Posts Series · 4 parts

The New OSS

The bottleneck in open source is moving from coding to diagnosis.

An investigation of what agentic execution does to open source, run against GitHub Agentic Workflows with public data and open methodology.

19 Bugs filed, fixed by agents
435 Community issues analyzed
158 Contributors in the dataset
0 Pull requests opened
The dispatch layer, drawn

Two paths through the same infrastructure. Bot-generated issues enter the queue automatically; community issues wait for one maintainer to promote them by hand.

Classification Eligibility Priority Bot-generated enters automatically Community 5.7% of issues ai-triage · automatic cookie label · one maintainer, by hand shared queue merged
#18980 promoted at the gate
71 minutes
#19017 waited at the gate
7.76 days — 157×

Same author. Same quality of diagnostic signals. Filed three hours apart. The eligibility gate decided.

01

The New OSS

The bottleneck in open source is moving from coding to diagnosis. This is what happens when it does.

19 bugs filed, 0 pull requests. Fixes implemented by agents, merged by a human.

5 min
02

The Agent Interface

What 435 issues reveal about writing work that agents can execute.

435 community issues from 158 contributors, methodology open-sourced.

7 min
03

The Dispatch Layer

Inside the routing architecture that decides which issues get fixed in minutes and which sit for days.

Issue #18980: 71 minutes. Issue #19017: 7.76 days. The 157× gap, located.

7 min
04

The Routing Problem

Five things that break when agentic triage systems scale beyond bot-generated workloads.

Five failure modes in the routing architecture, with structural fixes.

9 min