Privacy, sovereignty and direction: what the Anytype code, the Balderton thread and the governance structure actually tell us

zuollia · June 24, 2026, 8:44am

A few weeks ago I read the Balderton Capital / “vcware” thread on this forum end to end, then went deeper. I cloned all three repositories (anytype-heart, anytype-ts, anytype-kotlin), audited the telemetry / crash / analytics / push paths in the source, read the Swiss commercial register entry for the Any Association, pulled Balderton’s track record (Wonga, VanMoof, the $1.3B 2024 raise, Colin Hanna’s August 2025 departure), cross-checked the “AI replacing human support” rumour against Hacker News, Reddit and this forum, and read Steph Ango’s original “vcware” essay. A tarot card got drawn at the end for the hell of it.

This is the consolidated result. Not an accusation — an attempt to separate what is actually true (from the code and the public record) from what is marketing, hope, or anxiety.

Executive summary

Anytype is not a Signal-style foundation ring-fencing its IP from investors. It is a non-profit Swiss association (the “Any Association” in Zug) sitting alongside a VC-backed Berlin operating company, with overlapping governance. The structure gives users a credible fork-and-export escape hatch (stronger than a typical SaaS) and the founders retain voting control today — but the VCs hold real equity, a (now-former-Balderton) partner holds a board seat in both entities, and the source-available app license + association consent requirement mean the practical “sovereignty” depends on the founders continuing to hold the line.

The code audit surfaced three concrete privacy/sovereignty gaps that contradict Anytype’s “local-first, privacy-first” marketing:

Telemetry is on by default and there is NO user-facing toggle on desktop to disable it. The Go core respects a doNotSendTelemetry flag, but the desktop client (anytype-ts) never sends it.
Hardcoded Sentry DSN and Amplitude API key are committed in plaintext in the desktop source, not injected at build time.
Firebase Cloud Messaging is embedded in the Android app for push notifications — meaning Google (Firebase) sees your push token even in Local-only mode, which is philosophically inconsistent with a sovereignty product.

None of these are showstoppers, but they are real sovereignty gaps that the marketing does not acknowledge. Combined with the “Any Source Available License” (not OSI-open-source), the 90% free-tier cut (1 GB → 100 MB), the aggressive chat/AI roadmap, and the multi-year bug backlog, the direction indicators are mixed-to-cautionary.

PART 1 — Code Privacy & Security Audit

1.1 Telemetry & Analytics — The Biggest Gap

What is collected

Three independent telemetry pipelines run across the codebase:

Pipeline A — In-house telemetry (anytype-heart, Go core)

File: anytype-heart/metrics/service.go:31
Endpoint: https://telemetry.anytype.io/2/httpapi (hardcoded)
Key: INHOUSE_KEY (loaded from env / build config)
Trigger: metrics.Service.InitWithKeys(metrics.DefaultInHouseKey) at clientlibrary/service/lib.go:38 and cmd/grpcserver/grpc.go:88 — runs on every app start
What it sends: app version, platform, device ID, user ID (account ID), working directory path, technical performance metrics (timing samples, long-method warnings)
Interval: every 30 seconds, batched up to 500 events, throttled at 1000

Pipeline B — Amplitude (anytype-ts, desktop + anytype-kotlin, Android)

Desktop file: anytype-ts/src/ts/lib/analytics.ts:12 — URL amplitude.anytype.io (self-hosted Amplitude proxy, not amplitude.com directly)
Desktop key: hardcoded in src/json/constant.ts:5 → REDACTED
Android file: anytype-kotlin/analytics/src/main/java/com/anytypeio/anytype/analytics/tracker/AmplitudeTracker.kt
Android key: injected via BuildConfig.AMPLITUDE_KEY from apikeys.properties
What it sends: account ID (set as user ID), interface language, network ID, tier, plus hundreds of event types (CreateObject, ClickExport, LogOut, AddReaction, PinMessage, etc. — see analytics.event(...) calls throughout anytype-ts/src/ts)
Android mitigation: TrackingOptions().disableIpAddress() at AndroidApplication.kt:115 — IP collection is disabled
Desktop mitigation: none visible — no disableIpAddress equivalent, no user-ID hashing

Pipeline C — Sentry (crash reporting, both platforms)

Desktop file: anytype-ts/src/ts/app.tsx:91-106 — Sentry.init({ dsn: SENTRY_DSN, ... })
Desktop DSN: hardcoded in src/json/constant.ts:4 → https://[email protected]/3
Android: io.sentry:dsey-android via crash-reporting Gradle module, DSN injected from sentry_dsn in apikeys.properties
What it sends: crash stack traces, breadcrumbs (last 20 events), environment (production/development), network mode, packaged flag. URL is scrubbed (e.request.url = '' at app.tsx:97).

