Sorushi vs Reflect: a thinking partner vs a second brain
Reflect is a networked note-taking app with end-to-end encrypted notes and on-demand AI. Sorushi is a journal built around AI conversation as the central interaction. Here's the honest comparison, with a careful look at how the two products land on opposite sides of the same privacy trade-off.
Reflect is one of the most thoughtfully designed note-taking apps in the "second brain" category. It is built around networked notes (backlinks, bidirectional links, daily notes) and pairs that structure with a strong privacy posture: the contents of your notes are end-to-end encrypted, so no plaintext reaches Reflect's servers. Reflect also includes AI features powered by GPT-4 and Whisper, invoked on demand when you select text and ask.
A lot of people use Reflect as a journal. Daily notes plus optional AI plus end-to-end encrypted contents is a reasonable journaling stack, and for someone who already wants notes and journal in one networked workspace, it is a thoughtful answer.
Sorushi is a different shape of tool. It is not a note-taking app. It does not have backlinks or networked notes. It is a journal specifically: a writing space whose entire personality is reading what you wrote and answering back, holding long-form memory across entries, and synthesizing your writing into weekly and monthly reports automatically.
This piece is the honest comparison. Where Reflect is the better tool, especially for people who want a knowledge-base structure around their journaling, it is the better tool by a real margin. Where Sorushi is different, it is a deliberate difference in scope.
What each app is actually built for
Reflect was built around the idea of a networked second brain: notes that link to each other, daily notes for the journaling layer, voice transcription, and AI you can invoke against your own notes. The target audience is knowledge workers, researchers, and people who write long-form. The product treats the note as the unit of work, and the network of connections between notes as the value. Journaling is one use of the product. So are meeting notes, research, idea capture, and book notes.
Reflect's AI is on-demand. The interaction model is "select text, ask AI to do something with it": transcribe a voice note, generate an outline, list key takeaways, fix grammar, run a custom prompt. The product does not respond automatically when you finish writing a daily note. It waits to be asked.
Sorushi is built around a different idea. The product has one shape and one job. You write an entry, and the page reads what you wrote and answers. The response is not invoked; it arrives. The conversation that grows out of your writing over weeks and months is the product. There is no backlink graph, no separate note types, no template gallery. It is a journal, full stop.
Neither approach is wrong. They serve different impulses. If you want a single networked workspace where your journal lives next to your meeting notes and your research, Reflect is built for that. If you want a journal whose entire personality is reading and responding, Sorushi is built for that.
What Reflect does really well
Some honest credit, because Reflect has done genuinely careful work on a hard problem.
End-to-end encrypted notes. Reflect's privacy policy states that the contents of your notes are end-to-end encrypted, so no plaintext reaches their servers. This is rare in the AI-adjacent note-taking category and deserves real credit. For someone who wants the notes themselves private even from the company hosting them, Reflect is one of the few options in the space that genuinely meets that bar.
Networked structure. Backlinks, bidirectional links, and daily notes give Reflect the same affordances that made the networked-notes category popular: ideas connect across notes without you maintaining a folder structure. For people whose journaling lives next to their research and meeting notes, the network is the value.
Voice transcription via Whisper. Reflect uses OpenAI's Whisper for voice transcription, which is among the best in the industry. If you capture thoughts by voice and want them transcribed accurately into a journal entry, Reflect's implementation is strong.
On-demand AI for specific tasks. Reflect AI can outline scattered thoughts, list takeaways from meeting notes, fix grammar, and run custom prompts. For knowledge workers who want AI as a tool to invoke against specific text rather than as an automatic conversation, this is a clean implementation.
A particular aesthetic. Reflect's design is calm, focused, and built for people who care about writing. The product attracts a specific audience (researchers, writers, knowledge workers) and has clearly designed for that audience without trying to be all things to all users.
Cross-environment availability. Reflect ships an iOS app and a web app, with Chrome and Safari web clippers for capturing pages. For a journaling workflow that spans phone and laptop, the combination is workable.
If most of those things matter to you, Reflect is probably the right answer, and you should keep using it.
Where Sorushi is different
Sorushi is not trying to be a networked note-taking app. It is trying to be a particular shape of thinking partner.
The page reads back without being asked. In Reflect, AI is invoked when you select text and ask. In Sorushi, AI responds automatically to every entry. The difference between "select and ask" and "write and receive" is the difference between a tool you use occasionally and a partner you write with. For people who want the conversation to be on by default, "remembering to invoke AI" is a friction that quietly stops the practice.
Long-form memory by design. Sorushi's response to today's entry already knows what you wrote a month ago, three months ago, last year. Reflect's networked structure lets you link to and find old notes; Sorushi's memory does the linking on its own when it responds. Different model. If you want the journal to notice "you have been mentioning this same conflict for six weeks," that is what Sorushi is built for.
Automatic synthesis. Sorushi generates weekly and monthly synthesis reports of your writing, without configuration. A Reflect power user could absolutely run a custom prompt against the past week's daily notes, but that is a workflow you maintain, not a default. The same is true of goal-drift nudges and recurring-pattern flags.
One job, one shape. Sorushi has no backlinks, no daily notes as a separate concept, no graph view. The product does not aspire to be a second brain. For some people, the absence of all of that is unwelcome. For people who tried a networked note-taking tool for journaling and abandoned the system, the journal needed to be one thing, not a graph.
