This is the first post on a blog that will say how Perpetual Technologies is actually doing. Not a press page. The plain version, written as it happens.
Why write it down here
Most of what we do is visible in two places already: the code, and LinkedIn. The code is honest but nobody reads a commit log for the story. LinkedIn is where the conversation happens, and it is not ours: a post there is gone from the feed in a day and cannot be found again a month later.
So from now on every update gets a home here first, and the LinkedIn post points at it. The changelog says exactly what changed in each version. The roadmap says what is coming. This page is for the part neither of them has room for: why.
Where NeuralVault is today
NeuralVault is a local-first AI desktop app. Your notes stay as Markdown files on your own machine, an assistant called Cortex reads them with real tools rather than guessing, and every tool call is written to an audit log you can open. It runs a local model through Ollama, or a cloud model on your own API key, and the app itself never holds your content on a server.
The Free edition has been a public download for Windows since 20 August, as an early beta.
Why 1.1 is the release that matters
Everything since 20 August has gone into 1.1, and it is the release the product has been pointing at all summer. Three reasons.
The frontend was rebuilt. The first beta shipped on a frontend that was fine for proving the idea and wrong for living in. It has been replaced with a new shell built around the two things people actually do all day, chat with Cortex and work in the vault, and the settings, dashboard and library were rebuilt around that. It is faster to open, faster to move around, and it looks like a product rather than a prototype.
The Personal edition needed a front door. Selling an edition means a sign-in, a licence that the app can check, and a way to buy without leaving the app, and none of that existed. It does now: one account across the website and the app, a licence key issued and checked offline after the first refresh, and a Buy button in Settings that goes to checkout for the right tier.
Not everyone wants to manage API keys. A local model through Ollama is free and private, and a cloud model on your own key is the other honest option, but both ask you to set something up first. Managed routing on credits is the third path: sign in, pick a model, and it works, with the cost shown before you spend it.
The list under this post is what those three reasons turned into. When 1.1 ships, this post will switch to the exact version and the list will become the changelog entries themselves.
What comes after
Once 1.1 is out, the work turns to the thing the whole product is for: an assistant that can act in your vault, not only read it, with you approving each change before it happens and able to undo it afterwards. That is being built now, and it is the subject of the next post.