rishabh here, i run a small ai automation agency. most of my builds now start the same way: before any agent touches a workflow, i write one short runbook that says what the agent owns, what it never touches, and what done looks like. what changed was drift. an agent i built for quote follow-ups slowly started rewording client emails in its own style. nothing broke. the output just stopped sounding like us, and nobody noticed for two weeks. now every build ships with that one-pager baked into the system prompt, and when output drifts i check the runbook first, not the code. nine times out of ten the agent was doing exactly what it was told. curious how you all handle this. do you lock agent behavior down with docs up front, or steer it after it ships?

Kit is an email platform built for creators. People sign up through a form, you tag them, and you send them emails, either one-off or on autopilot. 👉 https://kit.com How it works: 1. Build a form or a landing page. Anyone who fills it in becomes a subscriber on your list. 2. Tag subscribers by where they came from or what they asked for. Segments group people by those tags, so you can email one slice of the list. 3. Send a broadcast. That's a one-off email to everyone or to one segment, like a weekly newsletter. 4. Set up a sequence, a run of emails that goes out on its own after someone joins. A welcome series is the classic one. Visual Automations chain sequences and tags together. For builders, Kit has an API (v4). An API key covers automating your own account, and OAuth is there if you're building an app for other creators. Webhooks can ping your own system when something changes. So an n8n workflow or a Claude Code script can add people to your list from any form or page you build. On cost: the Free plan covers up to 10,000 subscribers with unlimited forms, landing pages and broadcasts, plus tagging. Sequences, Visual Automations, apps and integrations and the Kit MCP start on Creator, at $33 a month for 1,000 subscribers (17% less paid yearly). What would you send a list if you started one this week?

Worktrunk lets you run several Claude Code sessions on the same project at once without them overwriting each other's work. It has 11,269 stars on GitHub. https://github.com/max-sixty/worktrunk The trick underneath is a git feature called worktrees. Each agent gets its own copy of the project folder on its own branch, so one can build a feature while another fixes a bug. Plain git makes this painful: you type the branch name three times just to start one. Worktrunk cuts it to one command. How it works: 1. wt switch -c -x claude feat creates a branch called feat, makes its own folder and opens Claude Code inside it. 2. Repeat with a different name for each job. Every agent works in its own folder. 3. wt list shows every worktree with its changes and how far ahead of main it is. 4. wt switch feat jumps you into any of them. 5. wt remove deletes the folder and the branch when the job is done. Worth knowing next: hooks can run a command whenever a worktree is created or merged. It can write commit messages from the diff. Ten worktrees can share one node_modules folder instead of copying it. And wt switch pr:123 jumps straight to a pull request's branch. Install it on Mac or Linux with: brew install worktrunk && wt config shell install On Windows it installs through winget as git-wt. It's free and open source. What would you hand to a second Claude Code session if you could run two at once?

Mine, learned the hard way: I feed it a broken input on purpose and see whether it fails loudly or quietly invents an answer. The inventing is the dangerous part, because everything looks fine right up until a client finds it. The other thing that caught most of my silent failures: one log line per automation per day, so I can see at a glance what ran and what didn't. Low-tech, but it works. This group talks a lot about building fast, but I want to hear about the trust step. Those of you running automations for clients, what's your check before you hand it over?

rishabh here, i run a small AI automation agency. mornings used to disappear into reading: client emails, group threads, notifications across three inboxes. i'd end up busy and still miss the one message that actually needed me. so i wrote out what genuinely needs my judgment and handed the rest to an AI agent: it reads everything, drafts the routine replies in my voice, and sends me one short brief with only the decisions left. mornings went from three hours of reading to fifteen minutes of deciding, and nothing important slipped since. curious how you all handle the reading side. do you build pipelines for it too, or does the inbox still land on your desk?