trip2g: marketing brief and content plan

What trip2g is (one line)

An open-source, self-hosted way to turn your Markdown/Obsidian notes into a federated knowledge mesh that AI agents can query across trust tiers (personal, friends, company, public) without uploading your files to anyone's cloud.

The name is trip to graph: every base is a graph, the bases connect into one larger graph, and agents walk it.

Two products, two go-to-markets (read this first)

This pack is the trip2g.com / OSS-core GTM. There is a second product on top: the sovereign B2B box (the simplepanel control-plane + agent fleet + federation core, sold as a self-hosted appliance). It has a different buyer, a different beachhead, and its own materials. Do not mix them.

• trip2g.com (this pack): OSS federated knowledge mesh for individuals and small circles. Beachhead = peer circles. Free / self-host. Lives here.
• The B2B box (separate): a self-hosted appliance for an operator who runs an agent fleet over context they own and needs cross-org sharing no SaaS will do. Beachhead = the self-hosted-agent shop outgrowing n8n/Dify (entered via fleet-ops), expanded into cross-org federation. Materials + research live in trip2g_simplepanel/docs/ (market_entry_analysis.md, product_truth_b2b.md, cis_market_research.md, and marketing/ there).

Three facts from the B2B research that strengthen this pack too: (1) no funded product peers two separately-owned knowledge bases under access control. That whitespace is real. (2) MIT NANDA (2025): 95% of enterprise GenAI pilots delivered no P&L impact, and 2026 analysts blame "missing or ungoverned context," not model quality. (3) MCP went from niche to ~97M monthly SDK downloads by Mar 2026. The "MCP-native, self-hosted" angle is timely.

Beachhead (who we sell to first)

Small trusted peer circles: 3-5 collaborators (co-readers, study partners, project builders) who already keep notes in Obsidian and are sick of the Git-sync bottleneck. They want a shared "community brain" their agents can search across, without any member giving up ownership of their files, and without a SaaS in the middle.

Why them first: they're the first group that actually exercises the two unique features, federation and per-base access control. (Solo users don't need federation; orgs are blocked by enterprise features we haven't shipped, like SSO/SCIM.)

Ranking from the research: 1) Peer circles · 2) Scattered researchers (easiest to reach, but frugal; they pick free Quartz) · 3) Privacy orgs (will pay $20-65/user, but need un-shipped enterprise features) · 4) B2B partners (high value, hardest to land).

Headline to test first

Federate your local notes into a private memory mesh for your AI agents.

Voice (how we write, non-negotiable)

• Human, specific, concrete. Real examples, real numbers, real commands. If a sentence could appear on any SaaS site, cut it.
• Skeptical and honest. We name the trade-offs and the one real risk (below). No hype.
• Plain language. Short sentences. Say the thing.
• Ban-list (AI slop): "in today's fast-paced world", "seamless", "unleash", "elevate", "game-changer", "robust", "supercharge", "revolutionize", "harness the power", "dive in", "it's important to note", empty tricolons, em-dash pile-ups, and listicles that pad.
• Show, don't claim. Prefer a query trace / a diff / a screenshot over an adjective.
• Cite. Articles link to primary sources (see the essay library). Don't assert market claims without a link.

Voice calibration (the founder's actual voice, match it)

Write like a builder's working log, not a brand page.

• First person, direct. "I did X. It broke at 3am. Here's the fix."
• Concrete over abstract: a real command, a file path, a number, a short code snippet, the actual URL. Show the thing, don't describe it.
• Honest and a little self-deprecating. Name what was hard, what failed, what came out cringe ("burned out for two years", "I did this badly the first time", "it looked cringe").
• Dry, deadpan humor. Understatement, not jokes.
• Short sentences. Plain words. Cut the throat-clearing.
• Ask the reader a real question now and then ("Do you automate this yet?", "What does the world look like in a year?").
• No hype, no corporate gloss. Assume a smart, technical reader who's tired of being sold to.

The energy: someone who has actually been awake debugging this, telling you what they found.

Vocabulary that lands (use these; they're load-bearing)

• "Digital sovereignty": your knowledge should belong to you, not a platform.
• "Local-first": your device is the primary copy; the network is optional (Ink & Switch).
• "Contentless CMS / no shadow copy": the server never keeps a private copy of your notes.
• "Shared second brain / collective memory": the bridge from personal PKM to the mesh.
• "Active middleware": a real-time semantic API gateway for agents (vs a static site).
• "Federated query / query trace": one agent question, answered across trusted bases.

Debates we take a side on (thought-leadership hooks)

• The Git-sync bottleneck: Git guarantees conflicts on files like workspace.json, which makes it broken for real-time note collaboration.
• RAG is a band-aid: passive retrieval is transitional; agents need structured, owned, federated context.
• The "original sin" of partitioning: folders + vector shards sever the connections an agent needs.
• Model layer vs context layer: the model is commoditizing; the context layer is the moat.
• All-or-nothing sharing: why must you choose between total privacy and total public exposure?

The one risk we don't hide

If frontier models ship 10M+ token windows with sub-second latency and no "lost-in-the-middle" decay, a federated mesh becomes overhead for everyone except air-gapped/extreme cases. Our honest answer: ownership + access control + cost don't go away even if context windows grow. We say this plainly. We don't pretend the risk isn't real.

Content plan (what lives in this folder)

Essay library: mirror their tone/structure and cite them

• Local-First Software: You Own Your Data, in spite of the Cloud (Ink & Switch, https://www.inkandswitch.com/local-first/): coined a category + a 7-ideal checklist.
• Building a Second Brain (Tiago Forte, https://www.buildingasecondbrain.com/): shifted "collect" to "organize for outcomes" (PARA).
• MemGPT: Towards LLMs as Operating Systems (Packer et al., https://arxiv.org/abs/2310.08560): the "LLM OS" RAM-vs-disk metaphor.
• Lost in the Middle: How Language Models Use Long Contexts (Liu et al., https://arxiv.org/abs/2307.03172): hard evidence against the "infinite context" hype.
• Evergreen Notes (Andy Matuschak, https://notes.andymatuschak.org/Evergreen_notes): knowledge as compound interest.
• A Brief History & Ethos of the Digital Garden (Maggie Appleton, https://maggieappleton.com/garden-history): library to garden metaphor.
• Data agency / AT Protocol (Martin Kleppmann, https://martin.kleppmann.com/): protocol shifts framed as data-agency rights.
• From Local to Global: A GraphRAG Approach (Edge et al., Microsoft, https://arxiv.org/abs/2404.16130): graph structure fixes a core RAG gap.

Structure to steal from the best of these: name a new category or metaphor → one sharp, falsifiable claim → evidence → a checklist or framework the reader can act on. That's why they spread.