The critical flaw — no user-facing telemetry toggle on desktop

The Go core does respect a doNotSendTelemetry flag:

anytype-heart/pb/protos/commands.proto:7783: bool doNotSendTelemetry = 7;
anytype-heart/pkg/lib/initialparams/initialparams.go:61: SendTelemetry: !req.DoNotSendTelemetry
anytype-heart/core/metrics.go:41: metrics.Service.SetEnabled(params.SendTelemetry)

But the desktop client never sends this flag:

anytype-ts/src/ts/lib/api/command.ts:4-13 — InitialSetParameters only accepts platform, version, workDir, logLevel, doNotSendLogs, doNotSaveLogs. There is no doNotSendTelemetry parameter in the function signature.
anytype-ts/src/ts/lib/analytics.ts:232 — calls C.InitialSetParameters(platform, ret.join('-'), userPath(), '', false, false) — doNotSendLogs=false, doNotSaveLogs=false, telemetry flag absent (defaults to false = telemetry ON)

I searched the entire desktop codebase for telemetry, disableAnalytics, enableAnalytics, optOut, setOptOut — no results. There is no Settings UI toggle, no preferences flag, no environment variable check. Telemetry is always on for every desktop user, in every mode including Local-only.

Privacy implication: A user who chose Local-only mode specifically to avoid server round-trips is still sending crash reports to sentry.anytype.io, performance metrics to telemetry.anytype.io, and behavioral analytics to amplitude.anytype.io — all from the desktop client. The Android app is slightly better (debug builds can disable analytics via config.enableAnalyticsForDebugBuilds), but release builds also have it on by default with no user toggle.

Recommendation

Add doNotSendTelemetry to the desktop InitialSetParameters call.
Add a Settings → Privacy toggle: “Send anonymous usage telemetry” (default off for Local-only mode, default on otherwise — or better, default off everywhere with an opt-in prompt on first run).
Move the hardcoded Sentry DSN and Amplitude key out of src/json/constant.ts into build-time environment injection.

1.2 Firebase Cloud Messaging (Android) — Google Dependency

File: anytype-kotlin/device/src/main/java/com/anytypeio/anytype/device/DeviceTokenStoringServiceImpl.kt
On every app start, FirebaseMessaging.getInstance().token is fetched and registered with Anytype’s push-notification service (Rpc.PushNotification.RegisterToken).
Dependency: com.google.firebase:firebase-messaging in device/build.gradle:20 + app/build.gradle:282

Privacy implication: Even in Local-only mode, the Android app contacts Google’s Firebase Cloud Messaging (FCM) servers to obtain a push token. Google sees the device’s IP and a persistent device identifier on every app launch. This is philosophically inconsistent with a “sovereignty” product — there is no way to opt out of FCM while still receiving push, and no way to disable it without losing notifications.

Mitigation: Offer a “No Google services” mode that disables push entirely (acceptable trade-off for sovereignty users), or switch to a self-hosted push path (UnifiedPush / ntfy / polling fallback).

1.3 License — “Source Available”, Not Open Source

All three repositories ship under the “Any Source Available License 1.0” (see LICENSE.md in each repo). Key terms:

Non-Commercial Use (personal, academic, research) → allowed freely.
Commercial Use → only allowed on “Allowed Networks” (networks explicitly authorized by the Any Association, listed at networks.any.coop).
The Any Association can add or remove networks from the allowed list “upon reasonable notice.”
You cannot sublicense or transfer rights.

This is not an OSI-approved open-source license. It is a protective, source-available license. The practical consequence:

You can read, audit, fork, and modify the code for personal use.
You cannot commercially host an Anytype-compatible network without the Any Association’s consent.
The Any Association (governed by founders + Colin Hanna) retains unilateral control over which networks are “allowed.”

Anytype’s founders have publicly committed to migrating to full MIT “over time,” but as of today the apps themselves remain source-available. Only the underlying any-sync protocol is MIT-licensed.

1.4 No Hardcoded Secrets in Source (with one caveat)

I searched for hardcoded API keys, passwords, and tokens across all three repos. Findings:

No hardcoded backend secrets (database passwords, private keys, JWT secrets) found in source.
Caveat: The Amplitude key (REDACTED) and Sentry DSN (https://[email protected]/3) are committed in plaintext in anytype-ts/src/json/constant.ts. These are not “secrets” in the traditional sense (Amplitude keys and Sentry DSNs are designed to be client-visible), but committing them in source rather than injecting at build time means they cannot be rotated without a code change, and they are trivially extractable from any distributed binary.
Android correctly injects its Amplitude and Sentry keys via apikeys.properties (not committed) — better practice than desktop.

1.5 Network Security

Android: app/src/main/res/xml/network_security_config.xml allows cleartext traffic only to 127.0.0.1 (the local gateway). All external traffic must use TLS. This is correct.
Desktop: No explicit network security config found, but Electron enforces HTTPS by default for renderer-process requests.
Local gateway: Listens on 127.0.0.1:47800 (configurable up to port 47900). Binds to loopback only — not exposed to the network. Correct.

1.6 Encryption

Anytype uses end-to-end encryption for all user content (files, objects, spaces). The encryption keys are derived from the user’s mnemonic (recovery phrase) and stored locally — Anytype’s servers cannot decrypt user content even if compelled. This is verified by the source: file blocks are encrypted with symmetric keys (core/files/files.go, pkg/lib/mill), and the key paths are stored in the file object’s metadata which is itself encrypted.

This is the strongest part of Anytype’s privacy story. Even under investor pressure to monetize data, the company literally cannot read user content. The telemetry gap (Part 1.1) is the weakness, not the content encryption.

