AI-Assisted Development

Pulse is built with AI tools, extensively and deliberately. This page exists so there is no ambiguity about that, and so nobody has to guess.

The short version

I am the sole maintainer of Pulse. I use AI assistance throughout the project. Code, tests, documentation, release notes, and issue replies may all be produced with AI tools in the loop. Everything ships under my direction, my review, and my responsibility. If something is wrong, it is my bug, regardless of what typed the first draft.

Why

Pulse monitors Proxmox, PBS, Docker, Kubernetes, TrueNAS, and vSphere. Around that sits a web UI, a mobile app, host agents, installers, release tooling, security processes, and a support queue. That surface area is well beyond what one person can sustain by hand at any reasonable quality bar.

AI assistance is the reason a solo project can cover that breadth, ship frequent releases, and usually turn a reproducible bug report into a fix in days rather than months. Refusing these tools would not make Pulse better. It would make it slower and buggier. Used responsibly, these tools help me sustain the project's scope and quality. They accelerate the work without removing the need for judgment, verification, or accountability.

Where it is used

Declaring the general position is easy. Declaring where it applies is more useful, so here is the breakdown.

• Code and tests. Much of both is written with AI coding tools, working from my direction and landing through the same review, test, and audit gates as anything else.
• Documentation and release notes. Drafted with AI assistance and checked against the actual code and behaviour before publishing.
• Issue triage and support. AI helps investigate reports, reproduce bugs, and draft replies. On GitHub, a response from an automated triage run rather than from me at the keyboard posts under the dedicated pulse-triage bot identity and links back to this page. Support email works the same way: a reply may be handled end to end by automation, in which case it is sent as Pulse Triage from the normal support address rather than as me, and links back here. Automated replies never present themselves as me personally. I would rather you know you are reading automation than wonder whether you are.

What stays human

The tools write a lot of the code. They do not decide what Pulse is. Product direction, architecture, what gets built and what gets cut, review of what ships, and every release decision are mine. When you report a bug, the investigation may be automated, but what the fix should be and whether it is good enough to ship is my call, and I own the outcome of every thread either way.

What it does not change

Pulse is judged the same way all software should be judged, by what it does. The test suites, the audit gates that run before anything lands, the release candidate process, and the issue tracker history are all public. If you find a bug, report it with steps to reproduce and it will get fixed. Specific concerns about correctness, security, reliability, or maintainability are welcome and will be investigated on their merits.

From here on

This document sets the standard from the day it landed, not a claim about the past. I used AI for a long time before writing any of this down, and some of that use was not declared, including automated issue replies that carried nothing to say they were automated. I am not going to rewrite history or pretend otherwise. What I can do is state the rules now and hold to them going forward.

If you object on principle

That is your call, and I am not going to argue anyone out of it. This document is not intended to persuade everyone, but to make Pulse's development process clear enough for users to make an informed decision.

AI assistance is increasingly common in software development, but disclosure is inconsistent. I would rather be explicit about how it is used here.

Judge Pulse on what it does on your infrastructure and on the track record in the issue tracker. Those are the things that matter, and they are the things I stand behind.

For what Pulse itself does with AI as a product (Pulse Patrol, Assistant, and MCP, all optional and off by default until you configure a provider), see AI.md.