@clairevo@davemorin@openclaw This is genuinely useful, and thanks Claire for doing it. But to be honest, I'll just have Claude Code running on the same machine and use that as my sysadmin.
It does a lot more than that. It does a first pass at my bookkeeping and researches and qualifies incoming meeting requests.
It's automated first-pass content creation, so I never miss the publishing schedule again, and it runs a plethora of checks on sites and services. If you set it up properly, it becomes genuinely useful.
My OpenClaw agent, James, works with his own team of specialist sub-agents he calls when the job demands it - operating more like a conductor than a soloist.
Sub-agent one, Ernest, is the writer. He knows my personal voice, follows strict writing guidance, including the Chicago Manual of Style, and handles all the work where tone, structure, and finish matter most.
One of the others, Sherlock, is the researcher. He has direct API access to tools like Brave Search, Perplexity and Exa (which he uses when relevant), and works through a highly disciplined research process with sourcing, citations, and links back to original material.
The advantages are: James, the core agent, no longer has to carry every context, instruction set, and skill in one set of instructions, which leads to bloat and forgetfulness.
Instead, the work is delegated to the right specialist who has the relevant deep knowledge, token use stays significantly leaner, and the output gets better.
All agents then write their output to an Obsidian vault that's synched via github to my MacBook so I have access to the output.
@icanvardar One of many things I use it for: listen for incoming https://t.co/VfW48fR34F requests. Research the person/company. Prepare a dossier using LinkedIn/Perplexity skills, post to Slack and open a new doc in Obsidian.
@JulianGoldieSEO Reason "nobody is talking about it", which is utter nonsense in itself, is that local models mostly suck unless you run them on $30k hardware.
@nextbrickx@mcuban That's incorrect. Just like with AI today, PCs were positioned as "efficiency gains", thinly disguised "you'll need fewer people". It never happened with PCs and I am fairly certain it'll never happen with AI either.
@thesamparr 1. We are all massive geeks and gravitate naturally towards this.
2. We enforce it - everybody commits to experimentation and we meet every Friday for show and tell.
3. It's how we earn money, so its in our best interest