PART 2 — The Balderton Capital / “VCware” Question

2.1 The Forum Thread

The community thread at community.anytype.io/t/balderton-capital-vcware-and-anytype-s-future/29542 (opened December 2025 by user Kyanite) is the most substantive public debate about Anytype’s governance. Key points from the original post:

Three funding rounds documented:
- Feb 2019: $200k from Techstars
- Mar 2020: $1.2M from Altair, System.One, Tiny VC, Ash Egan
- Aug 2023: $13.3–13.4M Series A led by Balderton Capital, with participation from Connect Ventures, Foreword, Inflection.xyz, New Forge, Protocol Labs, Script Capital, SquareOne, plus angels (Trent McConaghy/Ocean, Jutta Steiner/Polkadot, Luis Cuende/Aragon, Adam Wiggins/Heroku, Peter von Hardenberg/Ink & Switch, the Libermans siblings).
Free-tier cut from 1 GB to 100 MB (90% reduction) — matches the “subsidize-then-squeeze” vcware pattern.
2025 roadmap heavily features Chats, Discussions, AI integration, publishing — expansion into new verticals while core bugs persist.
Balderton Partner Colin Hanna joined Anytype’s board as part of the Series A. His background includes SoundCloud’s aggressive-growth period.
The “Any Association” in Zug, Switzerland is presented as governing the software, but sits alongside a cap table with multiple VC funds and a Balderton partner on the board.

The thread was closed twice by the system “due to a large number of community flags” — indicating significant community contention.

2.2 Founder Responses

Anton Pronkin (co-founder/co-CEO), post #2, 17 likes — the key official reply:

“We actually agree on one your point. If you care about digital sovereignty, you must be self-sustainable as a business. That’s exactly why we’re trying to make more money – not to ‘go vcware,’ but to stop depending on VC at all.”

“We, on the other hand, are not just building ‘an app.’ We’re building a network and infrastructure layer on which many applications can run. Anytype is simply the first client and a way to battle-test the stack. If this works, you get a decentralized network (think ‘Ethereum, but for apps like Anytype, Obsidian-like tools, etc.’).”

“Anytype is a mix of MIT-licensed and source-available components with a clear commitment to go fully MIT over time. Nobody can be prevented from using the core tech for free… Anytype is built by a non-profit Swiss association controlled by two founders and Colin Hanna, with founders holding majority voting power.”

kaye (admin), post #3, 13 likes:

“100%, it’s true that being VC-backed means that Anytype is like any other startup: it can fall prey to dynamics of fast capital. Especially if it loses voting control over its company, which fortunately is not the case today. But if we continue to take more and more funding rounds, then of course, eventually that will happen.”

2.3 The Colin Hanna Detail

A material fact the forum post did not reflect: Colin Hanna left Balderton Capital in August 2025 (per his LinkedIn: “Partner & CEO, Earendil, Aug 2025 – Present; Portfolio Board Member, Balderton, Aug 2025 – Present”). He is no longer a Balderton employee but has retained his portfolio board seats (including Anytype) as a “Portfolio Board Member.”

This slightly weakens the “Balderton is actively steering Anytype day-to-day” framing — the partner on the board has effectively transitioned out of the firm. However, the board seat and Balderton’s equity remain. The governance risk is reduced but not eliminated.

2.4 The Swiss Association Structure — Not a Foundation

This is the most important nuance. Anytype is not structured like Signal Foundation or Ethereum Foundation. It is:

The “Any Association” — a Swiss non-profit association (Verein), registered in Zug in 2023 (same year as the Series A). Swiss commercial register (Moneyhouse): any-20674867741, “active, founded 2023, Management: Colin Daymond Hanna et al., IT services.”
A Berlin-based operating company (Anytype) that employs the team and issued equity to the Series A investors.

The association holds the IP/license rights. The operating company holds the cap table. The relationship is: operating company (VC-backed, builds the product) ↔ association (non-profit, governs IP/license rights), with overlapping people (founders + Colin Hanna) on both.

What this means for sovereignty:

The core any-sync protocol is MIT-licensed — genuinely forkable. An acquirer cannot re-close it.
The apps (anytype-heart, anytype-ts, anytype-kotlin) are under the “Any Source Available License” — commercial use requires association consent. A hostile acquirer cannot simply buy the operating company and commercially exploit the apps without the association’s approval.
But: the association is governed by the same people who sit on the VC-backed operating company’s board. The “firewall” between association and investors is about licensing leverage, not independent governance. There is no legal mechanism (unlike a foundation charter) that prevents a future majority from approving a sale.
Anton’s claim that “founders hold majority voting power” is credible today, but kaye’s admission that “if we continue to take more and more funding rounds, then of course, eventually that will happen” (loss of control) is the real risk indicator.

2.5 Balderton Capital — Track Record

Who they are: London-based multi-stage VC, founded 2000 as Benchmark Capital Europe, independent since 2007. ~$7B total assets. One of the four biggest VCs in London. In August 2024 they raised $1.3B (Early Stage Fund IX $615M + Growth Fund II $685M) — the largest raise for a European VC at the time.

Notable successes: Betfair ($2B IPO), Bebo ($850M to AOL), Depop ($1.65B to Etsy), MySQL ($1B to Sun), Darktrace ($3.5B FTSE IPO), Revolut, Recorded Future ($780M), NaturalMotion (to Zynga). Strong track record on exits.