A different privacy trade-off. Reflect's notes themselves are end-to-end encrypted, which Sorushi's are not. But Reflect's privacy policy also openly states that text you send to its AI providers (OpenAI, Anthropic, Google) may be used by those providers as training data. Sorushi uses Anthropic's API under enterprise terms that prohibit training on your content. The two products land in opposite places on the same trade-off: Reflect trusts you to invoke AI rarely on a small selection of text, with the providers holding that selection under standard API terms; Sorushi reads every entry by default, with the provider holding it under no-training enterprise terms. There is no purely better answer here. There are two reasonable, well-documented choices, and which one fits depends on what you most want protected.
A short feature comparison
Only the axes people actually ask about. Anything subject to change is described as "current" rather than fixed.
| Axis | Reflect | Sorushi |
|---|---|---|
| Primary positioning | Networked note-taking (second brain) | Journal, opinionated |
| Responds to entries automatically | No (AI is invoked) | Yes |
| Long-form memory across entries | Backlinks + invoked AI | Built into every response |
| Weekly/monthly synthesis | DIY via custom prompts | Generated automatically |
| Goal nudges | DIY workflow | Auto-detected, drift nudges |
| Daily notes / linked structure | Yes (backlinks) | No (single feed) |
| Voice transcription | Yes (Whisper) | No |
| Native mobile apps | iOS + web | Web only (currently) |
| AI providers | OpenAI, Anthropic, Google | Anthropic Claude |
| End-to-end encryption (notes themselves) | Yes | No (AES-256-GCM at rest) |
| AI training on prompts sent to providers | Privacy policy states providers may use prompts as training data | No (enterprise API no-training terms) |
| Free tier | Paid subscription (see their site) | Free during public beta |
| Paid tier | Paid subscription (pricing on their site) | None currently |
A few notes on the table. Reflect's specific pricing is on their site and is worth checking directly before any decision, since subscription tiers and prices change. Sorushi being free during public beta is the current status, not a forever promise.
Privacy, honestly
This is the most interesting comparison in the whole /compare set. Both products take privacy seriously, and they land in different places on the same trade-off.
Reflect end-to-end encrypts the contents of notes, so no plaintext reaches their servers. This is a strong posture for the notes themselves. When you explicitly select text and invoke AI, that selection is sent to OpenAI's, Anthropic's, or Google's API. Reflect's privacy policy openly states that those providers may use those prompts as training data. For someone who invokes AI rarely and wants their notes themselves private from Reflect-the-company, this trade-off is reasonable and clearly disclosed.
Sorushi takes the opposite trade-off. Notes are not end-to-end encrypted; they are encrypted at rest with application-level AES-256-GCM. But every entry is read by the AI by default, so Sorushi uses Anthropic's API under enterprise terms that prohibit training on your content. For someone who wants the AI conversation as the entire product and is willing to trust the AI provider's no-training terms, the trade-off is different but not worse.
Neither posture is universally better. They reflect what each product is trying to do. If you want the notes themselves invisible to the company hosting them and you'll only ask AI about a small selection occasionally, Reflect's model fits. If you want every entry to be read and answered, with the trust placed in enterprise no-training terms at the AI provider, Sorushi's model fits.
If end-to-end encryption with no AI in the loop is your most important requirement, neither product is the right answer. A private notebook, Day One, or Apple Journal is.
Who Reflect is right for
Reflect is the right answer if any of these are true.
- You want one networked workspace where your journal lives next to research, meeting notes, and reading notes.
- You write in daily notes and link ideas across notes as a thinking practice.
- You want strong end-to-end encryption for the contents of your notes themselves.
- You prefer to invoke AI on a specific selection rather than have it respond to everything you write.
- You'll use voice transcription and want a strong implementation.
If most of that describes what you want, Reflect is likely the right pick. There is no version of this article that asks you to switch for switching's sake.
Who Sorushi is right for
Sorushi is the right answer if any of these are true.
- You want a journal that responds the moment you finish writing, not when you remember to ask.
- You want long-form memory across entries, weekly and monthly synthesis, and goal-drift nudges as defaults.
- You tried a networked note-taking tool for journaling and the structure became the obstacle.
- You're comfortable with Anthropic's enterprise no-training terms on the AI side, in exchange for a journal whose entire job is reading and responding.
- You want a journal that is a journal, not a workspace that contains one.
If those resonate, start a journal at Sorushi. It's free during public beta, your entries stay private, and the first response shows up as soon as you finish an entry of any real length.
Can you use both?
You can, and the pattern is straightforward. Reflect for the networked knowledge work (research, meeting notes, ideas that link across pages) and Sorushi for the reflective writing that wants to be read and answered. They are not the same job, even if they overlap on "daily notes."
If you would rather keep everything in one place, that is also fine. Both products work as standalone journals, and neither requires the other.
The honest bottom line
Reflect is one of the most thoughtfully designed networked note-taking apps in the category, with end-to-end encrypted notes, on-demand AI, and a strong implementation for knowledge workers who want a second brain. Sorushi is a single-purpose journal built around AI conversation as the entire interaction.
If you want a networked second brain that contains your journaling among other things, Reflect is the better answer. If you want a journal whose every entry triggers a substantive response and whose memory works across months without you maintaining a graph, Sorushi is the better answer.
The two products are not actually competing for the same job. Decide which one those sentences describes, and the choice gets a lot simpler.
If you're weighing other options at the same time, the same honest framing applies to Sorushi vs Notion (the other workspace people try to journal in), Sorushi vs Rosebud (the closest direct AI-journaling competitor), and Sorushi vs Day One (a dedicated journal with end-to-end encryption and no AI in the loop). Or read more about how Sorushi actually works.