Némos vs Capacities in 2026 — Object Notes vs On-Device Capture

Capacities pitched a sharp idea in 2022: notes shouldn't all be the same shape. A book is different from a meeting, which is different from an idea, which is different from a person. So Capacities lets you define object types — each with typed fields, custom layouts, and dedicated views. The result is a PKM that feels like Notion's databases without the spreadsheet aesthetic — more like a personal knowledge graph where every node has a proper schema. The German-built product attracted a strong following among academic researchers and writers who want structure without spreadsheet vibes.

The execution is good. The web app and macOS app are polished, the writing experience is fast, and the object-based mental model genuinely scales as your library grows.

Némos goes the opposite direction. Capture is the bottleneck for most knowledge workers, not organization. We optimize for the seconds between "I want to remember this" and "it's saved." Object types and typed fields would slow that flow down. Here's how the two trade-offs actually compare.

What Capacities gets right

Most note apps treat every note the same. Capacities treats notes as instances of object types — a book has a rating, an author, a status; a meeting has attendees, a date, action items; a person has a role, a company, a relationship history. The schema layer turns a note app into something closer to a personal knowledge graph.

The web app is sharp. The free tier is genuinely usable (not a teaser). The community around Capacities templates is active — you can download a "research project" object type or a "reading log" template and have it working in 5 minutes.

For users who think in structured objects and want to model their knowledge that way, Capacities is one of the strongest tools available.

Where the object model adds friction

The trade-off for Capacities' structure is capture speed. When you have an idea on a walk, the steps to save it are: unlock phone, open Capacities, pick the right object type, fill in the required fields, type the actual content. That's 10-15 seconds for an idea that might have evaporated by second 5.

Capacities' developers know this, and they have a quick-capture inbox. But the inbox is the unstructured layer — it bypasses the object types and defeats the purpose of the structured model. Most users end up with two competing patterns: structured objects for things they planned to capture, and a messy inbox for everything else.

Némos avoids this by being the inbox-first system. Everything captures fast and unstructured; on-device AI handles classification after the fact. There's no schema decision at capture time.

Where Capacities pulls ahead

Three real advantages:

1. Cross-platform reach. Capacities runs on web, Windows, Android, Mac, iOS. Némos is Apple-only. For mixed-OS households, Capacities is the right answer.

2. Object views. Want to see all your books sorted by rating? Or all meetings with a specific attendee? Capacities lets you build these views. Némos searches, but doesn't render typed views.

3. Free tier. Capacities' free tier is generous — most personal users never need to upgrade. Némos's free tier is also unlimited but doesn't have the typed-field power Capacities gives away free.

Where Némos wins concretely

For Apple users, the advantages stack:

• Apple Watch capture. Action Button on Watch Ultra starts a recording in 0.5 seconds. Capacities has no Watch app.
• Screenshot OCR. Every screenshot on your phone auto-imports with text indexed. Capacities stores screenshots as images.
• Voice transcription on-device. Voice memos transcribe locally as you speak. Capacities relies on cloud transcription.
• Offline AI. All Némos AI features (tagging, search, summarization) work without internet. Capacities AI requires connectivity.
• No setup tax. Open Némos, capture, done. No object types to define, no fields to configure.

The honest middle ground

For some users, the right answer is both. Capture into Némos throughout the day (Watch, screenshots, voice memos). Once a week, the items that deserve structured modeling get promoted into Capacities with proper object types. Random ideas and references stay in Némos as a searchable archive.

This works because Capacities' strength is at the organize-and-model layer, while Némos's strength is at the capture-and-search layer. Most users won't need both — pick one — but for power users running complex workflows, the combination is reasonable.

Privacy comparison

Capacities is cloud-only. Notes live on their servers. Capacities AI uses GPT via OpenAI. The privacy posture is comparable to Notion or Mem — strong policies, real cloud surface.

Némos has zero cloud surface beyond iCloud (E2E encrypted with Advanced Data Protection). For users where privacy is a hard requirement, Capacities doesn't compete; Némos is the only option of the two.

Real-world workflow comparison

A 31-year-old academic researcher uses Capacities as her primary thinking environment. Every paper she reads becomes a #paper object with typed fields for author, year, methodology, key findings, and citations to related papers in her graph. Every conference becomes a #conference object linked to attended sessions and follow-up actions. The schema discipline pays off when she writes literature reviews — a Capacities query surfaces "all #paper objects from 2024-2026 with methodology = qualitative interviews" and produces a structured table she can export to her draft. This is the workflow Capacities was built for, and it's genuinely good at it.