Notable failures / controversies:

Wonga (UK payday lender) — Balderton + Accel backed Wonga, which was attacked for predatory lending, forced to write off 330,000 customers’ loans, and collapsed in 2018. Balderton + Accel injected a £10M emergency lifeline just so Wonga could pay compensation claims. (The Guardian, FT, TechCrunch, Business Insider all covered this.)
Setanta Sports (liquidated).
VanMoof (e-bike company, went bankrupt 2023) — Colin Hanna is named on Balderton’s page as working with VanMoof.

Investment thesis (2024-2025): Heavily AI-tilted. Portfolio now includes Wayve (autonomous driving AI), Writer ($200M Series C, $1.9B valuation), Light (AI finance), Escape (AI security), Proxima Fusion, Quantum Systems. European AI startups now account for ~18% of all European VC funding, and Balderton is actively chasing this cycle. TechCrunch notes European VCs (Balderton included) are criticized for “Europe’s AI misses” — pressure to not miss the AI wave, which incentivizes AI-tilted investing.

Board practices: Balderton publishes a board-practices playbook and takes board seats as standard for lead-investor deals. A board seat is a contractual right that generally cannot be removed without the VC’s consent.

No smoking-gun article was found alleging Balderton forced an AI pivot or removed a founder at a portfolio company. The pattern is structural pressure (growth expectations, board presence, AI-cycle chasing) rather than overt coercion.

PART 3 — The AI-Support Hypothesis (Debunked)

There is a recurring rumour that Anytype’s support team uses AI heavily, possibly cited on Hacker News. This hypothesis is not supported by the evidence.

Anytype’s official support is a human email desk ([email protected], 24-hour SLA per anytype.io/faq) backed by a human-moderated Discourse forum. Team members anton, kaye, Filip, sambouwer, Dariia, Roman reply personally.
No AI support chatbot, ticket-deflection bot, or Intercom/Zendesk AI agent is documented anywhere — not in their FAQ, not in the community forum, not on HN, not on Reddit.
The only “bots” on the forum are standard Discourse plumbing: an auto-close system bot and an @AnyFriends helper account (which a team member clarified is “just a bot, no human will ever get notified if you reply to them”).

What the research did surface is a different but related issue: Anytype has been shipping AI as a product feature (an in-app assistant called “AI Ally”/“Bobrick” + an MCP server for external LLMs), and that draws heavy community criticism:

Forum thread “DO NOT Waste Time on AI” — 246 likes, 94 posts, 3,107 views.
Forum thread “AI Assistant” (2023 feature request) — 431 likes, 112 posts, 9,159 views. Top critical comment (21 likes): “Just want to voice my support to NOT implement AI into Anytype… Besides it being ridiculously overhyped, there are real issues with data privacy and copyright that go against what Anytype stands for.”
Reddit: “Can we please all tell them to stop forcing AI?”
The recurring theme from critics: AI work is prioritized over critical fundamentals (mobile feature parity, cross-Space search, the editor, CJK input bugs, find-and-replace which sat unaddressed for 3+ years).

The general “VC pressures startups to replace human support with AI” trend is real and documented (Crunchbase: VCs poured hundreds of millions into customer-service AI startups in 2024-2025; Salesforce cut ~4,000 customer-service roles; CNBC reports consumers rank AI customer service among the worst for convenience). But Anytype is not a documented instance of it. The criticism of Anytype + AI is about product AI, not support AI.

PART 4 — The Broader VCware / AI-Washing Pattern

Steph Ango (Obsidian’s CEO) coined “vcware” in his essay at stephango.com/vcware. The core thesis, verbatim:

“VCware is built with a five year horizon, it is not built to live on for decades… In the short term, VCware tends to subsidize pricing to acquire users… But this generally comes at the cost of hoarding user data, and locking in customers… To keep raising money, VCware startups must paint an increasingly enormous vision of their future, which becomes impossible to live up to. This leads to increasingly disparate priorities that gradually make the product worse. What starts off as a useful app becomes burdened with crap. Eventually all VCware must exit.”

The “AI-washing” phenomenon is now well-documented:

Built In: “AI feature creep — adding AI tools without clear user value — can bloat products, waste resources and erode engagement.”
RM Magazine (Jan 2026): “Criminally Overhyped: The Risks of AI Washing” — notes SEC’s first AI-washing enforcement actions (March 2024).
CNBC: Microsoft VC’s warning about dubious AI usage claims from companies.
PitchBook Q4 2024: AI companies received 50.8% of global VC funding.
Mintz: AI company funding exceeded $100B in 2024.

Anytype’s symptoms match the vcware pattern:

Subsidize-then-squeeze (1 GB → 100 MB free tier)
Expanding roadmap into new verticals (chat, AI, publishing) while core bugs persist
Board seat held by growth-focused VC partner
Feature-velocity prioritized over depth (CJK bugs, find-and-replace, export reliability — all documented in this conversation’s two bug reports)

But Anytype also has counter-signals that distinguish it from pure vcware:

End-to-end encryption means the company cannot monetize user data even if pressured
MIT-licensed core protocol is genuinely forkable
Founders publicly state the goal is to reduce VC dependence via paid membership
Colin Hanna left Balderton in Aug 2025, reducing day-to-day VC influence
The “Any Association” structure gives users a credible exit path (export + fork)

PART 5 — Direction Indicator

Healthy signals

E2E encryption is real and verified. User content is cryptographically inaccessible to Anytype, investors, or any future acquirer. This is the strongest structural protection.
Core protocol is MIT. The any-sync layer is genuinely open. An acquirer cannot re-close it. A community fork is legally and technically possible.
Founders retain voting control today and publicly commit to reducing VC dependence.
Local-first architecture means users hold their data. Export (when it works — see the export bug) produces a real, portable copy.
Colin Hanna’s departure from Balderton (Aug 2025) reduces active VC steering.

