Douglas Campbell

Handing off the work doesn't mean handing off responsibility. The agents did the implementation. I kept the copy, the taste and the calls. Full write-up with screenshots: projectfresh.com/blog/design…

The final check mattered most. The agents read all 3 sites side by side and flagged drift: a product "launched" on one site, a "pilot" on another. Podcast numbers that didn't match. Even my own title. I decided each once. They fixed it everywhere.

Show it, don't describe it. On RAIZOR that meant live agent-and-human workflow demos, a department explorer with before/after examples, and a clickable five-stage build diagram. All three sites have light and dark themes.

Real access, with approval gates. The agents managed DNS, mailboxes, files and backups. We moved two domains off WordPress with a way back at every step. Every live switch waited for my yes. We never needed to roll back, but we could have.

Shared memory. Every agent session ended with a short note: what got done, what I decided, what's still open. The next agent read it before starting. So agents could work in parallel, on my Mac and my server, without me re-explaining decisions.

I redesigned and launched 3 websites in 7 days of part-time work, almost all from a chat with Claude. Done the old way with an agency, I'd guess a month or more per site. Here's what made it work:

The Weekly Deep Dive: Can autonomous systems run your local corner store while new digital tools handle your office chores without making costly mistakes? #Automation #Robotics #CyberSecurity #TechTrends raizor.podbean.com/e/the-wee…

The Weekly Deep Dive: Why are major companies quietly rehiring the workers they automated? The true cost of replacing human labor might surprise you. #Automation #FutureOfWork #TechTrends #WorkforceStrategy raizor.podbean.com/e/the-wee…

The Weekly Deep Dive: Can you trust automated digital assistants to run your business when software begins making financial decisions without your direct supervision? #Robotics #MachineLearning #TechNews #FutureOfWork raizor.podbean.com/e/the-wee…

The Weekly Deep Dive: What happens when autonomous software agents make irreversible financial decisions alone? Smart teams must keep human judgment in the loop. #Automation #MachineLearning #TechTrends #FutureOfWork raizor.podbean.com/e/the-wee…

Full write-up, including the safeguards I use: projectfresh.com/multi-agent… How are you keeping context in sync across your AI tools?

My takeaway: shared AI agents with narrow tools and a clear job and permissions (such as an SEO researcher or blog drafter) are the safe sweet spot today. Connecting human teams's personal agents should be used with great care until they can strictly enforce owner-only rules.

These permissions can live both on the communication bridge, as well as on the personal agents themselves and require good research and testing.

#1 worked great and is immediately useful. #2's biggest lesson was about risk, not capability. Because you could add both personal agents and team agents to a shared chat, there needs to be serious consideration to the permissions you give them.

2. A shared chat. Using a tool called Bolter, I asked Claude to find where I'd worked on a project, and then instruct my OpenClaw agent restyle a homepage draft with brand colors pulled from the brain. About five minutes later it replied that the page was live.

1. A shared brain. My own version of Garry Tan's gBrain, on my own server. Each agent sends summaries in through one MCP connection, and any of them can search it and pick up where another left off.

Every new session on a new platform meant re-explaining the project, or hunting for where I built something. So I ran two experiments.

I spent most of my weekend getting my cross-platform personal AI agents to share a memory and talk to each other. OK, and played some pickelball too! The problem: my work is scattered across Claude, Codex, Grok Bot, OpenClaw and others.

The Weekly Deep Dive: Who is liable when an automated agent makes a wrong call? We look at the shift toward machine autonomy. #Automation #Robotics #CyberSecurity #TechTrends raizor.podbean.com/e/the-wee…