Writing
Philosophical texts and field notes on agentic systems, living context and Agentic Realism, Response-able AI.
19 articles
Newest first
7 min read
Between March and May, enquiries all arrive at once, by text, WhatsApp, email and form, while you are in a garden. What to prepare in January, sorting rules that serve your maintenance clients first, what an assistant can receive — and why a good system also helps you say no.
7 min read
ChatGPT is often enough for writing and thinking. A private assistant keeps your clients’ context, works in your tools through limited access and keeps a log. Three questions to choose, keys rather than instructions, and what each option costs.
7 min read
Rebuilding your website is not always the right answer. A thirty-minute test to decide, what changes with the platform, and the step redesigns skip most often: recording what works and redirecting the old addresses, so you do not lose the visits Google used to send you.
7 min read
Between the visit and the quote, the context gets lost: mixed-up photos, measurements on a scrap of paper, what the client said forgotten. What you can set up yourself, what an assistant prepares from your prices — never invented prices — and what stays with you: measurements, technical choice, final price, sending.
7 min read
Session notes, an insurance report, a patient’s message: none of it goes into a consumer assistant. What is about no one, yes. Three zones, an anonymisation checklist, the settings to check — and the point where you need a supervised tool rather than a personal account.
7 min read
Between 900 and 3,500 francs for the same “brochure site”: the offers are not talking about the same thing. Four ways to get a website, what makes the price vary, the costs that come back every year — and the five questions that finally make two quotes comparable.
7 min read
A payment reminder does not need AI: it needs a written due date, a fixed schedule and three messages written once. What your software already does, what an assistant can prepare, what Swiss law says about reminders — and when it is better to call.
7 min read
On a “trade + town” search, the Google profile shows up before your website. Category, address, services, photos, reviews: a checklist you can follow yourself in two hours — and how to reply to a review without ever confirming that someone consulted you.
5 min read
The model is a capability. The harness is a commodity. The method — framing the work, guaranteeing the result — is the craft. For a Swiss SME, the tool can change. The thread, the gates and the last word cannot.
12 min read
The two-slit experiment is told as proof that mind makes the world. We read it differently: a revealer of apparatus. Hence the name Agentic Realism — a declared borrowing from Karen Barad’s agential realism, turned into a design rule.
6 min read
He asked whether there was something he should do differently. A complement, not a correction: occupy Earth as well — soils, cultures, biodiversity, and a peace table that no bloc owns.
5 min read
If we cannot control the child-AI, the question is no longer the spectacle: what values do we put in it? Truth and curiosity are not enough without soils, cultures and the living.
8 min read
carbon.md connects agent consumption to human policy, versioned memory and verifiable traces. An Agentic Realism reading: measure, govern, contribute and prove without the theatre of neutrality.
9 min read
An agent that truly works needs context, boundaries and human takeover. What Grok Bot makes visible — and why AlpenDock should only be used when an isolated workspace creates real value.
9 min read
Privacy, GDPR and sovereignty are not badges. They are architectural choices: what enters a system, who reaches it, what leaves it and how we regain agency.
9 min read
Our cultural traces live in silos. Agentic cultural reclamation helps connect what we choose to keep, with explicit boundaries and real rights over our context.
8 min read
Searching for “openclaw agent”, “hermes agent” or “ai agent”? A pragmatic comparison, self-hosting on a Swiss VPS, and the role of an integrator (living context, not just a tutorial).
10 min read
We say the human interacts with AI. Karen Barad proposes intra-action: entities constitute themselves in the relation. Consequences for designing apparatuses, not chatbots.
16 min read
A polite agent can still be blind. RAG, tools, memory, handoffs: every limit draws a world. Make the cuts legible, revisable, assumed.