The Wend Manifesto

Execution is becoming free

Anyone can build software now. Describe what you want to a coding agent, wait, and it exists. Work that took a funded team a quarter takes a motivated person a weekend. The price of building is falling toward zero, and it is not coming back.

When execution is free, value moves to whatever cannot be executed into existence. You can vibe code a CRM. You can vibe code an inbox, a scheduler, a storefront, most of a company’s software stack.

You cannot vibe code a relationship.

The investor who returns your call at a bad hour does it because of ten years of history. The engineer who follows you into a risky startup does it because of what happened the last time you worked together. No prompt produces that. It is built one conversation at a time, over years, and the context behind it is the most valuable data you own. Today that context is scattered across your inbox, your calendar, and your fading recollection of a conversation at a dinner two years ago.

Agents are fungible. Your network is not.

Within a few years, everyone will have a personal AI agent. It will read your mail, book your meetings, and talk to other people’s agents. That is not a bold prediction anymore; it is a product roadmap.

Here is the part people miss: you will not be loyal to that agent. The best model changes every few months. You switched search engines, browsers, and phones the moment something better arrived, and you will switch agents the same way, ten times over.

Every one of those agents will happily remember things for you. That is exactly the problem. Memory inside an agent is the vendor’s lock-in, not your asset, and the companies racing to build agents have every incentive to hold your context and none to make it portable. A memory that belongs to you will not be built by the companies that profit when it does not.

Your network is the opposite of fungible. It compounds for life. The people you know at 25 become the operators, buyers, and backers you know at 45. A twenty-year asset should not live inside a two-year tool.

So the memory lives outside

The layer has to sit outside every agent: one permanent store that any AI can read from and write to, and that none of them owns. Neutral by construction, because neutrality is the one feature no agent vendor can ship.

Agent-agnostic, so the assistant you use today and the one you switch to next year see the same people, the same history, the same open promises. Portable, so you can take all of it with you in one export. Permanent, so the record outlives every tool change around it. Yours, because the relationships are yours.

That layer is what Wend is. It speaks MCP today, so Claude, ChatGPT, Cursor, and whatever ships next month can all plug into the same graph. When the protocols change, the substrate stays.

People are not a notes file

General-purpose memory tools flatten your life into text chunks and embeddings. People deserve more. A person has a name that changes, roles that accumulate, employers that come and go, things they told you, things you owe them. That calls for structure: a graph of people, organizations, and events, with typed relationships and dated facts. Ask who you know at a firm and get an answer with receipts, not a similarity search over old notes.

Trust is the hard problem

Storage is easy. The hard problem arrives when several agents can write into one memory: which one wrote what, from where, and who checked it. Left unsupervised, a memory of people fills with confident errors, and you will act on them in front of someone who matters. A memory you cannot trust is worse than none.

So governance is the product. The agent proposes; you confirm. Every fact keeps a link to where it came from: the email, the page, the note, the meeting. When two sources disagree, the conflict is shown to you, never silently resolved. And none of this can be retrofitted. A database of orphan facts has no way to earn provenance after the fact; a memory is either born accountable or it never becomes so.

Every fact has a source, and you approved it.

Who this is for

Today: people whose work is people. Founders raising, investors sourcing, sellers closing, recruiters placing, organizers keeping a room warm long after the event ends. Wend already earns its keep with zero AI agents in your life: it builds the graph from the tools you use, briefs you before meetings, drafts the follow-up, and remembers the promise you made in passing. The agent era does not make Wend viable. It makes it inevitable.

Tomorrow: your agent negotiates with someone else’s agent, and both sides need to know who is who, what was said, and what was promised. Every agent will need to know who you know. We are building the layer they will all ask.

Why we build in the open

The engine that holds all of this is open source under AGPLv3. Anyone can read the code that touches their data, run it themselves, and check that it does what we say. A memory built to hold decades of your life cannot ask for blind trust, and it should not depend on one company staying alive. If Wend disappears tomorrow, the engine and your export survive.

Two more commitments, in writing. Your memory is exportable, always, in full; a memory you cannot leave is not yours. And we never train on your data. Not our models, not anyone’s. It is in the terms, not just the marketing.

The agent changes. The memory compounds.

We are building Wend for the person ten years from now who can ask any agent, on any model, about anyone they have ever met, and trust the answer.

The people you already know are the compounding asset of your life. Give them a memory that outlives every tool you will ever use.