The same researcher captures heavily during the workday on her iPhone. She photographs a whiteboard from a lab discussion, records a voice memo about a hypothesis while walking between buildings, saves a JSTOR article for offline reading, and jots a quick Apple Watch reminder about a co-author email. Capacities' iOS app exists but is significantly slower than the desktop apps, doesn't OCR photographs, doesn't transcribe voice memos, and has no Apple Watch presence.

Némos captures all four into one indexed library. The whiteboard photo gets OCR'd locally so the equations and notes become searchable text. The voice memo transcribes in real time. The article parses for offline reading. The watch reminder lands in the same library. When she searches "qualitative methodology hypothesis whiteboard" three weeks later, all four artifacts surface. Capacities returns only structured objects because that's what Capacities indexes.

Most heavy Capacities users we talk to use Némos for capture and Capacities for synthesis — they don't compete on the same surface.

The privacy deep-dive

Capacities' data path: every object, every typed field, every query crosses Capacities' backend (German-hosted on Hetzner). Capacities AI features call OpenAI's GPT-4 API with object contents serialized as context. Capacities retains logs and stores backups. The GDPR posture is among the better in the PKM space because the company is EU-based, but the data surface still includes Capacities, Hetzner, and OpenAI. For confidential research with embargoed findings, the surface area is real.

Némos's data path: notes, screenshots, voice memos, and articles are stored locally in MMKV. iCloud sync (if enabled) uses CloudKit with Advanced Data Protection. AI runs on Apple's on-device Foundation Models via the Neural Engine. There is no third-party LLM. For academic work with confidentiality obligations or pre-publication embargoes, Némos is the structurally easier choice.

What happens on a long flight

Capacities is partially functional offline. The desktop app caches recently-viewed objects and lets you create new ones. Sync resumes on reconnect. AI features fail offline. The mobile app is more limited offline because the cache is smaller. New object types and field definitions cannot be created offline.

Némos runs identically online and offline. Voice memos transcribe. Screenshots OCR. Semantic search hits a local embedding index. Apple Watch capture relays via Bluetooth. iCloud sync queues encrypted deltas for landing.

The hidden costs nobody mentions

Capacities' $10-15/month (depending on annual versus monthly) is the visible cost. The hidden costs are subtler. First, the object type model has a steep learning curve — most users invest 10-20 hours designing their schema before they reach productive workflows. Second, redesigning the schema later is painful; field changes don't cascade cleanly across existing objects. Third, the AI features are an add-on billed separately. Fourth, the export is JSON-only and doesn't preserve the visual layout or graph relationships in a way other tools can import; leaving Capacities requires custom migration.

Némos has hidden costs too. No object type model — every Némos note is the same shape. No structured queries. No graph view. We trade structural richness for capture speed and on-device privacy.

Migration friction (a real timeline)

Week one: export your Capacities library as JSON. The export preserves object contents, type definitions, and field values but flattens the visual layout. Install Némos. Drop the JSON into Némos via the share extension — each object becomes a searchable note with type metadata preserved as tags.

Week two: keep Capacities running. Add new captures in Némos for screenshots, voice memos, and articles. Note which Capacities object types you actually use day-to-day.

Week three: assess whether Némos's flat-but-searchable model covers your retrieval needs.

Week four: cancel Capacities if your usage is mostly capture + retrieval. Keep Capacities if you actively use object type schemas for research synthesis.

Total switching cost: roughly six hours across the month.

What Apple users specifically gain

Némos was built iPhone-first. Capacities was built web-first with native Mac and iOS clients added later. The Action Button on iPhone 15 Pro and later starts a Némos voice capture in under one second. The Apple Watch app captures from the wrist. Live Activities pin recording state to the Dynamic Island. Spotlight indexes Némos system-wide. iCloud sync uses Advanced Data Protection. Foundation Models on iOS 26+ run on the Neural Engine. AppIntents expose your library to Siri Shortcuts and Apple Intelligence. None of this surface exists on Capacities. Apple users who pick Capacities get a strong desktop product and a thin mobile one.

The team-of-one vs team-of-many decision

Capacities is increasingly building team features — shared spaces, collaborative object editing, multi-player views. For a research lab or content team, those features matter. For a solo researcher, they're weight. Most Capacities heavy use is solo despite the team marketing. For genuinely shared knowledge work, Capacities is the better fit. For solo knowledge work with strong mobile capture, Némos delivers more value per dollar.