Unhealthy signals

Telemetry is on by default with no desktop toggle — contradicts the sovereignty brand, especially in Local-only mode.
Firebase dependency on Android — Google sees your push token even in Local-only mode.
“Any Source Available License” is marketed as “open” but is not OSI-approved. Commercial forks need association consent.
90% free-tier cut matches the vcware squeeze pattern.
Chat/AI roadmap prioritized over core bug fixes — find-and-replace unaddressed for 3+ years, CJK input bugs, export silently dropping files (documented in this conversation), local-only mode broken offline (documented in this conversation).
The association is governed by the same people who sit on the VC-backed operating company’s board — the “firewall” is licensing leverage, not independent governance.
No legal mechanism prevents a future majority from approving a sale. A Series B would dilute founder voting power (kaye’s own admission).

Overall assessment: CAUTIONARY — Monitor Closely

Anytype is not yet vcware, but it is exhibiting early vcware symptoms. The next 18-24 months are the critical window. Watch for:

Series B announcement — this is the single clearest inflection point. If Anytype raises again, founder voting control erodes and the vcware trajectory hardens.
License migration to full MIT — if the apps move to MIT as promised, the sovereignty guarantee becomes real. If this stays “coming soon” indefinitely, the source-available license remains a lever.
Telemetry toggle added — a genuine privacy-first product would let users opt out of telemetry. If this never ships, the sovereignty marketing is hollow.
Bug-fix-to-feature ratio — if chat/AI/publishing continue to dominate releases while export, local-only offline, and CJK bugs persist, the priorities are misaligned.
Colin Hanna’s next move — if he transitions fully out of the board, VC influence drops further. If a new Balderton partner takes his seat, influence returns.

The honest answer: Anytype today is a privacy-respecting product with real cryptographic guarantees, built by founders who genuinely seem to want independence, operating inside a governance structure that is better than a pure VC startup but worse than a foundation. The structure can survive founder turnover or investor pressure only if the MIT migration completes and the telemetry toggle ships. Until then, treat Anytype as “trustworthy but monitor closely” — encrypt your data, keep local exports (when the export bug is fixed), and be prepared to fork if the signals turn negative.

PART 6 — Tarot Reading

I randomized a Major Arcana card

Card Drawn: The Tower (XVI) — Upright

Keywords: sudden change, upheaval, chaos, destruction, dramatic change, revelation, awakening, loss and ruin, new start, unexpected events, foundational change.

Traditional meaning: The Tower is the card of abrupt, unavoidable disruption. A lightning bolt of clarity cuts through lies, illusions, and false assumptions a structure was built on, and the truth is exposed. Whatever was constructed on shaky foundations comes crashing down. You cannot escape it; the only path is to surrender to the collapse and let the old structure self-destruct so something real can take its place.

In a career/business/finance context: Expect sudden disruption, forced restructuring, or a wake-up call. The mess is often temporary and clears the way for a more stable, authentic organization long-term — but only if you do not resist the change.

The transformative aspect: The destruction is not punishment but necessity. Old foundations must be replaced with something more genuine. Growth, enlightenment, and a new perspective follow.

Advice: Do not resist or deny the change. If you are self-aware, you may see the cracks forming and transform before forced collapse. Let the false structure fall, then rebuild on truth.

Application to Anytype

The Tower upright maps cleanly onto Anytype’s situation. The current structure — a VC-backed operating company sitting alongside a non-profit association, with a source-available license, always-on telemetry, and a roadmap tilted toward chat/AI while core bugs persist — is built on partially shaky foundations. The “lightning bolt” could take several forms: a Series B with dilutive terms, a community fork triggered by a governance dispute, a security incident that exposes the telemetry gap, or a strategic pivot that breaks the trust of the privacy-first user base.

The card’s counsel is not that collapse is inevitable, but that clinging to the compromised structure makes the eventual collapse worse. The healthy path is to proactively dismantle the unstable elements — ship the MIT migration, add the telemetry toggle, fix the export and local-only bugs, resist the chat/AI feature bloat — before an external force imposes the change. The chaos of the Tower is the mechanism by which a more authentic product emerges. Anytype’s founders have the agency to choose whether that transformation is voluntary or forced.

The Tower does not say “Anytype will fail.” It says “the structure as it currently exists will not survive unchanged, and the sooner that is accepted, the better the rebuild.”

