FILE-G · 2026.05.04 · KNOVYA EDITORIAL Field Notes · The State of PKM, 2026

Eight tools, mapped.
Then we built Knovya.

From Notion's all-in-one workspace to Obsidian's local-first markdown turn, from Roam's networked thought to Mem's AI-first bet — the personal-knowledge space spent a decade circling a problem none of them quite solved at once.

Not a "Top 10." Not a sales pitch. An honest atlas of what each tool got right, where each falls short, and what Knovya was built for in 2026.

  • 8 tools
    reviewed
  • 10 yrs of pkm
    evolution
  • ~ 47k words across
    this archive
  • 1 conclusion
    at the bottom
The Path Here

A decade of tools, one missing piece.

  1. Notion launches.

    The all-in-one workspace arrives — pages, databases, sharing — and a generation discovers what knowledge work could feel like. The model is good. The architecture, less so as it scales.

  2. Obsidian's local-first turn.

    March 2020: Obsidian releases its public beta — markdown files you own, backlinks, graph view, no cloud required. The local-first movement has its flagship; sync, mobile, and AI become the asterisks.

  3. Roam, Logseq, Mem, Anytype arrive.

    Roam Research ships networked thought (2019). Logseq launches as open-source AGPL (2020). Mem.ai bets on AI-first capture and raises from the OpenAI Startup Fund (founded 2019). Anytype enters public beta with end-to-end encrypted P2P sync (2023). Each solves a piece. None solves the whole thing.

  4. MCP changes the rules.

    On November 25, 2024, Anthropic introduces the Model Context Protocol. For the first time there is an open standard for AI to read and write across your knowledge — and most of the existing PKM tools weren't built to take advantage of it. The category quietly resets.

  5. Knovya launches.

    Built for the era these eight tools didn't quite reach: MCP-native AI, end-to-end encrypted cloud sync, mobile parity, real teams, and the speed of a desktop app. The tool we'd build, knowing what each of them taught the field.

The Eight Entries

Eight tools, looked at from the outside in.

Each entry below is a careful read of the tool — what's documented as a strength, what surfaces as a recurring frustration in user communities, and where the architecture leaves a gap. The links lead to longer pages with feature comparisons, migration guides, and FAQ.

  1. FIELD NOTE 01 Founded 2016 · NYC · Notion AI added 2023

    Notion

    The all-in-one that became too much in-one.

    What shone

    Templates, databases, sharing — the canonical "second brain starter kit." Notion taught a generation what knowledge work could feel like, especially in teams. The page model is genuinely good.

    Where it falls short

    Slowness compounds at scale. AI arrived as a paid add-on in 2023, then got bundled into the Business tier in 2025 — pushing individual users from $10/month to $20/month for full access. Privacy stays a checkbox, never an architecture.

    The verdict Notion is right when "documents + databases + light collaboration" is the whole problem. Past that, the seams show.

    Read the Notion alternative dossier
  2. FIELD NOTE 02 Public beta March 2020 · Shida Li + Erica Xu

    Obsidian

    Local-first markdown — and everything that costs you.

    What shone

    Plain markdown files the user owns. Backlinks, graph view, a plugin ecosystem that respects power users. Speed of a local desktop app. Free for personal use, no cloud required — Obsidian raised the bar for what a vault could feel like.

    Where it falls short

    Sync as a paid add-on ($4–8/month). Mobile carries real friction compared to the desktop. AI as a plugin you wire yourself. Sharing requires GitHub, Publish ($16/month), or a workaround. Local-first turns into locally-stuck.

    The verdict Obsidian is correct if you live at one desk. The moment your work moves between devices, contexts, or collaborators, the cracks appear.

    Read the Obsidian alternative dossier
  3. FIELD NOTE 03 Founded 2008 · Acquired by Bending Spoons 2023

    Evernote

    The legacy elephant — beloved, then bought, then slowed.

    What shone

    Web clipper that worked. OCR on handwriting. A decade-plus of muscle memory and search across notebooks. For knowledge workers who came up before "second brain" was a phrase, Evernote was the second brain.

    Where it falls short

    Bending Spoons closed the acquisition in January 2023 and laid off 129 staff a month later. Slowness became aggressive. Pricing crept up. AI features arrived years late and sat behind the highest tier. The app got slower as the device got faster — a rare technical achievement.

    The verdict Evernote is what you stay with out of inertia. Most people who leave wish they'd left two years earlier.

    Read the Evernote alternative dossier
  4. FIELD NOTE 04 Microsoft, since 2003 · Copilot enterprise-tier only

    OneNote

    Microsoft's notebook — first-class on Windows, second-class everywhere else.

    What shone

    Free, integrated with the Microsoft 365 stack, infinite canvas, good handwriting on Surface. For Windows-first enterprise teams already paying Microsoft, OneNote is genuinely competent.

    Where it falls short

    Mac and Linux remain afterthoughts. Sync conflicts on iOS keep recurring. The sunset of OneNote 2016 fragmented the user base. AI integration is Copilot-only, locked to enterprise license tiers — not the personal thinking tool people want.

