what your machine needs to think locally
by default, your journal is processed on your own device: no cloud, no account, nothing to configure. that takes real memory, and on linux a GPU. here's the honest bar, and what happens if your machine doesn't clear it.
the floor
| platform | requirement | |---|---| | apple silicon (mac) | about 13 GB of free memory — in practice, a 16 GB machine to have headroom day to day | | intel mac | the solstone app runs here too — it's universal and doesn't bundle the journal's runtime. the journal itself needs apple silicon (true since 1.0.0, not new): use an apple silicon mac for the journal, or run the journal on supported linux instead | | linux | a supported GPU (AMD, NVIDIA, or Intel via Vulkan), roughly 6–8 GB of GPU memory as a practical guide | | windows | the journal on windows now installs the local thinking model too, on a compatible GPU, from local setup in the journal's thinking app. bring your own model works there as well, and you can still pair the solstone app on windows with a journal on a mac or linux machine instead |
the journal's thinking app checks your free memory before it loads anything, so a busy 16 GB mac can still come up short in the moment. that's expected, not a bug. there's no single hard cutoff on linux either: it depends on your specific card and backend, so the journal's thinking app checks your hardware and tells you plainly if it can't run local rather than guessing.
the journal's thinking app checks all of this during setup, before anything downloads.
if your machine can't clear the bar
nothing breaks. you just pick a different way to process your journal:
- bring your own AI engine. a key from Claude, Gemini, or GPT, your own ChatGPT account, or your own OpenAI-compatible endpoint. your billing, your control. a key stays in your journal, and sol pbc is never in the path. with ChatGPT, OpenAI sees the parts of your journal that processing uses.
- wait, and choose later. you can finish setup with no thinking engine chosen. the solstone app keeps taking in what you share, and it still goes into your journal. nothing gets processed until you choose. open the journal's thinking app whenever you're ready.
- confidential processing — available to approved scouts. processing on confidential hardware sol pbc runs, off your own device: no content retained, no human review, nothing used to train, and your journal stays on your device either way.
whatever you choose, the journal's thinking app never quietly reaches for a cloud model on its own: if nothing's chosen, it tells you rather than picking for you. see three ways to process your journal for the full picture.
still not sure?
open your journal (by default at http://localhost:5015) and go to its thinking app. it tells you directly whether this machine can run local, and why not if it can't. or see how much of your computer's memory your journal uses for the full picture.