Sources

Code (directly audited): anytype-heart (metrics/service.go, core/metrics.go, pkg/lib/initialparams/initialparams.go, pb/protos/commands.proto, core/files/, space/spacecore/localdiscovery/) · anytype-ts (src/ts/lib/analytics.ts, src/ts/lib/api/command.ts, src/ts/app.tsx, src/json/constant.ts) · anytype-kotlin (analytics/, AndroidApplication.kt, DeviceTokenStoringServiceImpl.kt, app/build.gradle, network_security_config.xml)

Forum & community: Balderton Capital, VCware, and Anytype’s Future · DO NOT Waste Time on AI · AI Assistant · Please give us a way to turn AI integration off completely if you're going to add it · https://www.reddit.com/r/Anytype/comments/1tqz1tx/can_we_please_all_tell_them_to_stop_forcing_ai

Funding & governance: https://blog.anytype.io/anytype-raises-13-4million-usd-funding · Anytype announces $13.4 million round following launch of open Beta - Balderton Capital · The anthesis web: Anytype raises $13.4M to restore the Internet - Tech.eu · https://blog.anytype.io/our-open-philosophy · https://legal.any.coop · Any in Zug - Reports | Moneyhouse · Balderton Capital - Wikipedia · Euro VCs welcome Balderton's fresh $1.3B but grumble about Europe's AI misses | TechCrunch

Balderton controversies: Private equity firms behind expansion of Wonga.com | Wonga | The Guardian · Wonga investors inject £10M so cash-strapped payday lender can fund claims | TechCrunch · Awkward Silence As Top European VCs Are Asked About Failures at Wonga - Business Insider

VCware / AI-washing pattern: 100% user-supported — Steph Ango · How to Beat AI Feature Creep and Build Products That Truly Matter | Built In · AI washing: A Microsoft VC's warning about dubious, and rising, corporate artificial intelligence claims · Risk Management Magazine - Criminally Overhyped: The Risks of AI Washing · VCs Put Customer Service AI On Speed Dial · 'I hate customer-service chatbots': The consumer-AI refund relationship is off to a rocky start

Tarot: https://www.biddytarot.com/tarot-card-meanings/major-arcana/tower/ · https://www.trustedtarot.com/card-meanings/the-tower/ · https://www.labyrinthos.co/blogs/tarot-card-meanings-list/tower-meaning-major-arcana-tarot-card-meanings

Colin Hanna: colin daymond hanna - Earendil | LinkedIn · Get to know Balderton’s Colin Daymond Hanna, our first Partner outside London | Balderton Capital

Zak-from-Zork · June 24, 2026, 7:53pm

This is information I was always looking for!

If that is all correct it is very interesting stuff and took a lot of work to gather.

Thanks a lot for that!

Astro-L · June 25, 2026, 12:27am

I have absolutely no idea how to evaluate the veracity of this, and I have next to zero context in which to frame a lot of it. Brand-new user, almost uncannily clear writing, and weird tarot digression? I’m taking this with a mine’s worth of grains of salt. But it was a fascinating and easy read, and it’s at least the kind of information I very much appreciate knowing, though of course I would love to have more confidence in the content.

[Edit:] Of course there’s a lot that I could verify via the source-links at the bottom - I do very much appreciate that those are there. But I think whatever I have the capacity to confirm or disconfirm by doing so is on a level below what I would really need to feel confident in the analysis. Like, what it means for the code to be this way or the investors to have done that thing. And I’m feeling some kind of Spidey-sense about the irony of the OP having done this massive job intending to shed light on things that were previously unclear while being personally a cipher. When they say “I read”, “I cloned”, etc., I have no sense of who or what that “I” is. Ultimately if the facts and analysis are accurate, I suppose it doesn’t really matter so much, but I’ll wait and see how this thread plays out.

ACstudio · June 25, 2026, 8:56am

I share your view. I have the same doubts, but I appreciate the work done and hope that Anytype becomes increasingly open and transparent. Right now, it’s the app I use for work every day.

I apologize for my “translator-style” English…

kaye · June 26, 2026, 9:46am

Thanks for taking the time to put this together and for sharing. It’s wonderful to see such engaged community members. I apologise that I cannot address every detail, because there are simply too many and a few are not quite relevant/accurate.

Investor influence and the board

This is not the best framing to give the community because it overly focuses on the board. Focusing on the board seat is a detail, sure, but it misses the forest from the trees.

As long as a startup is an unprofitable business that cannot sustain itself from customer revenue, it must rely on taking in external money to pay the bills. As long as that paradigm exists, a startup must juggle two things: please customers to gain traction and please investors to keep the product alive. These two are interconnected and ideally aligned, but it’s not always the case (for many reasons too verbose to explain here).

As stated in our previous post you referenced, this is why focusing on revenue/profitability is important. It’s not because people are greedy and just want to milk the community, it’s because it’s a way to stay alive and aligned to a mission.

In short, as long as Anytype relies on external money to pay the bills, it will always be under investor influence—just like any unprofitable startup. A way reduce/remove this dynamic is actually very simple: get customers to pay. However you can see how this solution feels ‘against users’ at times, because users naturally want as much free stuff as possible—we’ve all become accustomed to it due to VC-backed startups and data harvesting products being able to offer products for free.

It’s easy to paint the investors as the villain, but the reality is that every business wants to be able to sustain itself. Anytype is not profitable today. I understand why looking at the ‘90% free-tier cut’ seems like an unhealthy signal. And indeed, it’s a decision to try and spur more users to pay for the product. But I would argue it’s less of a VCware bait-and-switch thing. It’s just fundamentally a business sustainability thing.

VC backed startup

Again, this framing is too simple and that looks at things in a linear way. We just need to think about it from fundamentals.