    The verdict OneNote is fine if your life happens inside the Microsoft 365 tenant. Outside it, every cross-platform compromise hurts.

    Read the OneNote alternative dossier
  5. FIELD NOTE 05 Open-source AGPL-3.0 · Launched 2020 · Tienson Qin

    Logseq

    The open-source outliner — beautiful idea, painful logistics.

    What shone

    Outliner-first thinking. Block references that work. Open source under AGPL-3.0, Clojure under the hood, a small but loyal community. Logseq thinks in a way most note apps don't, and the data is yours forever — plain markdown on disk.

    Where it falls short

    Mobile a long-running asterisk. Cloud sync fragile. Performance degrades with vault size. AI integrations unofficial. The "outliner-first" model is brilliant on desktop and painful on a phone.

    The verdict Logseq is a beautiful outliner that asks more from your devices than most devices want to give.

    Read the Logseq alternative dossier
  6. FIELD NOTE 06 Launched 2019 · $15/mo Pro · Bidirectional links pioneer

    Roam Research

    The networked-thought pioneer that stopped shipping.

    What shone

    Bidirectional links and block references that genuinely changed how a generation took notes. The daily-notes model is still one of the cleanest UX patterns in the space. Roam started the conversation.

    Where it falls short

    $15/month with no free tier. The release cadence slowed. Mobile parity never landed. Performance issues at scale. The roadmap drifted while much of the community moved to Obsidian and Logseq. The category leader stopped leading.

    The verdict Roam was the first and is no longer the best at what it invented. Most former Roam users are looking for what comes next.

    Read the Roam Research alternative dossier
  7. FIELD NOTE 07 Founded 2019 · OpenAI Startup Fund-backed ($23.5M, $110M valuation)

    Mem.ai

    AI-first capture — closest in spirit, furthest in execution.

    What shone

    The first PKM tool that genuinely got AI as a primary surface, not a bolt-on. Capture-first UX. Auto-tagging and smart references. Mem's product instinct is right: recall should be a property of the system, not a chore for the user.

    Where it falls short

    No MCP. No end-to-end encryption. AI quality variable. Pricing steep. Mem's bet was right too early — the AI capability they needed didn't fully exist yet, so the product had to over-promise what the model could deliver at the time. Closest spiritual sibling, but not for the era of MCP.

    The verdict Mem proved the category. The next chapter belongs to the tool that adds MCP, encryption, and the AI capability of 2025+.

    Read the Mem.ai alternative dossier
  8. FIELD NOTE 08 Started 2018 · Public beta 2023 · AnySync MIT license

    Anytype

    Local-first, encrypted, open — and quiet on AI.

    What shone

    End-to-end encrypted by default. Local-first sync via P2P. Object-based data model. Free to self-host, AnySync protocol open under MIT, multiplayer arrived in 2024. Anytype is what you'd want if you didn't want to give your knowledge to anyone — including its developers.

    Where it falls short

    No AI features beyond simple search. No MCP. P2P sync expects one device online or the phone falls behind. The learning curve for object-types is real — closer to a database than a notebook. Privacy is the product; AI was never the bet.

    The verdict Anytype is the right answer for the user who wants encrypted local data and is content without AI. For everyone else, the trade is too steep.

    Read the Anytype alternative dossier
At a Glance

The eight tools, summarised in one line each.

Tool Best for Where it breaks Knovya wins on
Notion Docs + databases + light collab Speed, AI architecture, privacy AI-native, MCP, E2E
Obsidian Single-device markdown power user Sync, mobile, AI, sharing Cloud + AI + mobile parity
Evernote Long-time users with .enex archives Speed, pricing, AI lateness Modern speed + AI-native
OneNote Windows-first M365 enterprise teams Mac, Linux, iOS, AI access First-class on every platform
Logseq Outliner-first desktop thinkers Mobile, sync, AI Outliner + sync + AI
Roam Research Networked-thought pioneers Pricing, roadmap, mobile Bidirectional links + a future
Mem.ai AI-first capture believers MCP, E2E, AI quality MCP + encryption + 2025-era AI
Anytype Privacy-first encrypted local AI features, MCP, sync model E2E encryption + AI together
Cross-References

Two on your shortlist? See the head-to-heads.

Sometimes the question isn't "which alternative" — it's "which of these two." We wrote the verdict for the most-asked head-to-head matchups.

The Conclusion

Knovya isn't here to replace any of them.
It's built for the era they didn't quite reach.

Each tool above got one thing right that no one else got first — Notion's all-in-one ambition, Obsidian's respect for your files, Roam's networked thought, Mem's AI capture instinct, Anytype's privacy bet, Evernote's web clipper, OneNote's enterprise reach, Logseq's outliner discipline.

On November 25, 2024, Anthropic introduced the Model Context Protocol — the first open standard for AI to read and write across your knowledge. Knovya launches in 2026 for that era: MCP-native AI, end-to-end encryption, cloud sync without compromise, mobile parity, real teams, and the speed of a desktop app. Built on the lessons each of these eight tools taught the field.

That's the report. The rest of the archive is at /features and /use-cases — if you want to see what was built, and what it's for.

Start free

— Filed by Knovya Editorial · 2026