Hi All! It’s great to be here. I’d love to hear about what your goals and greatest challenges are? I’m keen to connect and meet with other likeminded people.

Built a Daily AI Industry Digest that runs every morning at 9:00 AM BST on its own. No prompting, no babysitting. It pulls and summarises AI news across categories like Models & Research, Products & Launches, Funding, Policy and UK Events, sourcing from places like Fortune, Axios, TechCrunch, CNBC and Bloomberg. Opened my laptop and the digest was already sitting there waiting. Took a second to sink in! Scheduled task or loop? Actually used both. The digest itself runs on a schedule, but inside the project I used a loop to continuously check for runtime errors and logs while it runs. That way if something breaks mid-run, the agent catches it rather than silently failing. Much better than finding out after the fact that nothing got scraped! What surprised me? Seeing it had already completed before I even started my day. There is something different about watching an agent handle things quietly in the background compared to manually running prompts yourself. It has been firing every morning since 19 May without a single miss. Going from "I run the tool" to "the tool runs itself" is where automation starts to feel real. #AISChallenge

https://test-website-eight-gray.vercel.app I did use all of the hacks really. Yeah, that is a lot diffrent. Did not really take long time either. Everyting did make sence and its way more fun then I thought this would be

The internet wants you to believe that n8n is dead and everyone has moved on to pure Claude Code or custom scripts. But real production environments tell a different story. I recently launched my LinkedIn Automation SaaS (https://inboundy.app/). When planning the architecture, I chose to build the entire backend infrastructure using n8n workflows. Why? Because of visibility. When you scale outreach automation, you don't want to blindly hope an LLM handles the orchestration perfectly. When something breaks, I don’t want to jump through endless nested functions to find the bug. In n8n, I see it instantly. I can trace the flow, fix the node, and keep moving. How it started: I originally built this custom n8n setup just to market my GLB-Optimizer. It worked crazy well for client acquisition. It handled keyword-based connections and hyper-personalized the messages by matching the recipient's profile summary with my own context so it actually felt human. It brought in so many clients that I decided to pack it into a clean, cloud-based SaaS platform. Building in the AI space right now isn’t about chasing the latest hype tool. It’s about choosing the right architecture for the job. For us, n8n is still a production beast. Curious to hear from the community: Who else is still running their core SaaS backends or heavy agency workflows on n8n? What’s your go-to stack right now?

I’ve been using RIGGS — a Claude-backed agent running in my terminal — as an orchestration layer, not just a coding assistant. For Phase 2 of the lab, it: provisioned a Windows VM over SSH installed Sysmon installed and configured the Splunk Universal Forwarder shipped telemetry to Splunk on my home server installed Atomic Red Team ran T1003.001 (LSASS credential dump via ProcDump) My role: physical access + final decisions. The most interesting part was watching it hit real-world blockers and work through them. Defender blocked the attack — and also blocked the usual paths used to disable it. RIGGS surfaced each blocker, explained why it failed, proposed the remediation, and executed when approved: execution policy Defender exclusions Tamper Protection LSASS PPL SSH filtered admin token / missing SeDebugPrivilege Once those were cleared, ProcDump dumped LSASS in 1.7 seconds. Then RIGGS checked Splunk and confirmed 19 events. Full kill chain visible. The split was clean: AI handles execution and verification. Human handles approvals, destructive actions, and judgment. That feels like the real opportunity with AI orchestration. Not “AI replaces the operator.” More like the operator stops doing the low-leverage work. Anyone else running AI this way — as an ops layer with human-in-the-loop checkpoints?