Five vessels. Three shapes of work.

An operator's context is scattered across ten surfaces — inbox, calendar, three messaging apps, a notes tool, a Drive, a CRM, a stack of dashboards. Every system asks them to start from zero. Nothing remembers what happened yesterday. Switching surfaces means rebuilding context by hand.

And the AI tools they could reach for — ChatGPT, Claude, Gemini — sit adjacent to the work, in a tab, with no continuity between sessions. Each conversation begins blank.

A single private vessel. Their data lives there — email, calendar, Drive, channel histories, journal, notes, files. Kay lives there too, not as a chat app but as a resident agent with continuous memory, running daemons that compose a morning briefing, garden the day's signal each night, and keep the operator's world model continuously fresh.

And the surfaces are mouths for the same brain. The same Kay is reachable from a web workspace, an iOS app, email replies, Slack DMs, and WhatsApp messages — every channel writes back into the same conversation history, every channel reads from the same memory.

When the operator switches from their laptop to their phone to a Slack thread, they pick up where they left off. The brain is the constant.

Two people — one operator and one Kay — run a network of more than fifty customer environments from a single kitchen table. Daily morning seed, intra-day daylog, nightly gardener pass. Every channel reaches the same brain. Continuity survives across days, weeks, surfaces.

A team of operators living between Shopify exports, Amazon Seller Central screens, Meta Ads Manager tabs, Google Ads dashboards, retail-velocity PDFs, and a stack of Looker dashboards that took someone half a day to assemble each Monday.

Every question that mattered — how's the new retail endcap doing?, is the new creative converting in the UK?, what's our true blended ROAS? — required a human to stitch four data sources together by hand. The leadership team had a story about the business that was always a week out of date.

A single vessel that ingests every system the brand runs on. Shopify orders + line items + variant inventory. Amazon Seller Central + ad spend + product-detail performance. Meta Ads + Google Ads + TikTok at campaign and country level. Retail velocity from Spins. Reformulation tracking. Accounting tie-out.

On top of that, a Kay that knows the business. She holds the names of every SKU, every channel, every retail partner. She knows what a normal week looks like. She knows what the CEO cares about in the morning.

And because the data is joined inside the vessel — not stitched at query time — questions that no single tool would answer become one-shot questions.

A 90-day country-level GEO view that joined Meta + Google + Shopify revealed $37K of leakage across 79 long-tail countries — invisible in any single platform. The same view surfaced a GDPR-tracking gap that no platform-side dashboard would show. The daily morning brief now lands in the CEO's inbox by 7am UK; ad-hoc questions get answered in seconds, with the underlying numbers shown. And the finding is structure now, not a memory — next quarter's questions start where this quarter's answers left off.

The growth team was generating leads from Meta and Google ads at decent volume, and qualifying them through a survey on the landing page. But scoring those leads — deciding who to call first — meant a manual stitch each morning between the ad platforms, the survey results, the CRM, and the rep's gut.

The sales lead spent the first ninety minutes of every day rebuilding the same picture. By the time he was ready to call, the leads were already cooling.

A vessel that ingests every channel the growth team runs on — Meta Ads, Google Ads, the survey, HubSpot — and runs a resident lead-scoring agent: its own Linux user inside the vessel, with its own files and credentials, scheduled via systemd to run nightly.

For each new lead, the agent runs a Sonnet-powered enrichment pass with live web search — building a sharper picture of the prospect than any static form could capture — and writes a score with reasoning into the vessel database.

The sales lead opens one page each morning: a ranked, filtered, enriched view of every lead, in priority order, with the reasoning attached.

One page replaced four. The sales lead's daily session moved from rebuilding-the-picture to working-the-leads. Daily engagement on the vessel runs at a level you don't see on dashboards built to be looked at once a quarter. Each week's worked leads teach the scoring what a good lead looks like for this business — the engine is better in month three than month one, which no off-the-shelf tool gets to be.

A fast-growing brand (£190M → £270M) with a deeply founder-led identity, generating thousands of comments per day across half a dozen social and reviews surfaces. The brand sentiment, share-of-voice, and theme-detection questions were split across every channel. The Monday leadership report — a 25-slide deck — took a day to assemble and was already stale when it landed.

And the comments themselves were a vast pool of signal nobody had time to read. Is the new launch hyped or hated? required someone to scroll, by hand, for an afternoon.

A vessel that harvests every surface nightly — IG product posts, TikTok mentions, Reddit threads, Trustpilot, YouTube comments, Yotpo reviews — and runs Haiku-batched sentiment classification across thousands of comments per night, with sentiment-override learning so the model improves on the brand's own vocabulary.

Seven Yotpo storefronts (CA, EU, UK, US, DE, AU, ROW) backfilled with 3,113 reviews, including DE→EN translation. A live Daily Pulse dashboard with comment-level drilldowns. A Founder Pulse view tracking the founder's own mentions. Realtime coverage gates that catch viral posts within an hour for less than a dollar each.

The 25-slide weekly report is being replaced by a live brain that six people across product, marketing, and the founder tier log into during the week — not once a week. Viral launches surface within the hour they break. Yotpo reviews are sentiment-tagged in seven languages, in one view. And what a launch teaches stays learned — the brain holds it, so the next launch is read with everything the last one taught.

A WhatsApp group nobody could read all of. A Notion always six weeks out of date. A Substack archive nobody knew where to start with. Member profiles in a Google Sheet that someone had stopped updating.

The founder was holding the community's coherence personally — remembering who knew whom, what was said when, which threads had legs. The community was thriving and the surface was failing.

A vessel that holds the community's life as it actually flows. A members directory with every member's Q&A profile, tags, and connection history. An events calendar for programs, sidebars, retreats, IRL meetups. A library of every Substack article, shared link, and audio recording, with AI-generated summaries so the archive becomes searchable rather than buried.

The active WhatsApp group ingested daily, with weekly AI-generated digests surfacing key threads. Kay as the community's working memory — the operating team can ask who in the community has worked with this kind of research? or what was said about embodiment at the last retreat? and get an answer drawn from the actual material.

The community has a memory. The operating team holds less in their heads, can spot more across the whole community, and have time for the relational work that's actually their job. First production cross-vessel federation in the network: the founder's individual vessel talks to the community vessel through Sophia — he asks personal questions; the answers draw on the shared brain. Live since May 2026.