Stop Using Git. Stop Writing Tickets.
Let AI do them for you—and do them better.
Robert Matsuoka May 04, 2025
Let me tell you what changed. Everything about my development workflow just became exponentially more efficient.
I hit a bug with the Gemini API: the tool was throwing 404s because it tried to use a model that didn't exist. Pretty standard debugging scenario. But instead of writing up the ticket, switching tabs, and babysitting the Git flow, I told Augment what went wrong.
It handled the rest:
Filed a clear, labeled GitHub issue
Proposed fix strategies
Created a bugfix/ branch
Made granular, well-scoped commits
Wrote thoughtful commit messages
Tagged the issue and opened a clean PR
It didn't just remove overhead. It produced artifacts I'd be proud to ship.
Old habits die hard—but they should
Back during COVID, I worked with a team of top engineers at Tripadvisor to build a social distancing app called Crowdfree. We all volunteered our time. It was a great project—fast-moving, high-impact—and I filed every ticket manually. Wrote up every bug. Managed every commit. It worked, but it was inconsistent. Some issues had context. Some didn't. Most were rushed. I spent hours each week just on organizing issues that should have been spent solving them.
That was the best we could do back then.
Now?
Augment writes better tickets and manages Git better than I ever did.
It catches edge cases I forget. It names things clearly. It preserves intent—not just code.
AI as your Git partner
The unsung hero here is what Augment does between commits:
It scopes changes intentionally
It writes accurate, non-generic commit messages
It links every commit to the issue it's solving
It manages branch hygiene without a prompt
It even tags releases when a fix gets merged
Git hygiene used to be a tax.
Now it's a strength—and I don't touch it.
Let's be honest:
You don't need help writing code. You need help managing the mess around it.
This insight changes everything about how we should approach AI for development. The low-hanging fruit isn't in auto-generating functions—it's in eliminating the administrative overhead that drains our focus.
This isn't automation. It's orchestration.
I didn't script the GitHub API. I didn't set up a bot net.
I just taught Augment what "file an issue" and "manage a Git flow" mean—and it ran with it.
Now when I say: "There's a bug with Gemini. File it." "Start work on #8." "Propose a fix and open a PR."
It happens correctly. Every time.
Quick aside:
In the case of Augment, I still click "Run" to approve its GitHub issue actions. That's minor—and I actually prefer having a review loop before issues go live.
On the Git side?
It's fully autonomous—commits, branches, messages, and all—based on a pattern I trained it on. No prompts. No approvals.
Cleaner workflows, higher quality
Here's what the difference looks like in practice. A typical developer ticket might read:
"404 error on Gemini API calls."
While Augment generates:
"BUG: Gemini API throws 404 when requesting non-existent model variant 'gemini-1.5-pro-exp'. Affects: API client module, image generation feature. Reproduction steps included. Priority: High."
This transforms:
Issues with real structure
Branches named with purpose
Commits you'd actually want in a changelog
PRs that map cleanly to work
This isn't about speed.
It's about consistency, clarity, and leverage.
You stay focused on the architecture and the edge cases.
Your agent handles the rest—with aplomb.
One last thing
Even if you're hesitant about having your AI do some coding—and you shouldn't be—having it handle the details of your workflow will boost your team's productivity by at least 25%.
No new tools. No process revamps.
Just a fundamentally better way to ship quality code.
Worth noting: I'm also training Claude Code to handle the same workflows, and it's working without issue. The approach isn't tied to any specific tool—it's about delegating the right tasks to AI regardless of your preferred assistant.
The most valuable AI isn't the one writing your algorithms—it's the one eliminating the cognitive friction around them.
I used to write good tickets and clean commits. Now I don't write them at all.
And somehow they've never been better.
The bottom line
We're entering an era where development workflows can be delegated without compromise. The tools like Augment are mature enough that they don't just match human quality—they consistently exceed it. For teams that want immediate AI benefits without overhauling their tech stack, workflow management is the perfect entry point: low risk, high leverage, and surprisingly straightforward to implement.