A startup that takes VC money in the first place is a very specific kind of business. VCs take very big bets in hopes of very big returns. This is a completely different paradigm from taking $20k from friends and family to setup a food truck. VCs are not interested in small % year-on-year growth otherwise investors would just invest their money into other asset classes. This is why VCs overwhelmingly invest in software companies that have very high growth potential models, they don’t invest in boutique fashion stores.

VCs will always look for some kind of exit where they earn big returns on their investment. The framing of “VCware trajectory hardening at Series B” may hold true from a narrow frame, but it misses the main point: any VC-backed startup is pushed toward big growth from day one. Anytype is no exception, and that fundamental dynamic persists across all funding rounds. Importantly, the Anytype team couldn’t have built what is, dare I say, genuinely innovative local-first sync technology without external funding—because it’s incredibly expensive to do so (many years of engineering and its still a work in progress).

To be clear, ‘big growth’ doesn’t necessarily mean ‘bad’. Although I understand and can agree with that perspective in many circumstances, especially from a sustainability perspective (shoutout to fellow solarpunkers). In the case of Anytype growth, if more people in the world used local-first software vs. cloud software, I think it’s growth in a more positive direction.

I would argue the better framing is: you should not use a product from a VC-backed startup that is unprofitable if you worry about high velocity growth dynamics, period. Instead, look for bootstrapped startups or already profitable businesses, you’ll likely be better served there. 37signals writes a great post about this.

Roadmap priorities and bug fixes

To set the stage first: what one part of the community sees as a priority is not what the other part sees as a priority—there is no such thing as ‘one Anytype community opinion’. Our recent community survey shows that the majority of our users want new features, they don’t want us to focus only on bug fixes (20% of respondents). I will share these results in the next Town Hall.

To your point, your framing states: if we work on a chat-related feature instead of local-only mode, we are not prioritising correctly and exhibit VCware velocity prioritising behaviour.

Our local-only user base is close to 1% and the amount of users using chat is many orders of magnitude larger than this—these are objective stats. I know it doesn’t sound accurate to community sentiment, but it’s because local-only users are over-represented with their data issues. But by simple maths, the reality is actually the opposite of your position: prioritising development on local-only over chat, is in fact, not aligned with our community needs.

This proposition of ‘bug fixing-to-feature ratio being a VCware pattern’ is, frankly, just false.

I’ll add the obvious point that we’re able to know this objective information, have this discussion with you, and make better product decisions because we have product telemetry. Without telemetry, we would not know how many active local-only vs. chat users exist.

Again, it’s important recognise that every user feels like they represent the average Anytype user, but we have a very diverse user base.

Telemetry

I wrote a very detailed post summarising general points on how we approach data security, privacy, and anonymity—it’s worth reading for context.

To be clear to non-technical readers, telemetry is not visibility into the content in your spaces. Your content remains private to you. Telemetry is anonymised product usage data that is not correlated with user identifiable information and is compartmentalised on our internal systems. It is viewed in aggregate to inform product decisions (such as how many people are using local-only mode) and this data is obviously not sold on. All of this is covered in our privacy policies, which are clearly stated to all users and not hidden.

There’s little point in debating the legitimacy of collecting telemetry data here. However, I will say it’s a strong position to hold that ‘sovereignty marketing is hollow’ if a product collects anonymised telemetry. From my point of view, sovereignty is not a binary state, it’s a spectrum with multiple dimensions. This argument ignores the many aspects of sovereignty in Anytype—local-first, permissionless sign in, self-hosting, etc.

I think a fairer framing could be: sovereignty will improve if telemetry was turned off. In relative terms, I feel Anytype is a high sovereignty product in its class compared to what is available in the marketplace. I understand this is a subjective opinion.

A separate but relevant point: recently we removed Graylog because we found a better way to go about it.

Licenses

To add clarity: today, anybody is able to use Anytype and modify it to their hearts content—for use in their personal or work life. The big ‘sticking point’ of the ‘source available’ license is that if you want to earn money from it (commercialisation), you have to get permission. This is what’s being described as the ‘lever’. How much this change of commercialisation permissions would be meaningful to Anytype users is a matter of perspective and opinion. Yes, MIT License and open source is great, but I think adding specifics on what the practical difference would be to the average user is important.

--

I hope this additional context was helpful to the community.

atmos · June 26, 2026, 8:38pm

I hopefully can speak for at least some of the 1% here, and I want to share how this looks from our side.

After reading everything, the honest conclusion is that Anytype may not be the product for people who don’t want to be dependent on a cloud server. I understand that local-first is still better than any commercial software where everything sits in the cloud unencrypted. But there’s a gap between what the marketing says and what the product delivers, and it’s widest exactly where local-only mode lives.

The homepage says “Inspiration doesn’t need Wi-Fi. Deep in the mountains or off the grid, Anytype never stops working.” The FAQ says local-only mode “means that data will only be synced across verified devices in your local network.” The Why page says “application developers are the keepers of our keys… they determine who has access.” These are the reasons the 1% chose local-only.

