The folder of zips and grep
Every solo high-volume Claude user already has a folder of exported conversations. Some grep it. Some have a script. Some hit Cmd-F in their downloads folder and hope. The folder is already the product. The question is whether it can answer questions about itself.
Open the downloads folder of any solo operator who has used Claude seriously for six months. There is a directory. It is called claude-exports, or chats, or it is not called anything and the files live next to a year of PDF receipts. Inside it: zip files. Each one a single conversation, exported by hand because the operator wanted to keep the part where the model finally got it right.
The folder grows. Nobody curates it. The names are timestamps and uuids. The operator knows three of them by heart and the rest are dark matter. When they need something, they remember a phrase, open a terminal, and run grep -r "the phrase" ~/claude-exports. Sometimes it finds it. Sometimes the phrase was paraphrased and the grep returns nothing and the operator gives up and re-asks the model.
This is the competitive landscape for Cairn. Not a SaaS dashboard. Not an enterprise platform. A folder of zips and grep.
We say this without irony. The folder is the incumbent. It is free, local, private, and already installed. It loses on one axis only: it does not know what is in itself. The zips are opaque. The grep is literal. The operator's memory of which phrase is in which conversation degrades by the week. Cairn has to make that folder smarter without making it worse on any of the other axes.
1. The product is already shipped
Every solo operator who paid for the API and used it for a quarter has built a personal knowledge base by accident. The conversations are the artefact. The export button is the ship pipeline. The folder is the warehouse. The grep is the query interface.
The accidental product has properties Cairn cannot improve on. It is on-device. No vendor sees the contents. No migration is required. It is in a format the operator's own tools already read. It costs nothing per month and never expires.
It has one failure mode Cairn can improve on, and only one. The folder cannot answer questions about itself. "Did I solve the websocket reconnect thing in March or in April." "Show me every conversation where I was working on the same auth bug." Grep matches strings. The operator wants matches on ideas.
That gap is the entire wedge. One missing capability sitting on top of a product the operator already trusts. Cairn's job is to fill that gap without breaking the properties that made the folder usable in the first place.
2. Why a team product does not help
The first idea a product team has, looking at the folder, is to make it shared. Five engineers, one collective knowledge base, search across each other's conversations. This is the wrong shape for two reasons.
First, the conversations are not shared knowledge. They are private. The operator's chat history with Claude contains half-formed ideas, embarrassing dead ends, the customer name they pasted in by accident at 2am, the password they pasted in and then immediately regretted. The operator does not want a teammate to see any of it. The teammate's manager does not either.
Second, making the conversations shareable means sanitising them. Sanitising means stripping. Stripping means losing the context that made the conversation useful. A conversation rewritten for a teammate's eyes is a summary, and summaries cannot be re-grepped against next month's problem.
A team-shared knowledge base is a different product. It cannot be built by gluing five folders together. The five operators each need their own folder, smarter. The team's collective memory is what Slack is for.
3. Why an enterprise product does not help
The second idea a product team has, looking at the folder, is to sell it to IT. Centralised search across all employees' Claude conversations. Audit. Retention. Compliance. The pitch writes itself.
The pitch dies in procurement. The enterprise sales cycle for a tool that touches employee LLM conversations is six to twelve months. By the time security, legal, data-residency, and seat-pricing reviews conclude, the underlying surface has drifted. Claude's API drifts faster than IT approves new tools. The thing the operator was complaining about in March is gone by November.
An enterprise procurement timeline cannot track a quarterly API release cadence. The tool ships stale or it ships obsolete. The rule in our own playbook: do not select buyers whose decision timeline is longer than your runway, and do not select buyers whose decision timeline is longer than the product's surface lifespan. Enterprise fails both tests for Cairn v1.0.
4. Why "AI-native small team" does not help either
The third idea is the fashionable one. Five-person AI-native startup. Everyone uses Claude. Everyone is technical. Surely they want shared LLM memory.
They do not. A team of five does not share a brain. They share a Slack channel. The decision-grade artefacts from each person's Claude conversations end up in Slack threads, Linear tickets, and pull request descriptions already, by the operator's own hand, with the operator's own filtering. That is the existing mechanism for "the team should know this." It works. It is not the bottleneck.
The bottleneck for each member of that team is the same bottleneck the solo operator has: their own folder of zips does not know what is in itself. Five people on the same team are five separate instances of the same problem, not one bigger problem. Cairn is not trying to replace the Slack channel. Cairn is trying to make each person's private grep folder smarter, and the small team is just five solo operators standing near each other.
5. Why we pick the solo operator
The solo high-volume operator wins on every axis the other two segments lose on.
The buying motion is hours, not weeks. The operator sees the landing page, recognises their own folder in the fold composition, installs the extension, exports a conversation, and decides whether the search is worth nine dollars a month. That entire arc is one evening. There is no procurement step because there is no procurement function.
The decision context is one they can answer themselves. The question "I already grep my zips; would I pay $9/mo to grep them faster and better" is a question the operator already has the data for. They know how often they grep. They know how often the grep fails. They know what the failure costs them. Nobody else has to be consulted. Nobody else has to approve.
The distribution surface matches. Cairn lives on a Chrome Web Store listing and a landing page. Both of those are single-operator install surfaces. They are not channels for team-admin force-installs and they are not channels for enterprise outbound. Building Cairn for solo operators is building it for the only audience the existing distribution can reach.
The product object matches. The hero image on the landing page is one user's downloads folder. ~/Downloads/claude-chat-8f3a2e91-everything.zip. One filename, one user, one filesystem. Anything other than the solo operator would force a rewrite of that fold, and the fold is the strongest design asset we have.
6. The folder is real
The folder of zips and grep is real. We have one. Our friends have one. Every operator we have shown the Cairn prototype to has, within thirty seconds, opened their own folder to show us they have one too. The folder is not a metaphor and it is not a strawman. It is the product the operator has already built for themselves, by hand, because nothing else was good enough.
Cairn does not pretend the folder does not exist. Cairn does not claim to replace it. Cairn extends it. The zips stay on disk. The export pipeline stays the same. The operator's filesystem stays sovereign. What changes is that the folder gains an index it did not have, and the index can answer the questions grep could not.
You already have a folder of zips and grep. Cairn is the version of that folder that knows what is in it.
That is the entire pitch. Not a platform. Not a workflow. Not an AI-native anything. One missing capability, added to a product the operator already trusts, sold to the only buyer whose decision timeline matches the speed at which the underlying surface is changing.
The folder is the competition. We are fine with that.
∞