Activity
Mon
Wed
Fri
Sun
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
What is this?
Less
More

Memberships

AI Accelerator

19.8k members • Free

Lead Gen Secrets

23.9k members • Free

AI Automation Agency Hub

329.2k members • Free

AI Automations by Jack

3.3k members • $87/month

AI Automation Society

426.3k members • Free

The CEO Clubhouse

326 members • Free

Le plan V

64 members • Free

Digital Income Collective

868 members • Free

AI GameDev Lab

272 members • $49/month

4 contributions to AI Automation Society
Nobody warns you that a broken automation doesn't tell you it's broken
I've been running a handful of small agents for my own business every day for a few months now (a morning brief, some inbox triage, a couple of scanners). Not selling them to anyone, just my own ops. Sharing the one thing that actually bit me, in case it saves someone here the same surprise. When an automation breaks, it almost never breaks loudly. It just goes quiet. No error, no alert, nothing red. One of mine stopped running for something like two weeks and I had no clue, because "looks fine" and "is actually running" are the exact same thing when the thing is invisible. I only caught it because I went looking for something it was supposed to have produced, and it just wasn't there. The fix was almost embarrassingly simple: I made every run end by messaging me its result. Even the boring "nothing to report" runs, especially those. Because a silent agent and a dead agent look identical from the outside. The only way to tell them apart is if the healthy one keeps saying "still here, here's what I did today." The all-clear message IS the heartbeat. Honestly this has got me thinking way more about "how do I know this is still working" than "what's the next cool thing I can build." Feels like that's the boring little line between a demo and something you actually trust to run without watching it. Curious if anyone else here has been burned by a silent failure, or if you've found a cleaner way to keep tabs on background stuff than just emailing yourself constantly 😅 (French-Canadian here so go easy on my English 🙂)
Creating a website page to add to my existing site
I'm watching Nate's video on building beautiful websites and have a few questions. I've already got my own website / blog on wordpress, with a drag and drop theme builder. I want to eventually get rid of the theme builder and just use Claude Code. 1. can i take the html code from Claude and copy/paste that into WordPress? Or should I follow the steps Nate says in the video and go GitHub --> Vercel? What's the difference in security? I already have my site hosted somewhere else, so Vercel becomes the new cloud host, correct? 2. can i add integrations to push emails to my CRM? I've got Active Campaign as my email system...I'm assuming I can build in form integrations with Claude Code? What else should I watch to understand SEO, security, etc for using Claude Code for my website going forward?
1 like • 13d
@Jen Bessire You did your homework — Astro is exactly what I'd point you to for your case 👍 I build with Next.js myself, but for a content-heavy site like yours (157 posts!) Astro is honestly the cleaner fit: it's made for fast content sites, and your blog becomes simple markdown files. Good news about those 157 posts: you don't copy-paste them. WordPress has a built-in export (Tools → Export), and that export can be converted to markdown — exactly the kind of job to hand Claude Code ("here's my WordPress export, convert it to Astro content"). It's an afternoon, not a month. Forms, landing pages, sales pages: Claude builds those as pages in the same project. And your checkout embed works fine — a static site will happily hold any HTML embed code. The one thing I'd underline: don't flip everything at once. Build the new site on a subdomain first (new.yoursite.com), migrate the posts, click around until you trust it — THEN point your main domain at it and retire WordPress. Zero downtime, nothing burned. FTP to your current host works if it serves plain files, but GitHub→Vercel gives you the "edit → auto-deploys" loop, which is really nice once Claude is your webmaster 😄
1 like • 10d
@Jen Bessire Astro's whole claim to fame is that it ships almost ZERO JavaScript to your visitors by default. Your pages go out as plain HTML and CSS, and JavaScript only loads on the specific pieces you deliberately make interactive (a form, a cart button). That's why the big names use it — pages load fast, Google loves it, and there's barely anything for attackers to poke at. The JavaScript you probably read about (Node.js) is just the build tool — it runs on YOUR computer (or GitHub's servers) when the site gets generated, never on the live site your visitors touch. So no concern at all — honestly it's one of the main reasons Astro fits your case. Less code running in the browser than WordPress + a theme builder by a mile 👍
AIOS
Hey guys, any ideas on how to go about learning Claude Code? Im new, would love some tips!
3 likes • 13d
Everyone's right about the 7-day challenge — do that first 👍 I'll add the 3 things I wish someone had told me on day one: 1. Learn by building something YOU actually need (a small tool, your own site, a daily brief). Tutorials don't stick — real problems do. 2. Ask it to make a plan before it writes code, and to verify its own work after (run it, test it, look at the result). Night-and-day quality difference. 3. Put your project's rules in a CLAUDE.md file at the root — Claude reads it every session. That file getting smarter every week is where the real compounding happens. Treat it like a very fast junior dev whose work you review, not a magic button, and you'll level up quick 🙌
Anyone running scheduled/cron-based agents with Claude? What's your stack?
I'm looking into setting up an agent that runs on a schedule (cron-style) using Claude, and wanted to hear what others are using in practice. A few things I'm curious about: - Are you using Claude Code, the API directly, or something else to trigger the runs? - What are you using for scheduling (cron on a VPS, GitHub Actions, a serverless function, some orchestration tool)? - How are you handling state/memory between runs? - Any gotchas with rate limits, costs, or reliability worth knowing before I set this up? Would love to hear what's working (or not working) for people. Thanks!
1 like • 14d
I run a handful of these every day for my own business ops (a morning brief, inbox triage, a couple of scanners), so answering from the trenches 😅 One thing nobody's mentioned yet: Claude Code has native scheduled tasks now — you create the routine right inside it (cron-style) and it fires in its own background session. No VPS, no GitHub Actions wiring. Since you're already in Cowork, that's probably your shortest path to v1. The state-file pattern Jay described is the right one — I'd just add: split the memory in two. An append-only log (dated one-liners, newest first) + a "working state" file each run overwrites. The log is history, the state file is the live snapshot. Prune the log now and then or it slowly eats your context. And the reliability lesson I learned the hard way: make every run END by emailing you its result — even the "nothing happened" runs. A scheduled agent that only reports on problems is indistinguishable from one that quietly died three weeks ago. The all-clear email IS the heartbeat. On costs/safety for fully autonomous: cap what one run is allowed to do (mine are read+draft only — anything that sends or spends stays approval-gated) and space heavy routines apart instead of stacking them all at 7am. Hope that helps 🙌
1-4 of 4
@daniel-germain-8705
From Québec/Canada

Active 40m ago
Joined Mar 29, 2026
Québec/Canada
Powered by