6 AI agents are working for me right now. In parallel. Like 6 direct reports — but faster, more patient, and available at 7:15 pm on a Thursday. That screenshot is my monitor tonight. Six Claude Code windows running simultaneously — reformatting a 30-page document, deploying a "kiddo points" web app for my kids, analyzing my health data, crunching LinkedIn analytics, cataloguing every film I've watched, and researching a VC firm's team. To bring the last one to life, in <60 minutes of clock time and <10 minutes of my actual supervisory time, Claude Code helped me to do some in-depth research on the 60+ person investing team at Bessemer Venture Partners: First, it pulled all the bios (easy for humans, but would take >60 minutes of mindless copy-and-pasting) Then, it created the So Whats of their team composition across educational backgrounds, prior employers, and uncovering Quirks and Hidden Patterns: https://lnkd.in/dJjqKqQh I’ll bet even folks who work at BVP - hello, Shannon Brayton, Lauri J. Moore, Steve Kraus, Kent Bennett, and more - will learn new nuggets from the Quirks section at the bottom of the Notion doc (BTW, the Claude Code-Notion integration is incredible) This "So Whats" analysis would have taken me 5+ hours to replicate, and, realistically, the quality of my work would have been far lower, and I would never have invested the time to begin with. And that was the output from just one of six windows running at the same time. 6 agents in parallel probably sounds crazy to many of you. But with the current pace of AI innovation from Claude Code/Cowork and others, I'm confident it'll be commonplace for everyone reading this by June. And 6 parallel windows will soon become 60. And then 600... The bottleneck isn’t technical skills anymore. It’s the ability to generate ideas worth building and to use ever-improving AI tools. If you're reading everything about Claude Code — as of today, with Opus 4.6 now available! — but not yet using it, you're missing out on the single most powerful product ever built. I've never entered a Computer Science building and can't write a line of code on my own. Now I ship new analysis and new products every single day, with zero other humans involved. #JFDI
Did you close the bvp deal?
How are you Supervising these agents?
This is the real thing. Embrace and run 10x faster. The only thing I miss in Claude is pro-active triggers. I built a personal PA, would love to have it send me a message each morning on whats important, like a PA cleaning the time table. There is a list of similar tasks. Did you come across these and if so, how did you work through it?
James Raybould I I just went through a prolonged sprint for my startup, and so I used Claude Code to build myself a health concierge that looks at my genetics, blood labs, food intake, daily schedule, and built out a custom supplement stack and schedule to work to optimize at least one good night of sleep a week to blunt the deleterious effects. It's incredible how powerful and easy it makes building custom software now.
I was always available at 7:15pm on a Thursday!
LOVE this, James! Do you turn off agent desktop notifications so you can decide when to come back to your Claudes on your own time/ limit context switching?
Very cool .... your posts are always inspiring .... out of curiosity, why BVP James Raybould PS: Crunching your health data is something I and likely many others want to do ... would you please consider a separate post on that
Claude code is truly amazing! But how do you deal with all of the context switching with 6 different instances? It's been a real challenge for me to work on even two deep things with CC at a time...
None of those are coding tasks, so I'm curious: Why are you running them in Claude Code?