For what it's worth no issue tracking system I am aware of has avoided buganizers huge gravity well. I expect this trend to continue.

1: https://issuetracker.google.com

> For what it's worth no issue tracking system I am aware of has avoided buganizers huge gravity well

Even some small, well managed, projects will develop hundreds upon hundreds of issues in their trackers. Can't imagine how you'd deal with that at such a massive scale.

AI assisted prompts to users to provide more structured context[1], de-duplicate, cluster, prioritise. All things people _should_ be doing but can't be bothered to or aren't incentivised to.

Who here hasn't been in a standup where a task was created against a CL that only had a title "FIX" (with the intention of filling it out later)

Complex Project Management in other sectors (construction) is starting to see life in this regard. Unfortunately much of the digital transformation of PM to date has focused on capture and reporting. Capture is capture but the problem with reporting is that nobody read the reports before and aren't going to read them now they are in dashboard with endless possibilities to slice and dice.

Automating insight is next but insight is difficult to draw when you have tens to hundreds of people building an often unreliable, usually bespoke, and piecemeal semantic model over time; quality and context is scattershot and the datasets not large enough on their own and not comparable to others.

[1] e.g. Clippy but AI that gets better as teams provide feedback: Does this issue relate to x or y? Is this issue A more important than this issue B? Was this issue A fixed by this PR B?

Reply on news.ycombinator.com