Hopper

Software engineering assistant built in 2017 — before GPT, before Copilot, before everyone was doing this.

What It Did

• Issue Duplicate Detector - Match new issues with existing ones to avoid duplicate work
• Issue Labeler - Predict whether an issue is a Bug, Feature, or Chore
• Commit Generator - Generate commit messages from code diffs
• Standup Summarizer - Summarize daily standups automatically

The Technical Approach

This was 2017. No transformers yet. We used:

• Seq2Seq architectures
• Neural Machine Translation models
• Attention mechanisms (Bahdanau attention)

The same problems Copilot solves now, but with 2017 tools.

Demos

Why This Matters Now

I was building AI coding assistants before it was cool. The core problems haven’t changed - understanding code, generating code, helping developers. The models got better, but the problems I was solving in 2017 are the same ones Claude Code and Copilot solve today.