My networking agent has drafted 15 follow-ups this week. I've sent 2. Looks like the more I use agents the more I turn into my own bottleneck.

Ardales with makers I met in Colombia in 2022. Sebas is now at a Valencia startup building knowledge systems for AI agents. Allison and Jose David came too. Lake, picnic, skating in Malaga. The ones you meet building early keep showing up.

Your CRM can change its pricing. Your sequencer can sunset a feature. Your enrichment tool can change its API. Your AI chatbot can switch models overnight. Nothing wrong with any of them. That's just how software works. But remember: none of that can happen to context you own. That was the point of last week's Clay post. I'm not hating on Clay, just vouching that you're conscious on where your GTM foundations live: - Personas - Offers - Segments - Buying triggers - Value propositions - References - Competitors Etc. Here's the part I left out: owning your context is only half of it. The other half is keeping it current. A GTM second brain you own but never update is just a database of old assumptions. So for me it comes down to 3 parts: → Own the layer → Update it consciously, and often → Build systems that allow reliable updates (without slop) Tools come and go. The context you keep alive stays.

Just joined Social Hub cowork in Barcelona. Checked five or six places first. Social Hub never replied, so I followed up twice. Then they confirmed. Persistence works, even for a desk. First time settling after nomad life.

Ran Barcelona before sunrise. First half marathon is done. Training for the second one before the year ends. The city is empty at dawn. Just me and the sea. Consistency is the whole trick. New quarter, same discipline.

Trying out a new sports modality in Barcelona, I call it Vibe Skating

Apple season near Bern. Went for a run, then picked apples straight off the tree. Everyone talks about low-hanging fruit. Today I actually went and grabbed it.

My wife called me Globi once. So I bought a book while I was in Zurich about this Globi character. Turns out it's a Swiss kids cartoon, a parrot who is kind of loud, over-engineers everything, doesn't quite fit in, and it's kind of funny as well. And then it hit me, "Oh my God, Globby is not Swiss!" He cannot be Swiss. Globi must be Latino, definitely migrated from somewhere in South America. It's a blue parrot guys! Now I kinda relate with Globi🤣

You don't need one superpowered agent, you need agents that know their lane. A Chief of Staff that routes. Specialists that know their topics best and own their lanes of work. Treat your agents like you'd treat your team (meetings, managers, operators) and you'll start seeing great results. Here's a reference from Oxygen, Appsaavy's partner company 👇

Highly recommended for solopreneurs: have your chief of staff in Grokbot interview you ("grill me") to reflect on your last Q (strengths, achievements, weaknesses). then move on to next Q's planning. A 20-min conversation will end up in a direction, clarity and goals locked

Day 2 of setting up my Agentic Team to support my goals. I now have my Grokbot Chief of Staff as the orchestrator, and specialist agents. They are using Claude via CLI so it can access all my previous skills (and a better LLM)

While @clay was happy around, flexing their new shiny features, did the intern take over the engineering?! They suddenly removed the "webhook" setup out of their campaigns setup, one day to the other, fucking up everyone with active campaign response systems. GREAT WORK GUYS!👏 I swear these type of issues are happening EVERY month to real Clay users, no amount of tokens or investment will fix it.

The real question is where context lives. Inside one tool → works in that tool Owned + tool-agnostic → every agent reads it (Claude Code, n8n, Clay, Hermes) I'm betting on owned. xitter.cf/JGrandillon/status/210…

Met an agency founder last night. She said the bigger it gets, the more scared she is of scaling. Everyone says grow. Nobody asks why. Before "how do I scale?", ask "what shape of business do I want?" Agency → linear Product / AI-native service → different curve Company of one → valid answer What did you pick?

Gave grokbot access to my Claude Code for it's deep-work sessions.

Muse, Instinct and GrokBot didn't out-build open agent harnesses. They out-designed them. Most of what they do, a well-set-up Hermes or OpenClaw can already do. The difference is you don't need to be technical to start, and it lives where you already chat. Open-source agents still win on control and data ownership. But without a mobile app, a clean desktop UX and 5-minute onboarding, that won't matter to most users. The backend is commoditizing. Good experience is (and has always been) the moat.