rjhuantan

I built an AI agent that has reached 45,000 people — 7.5 million messages and 735,000 calls. Some things I learned along the way:

1. Reading the conversations is irreplaceable.

We anonymize the logs — reading user conversations feels a little dirty, and you want to be responsible about it — but it is so, so helpful. Evals and QA only catch so much. Real users will find a zillion things your testing never will, and reading how the agent actually handled them is how we decided what to fix, what to add, and what to leave alone.

2. Managing an agent is a lot like managing a person.

That framing sounds cute but it's genuinely how the work goes. When I wanted something done — reach back out to people who'd gone quiet, handle a certain kind of request differently, soften the tone — I'd give the agent direction, it would go do it, and I'd check its work in the logs. Some direction is...

Hey everyone,

I’m pretty new to AI and automation, and I'm really fascinated by AI agents. I understand the concept, but I’m struggling to visualize how people actually use them in their day-to-day jobs.

I'd love to hear about your real-world use cases. What kind of daily tasks or workflows have you successfully automated using AI? What tools are you using to make it happen, and how does it actually work from start to finish?

Just trying to learn from the community and get some inspiration for my first setup. Thanks!

Is there a trend that eventually, all Agents will converge in functionality? Getting started with ZCodex feels similar to codx.

User experience: What exactly is ZCode? It is an AI programming assistant launched by Zhipu AI. The newly released version 3.0 is deeply adapted to the GLM-5.2 model. Its positioning is not just another IDE plugin—it is a desktop AI programming Agent that can understand project structure, plan code architecture, write code, run tests, and submit PRs, all without human supervision. ZCode runs on macOS. MoonBit, with 16.9k stars on GitHub, follows a similar approach, but ZCode is more radical—it turns AI into a desktop operating system, with task management, file system, terminal, and browser all in one window.

Three Core Highlights

1. Model is its own GLM-5.2 Zhipu AI is one of the earliest companies to develop large models in China. The GLM series has...

Job searching was eating hours a day, mostly on copy-pasting the same info

into slightly different forms. So I built an agent that:

• finds open roles across job boards

• rewrites my resume for each job and generates a fresh PDF

• fills out the actual application (Ashby, Greenhouse — dropdowns, comboboxes, the "why do you want to work here" boxes)

• stops before submitting so I review every application — it's an assistant,

not a spam cannon

Built with Python + Playwright + an LLM doing the form-field mapping. The

hardest part by far was handling how differently every ATS renders its forms.

AMA about the build — curious whether people here would trust it to submit

autonomously (I don't, yet).

Hey All - I wanted to share my experience over the past few years of doing automation consulting part time (while still having a 9-5). To start, I come from a cloud / platform engineering and development background working with Fortune 500 financial services companies, so when I saw the multiplier that things like Claude Code (when it first came out) paired with true best practices and principles could do, I was totally hooked.

To be clear, everything I'm about to go over is just my experience and what I focus on.

I started out like a lot of folks do targeting small businesses trying to help them automate their processes and procedures. It was painful. Most only loosely knew what automation was and what automations were all about. This was around 2023 so it was still earlier days for things like ChatGPT, etc. but I found that for smaller businesses they either:

didn't have the money t...