But the default is Anytype’s servers, nothing wrong with that. The app makes external connections even in local-only mode. Telemetry sends data to three servers with no opt-out on desktop. Firebase sends push tokens to Google even in local-only mode. It seems against the mission, but without the telemetry the product cannot be correctly improved, that’s understandable. Images fail to load without a network interface. Export drops files silently. The one mode that comes close to delivering what “Inspiration doesn’t need Wi-Fi, off the grid, Anytype never stops working” promises is labelled experimental, is being considered for removal, and users are told “you don’t actually need this, use the default mode instead.”

To be fair, Anytype’s marketing is carefully written. It says “local-first,” not “local-only.” It says “offline-first,” not “offline.” The FAQ describes local-only as a choice, not a promise. The homepage says “no server means no lag,” which is about performance, not sovereignty. The language is precise enough that you can read it as a performance claim rather than a sovereignty claim. But when the homepage also says “no server, no gatekeeper” and “nobody is mediating the connection between your devices,” and then the product requires a server in default mode and is broken in the one mode that doesn’t, the gap is there whether the marketing intended it or not.

The Open Philosophy blog says “By opening our source code, we ensure that our users have complete autonomy and independence from the Any Association” and that this guarantees “uninterrupted access to the tools and data they generate and store, shielding them from any potential restrictions.” Read in context, this is specifically about the ability to fork, compile, and run the code yourself, not about the default product experience. But even in that narrower sense, the source-available license (not MIT) means commercial forks need association consent, so the “independence” is partial. The MIT migration, if it happens, would close that gap.

On the 37signals link. I appreciate the honesty in pointing users there. The 37signals essay says “when you have to source raw materials from a very limited number of suppliers (investors), the money comes with all sorts of strings attached. Money with strings attached isn’t really yours.” That line describes the dynamic this whole thread is about. Anytype took VC money. The strings are growth. Growth pulls toward features that attract new users, not features that serve 1% of the existing ones. That’s not a moral failing. It’s structural. But the 37signals model and the Anytype model are different paths, and the outcomes are different too.

The pattern is genuinely frightening for those of us who built our digital lives around the sovereignty claims. I don’t say that lightly. When you read that local-only is “not worth it,” that it’s “the cause of a lot of support tickets,” that removal is “highly considered,” and then you look at the homepage that says “off the grid, Anytype never stops working,” it doesn’t leave you sleeping nicely at night. It feels like the thing you relied on could end one day, and the marketing that made you trust it will quietly be updated to match the new reality.

So the biggest question for the 1% is probably this: how much influence does the VC actually have right now? Not in a paranoid sense, but in a practical one. Balderton’s portfolio is heavily AI-tilted. Anytype ships chat and AI. Balderton expects growth and exit. Anytype chases growth and is unprofitable. The intentions may be good. The founders may hold voting control today. But the patterns and the mission tell a different story that could unfold, and the 1% are the canaries in that coal mine. When local-only goes, the sovereignty promise goes with it, and the homepage will need to be updated to match.

I don’t expect the team to change their priorities for 1% of users. I just hope the marketing can be honest about what the product is and isn’t, before the gap gets wider. The docs update is a good start. The homepage is what new users see first. So looking at the code, the GitHub PRs and the GitHub issues, AI and chat have high priority. Local-only is just there. So if we are being honest here, why is it being offered? If it’s such a burden to maintain, experimental for years and neglected with critical bugs, then I cannot fathom why it exists even in the first place. Some people may have requested it. It’s a nice addon to have. But I cannot remove the gut feeling that something is going to happen to Anytype. I don’t feel good using the software. It’s not even the bugs. It feels off. I’m in no way trying to make the product bad. But I don’t feel, in all honesty, confident using the product, despite it being on the market for so long. That’s my honesty. I really didn’t want to answer any of this or even say how I feel. Maybe some might understand it.

So it might be just better to remove the local-only mode altogether for peace of mind, but that could add even more backlash than needed. Maybe it’s best for the users to decide in this forum what to do with the local-only mode. But the majority aren’t in this forum. Most don’t know what local-only means, and as the numbers suggest, 1% knows. The other percentages probably do not or don’t care.

Sadly, I really liked the UI/UX, how fast it’s loading compared to Obsidian default, and that’s the only thing that somehow kept me considering using Anytype. But the product doesn’t want me. Yeah. These are my thoughts. Yeah, they’re messy and aren’t perfectly crafted. I know it sounds like I shit a lot on Anytype, only pulling it into the negative. Well, I hope this doesn’t come off as too aggressive or mean. Maybe pen and paper is the truest way we could choose, I don’t know.

Zak-from-Zork · June 28, 2026, 9:56am

I think it’s a strong point that Anytype claims to be a ‘safe haven’.

So, at least to me, local only and ‘on premise’ are two strong pillars, at least emotionally.

And since I am not a tech guy, I have no chance to run my own ‘local only’ instance or on premise without professional support in the long term.

I think most ordinary users are in the same position.

So what about paid support services?

Or paid update services for on premise and local-only?

I think there could be useful plans for that. Or it could just be part of the regular subscription.

Maybe with added special support for on-premise solutions.

If local only generates so many tickets, why not have a subscription for local only support or a little pay per ticket fee?

Or paid updates?

It’s about the eternal principle of giving and taking, isn’t it?

And if the Anytype team does great work on software and services, we should be honest enough to give something back that helps them even to be sovereign enough, not to take any more VC money.

P.S.: Just started a subscription.

kaye · June 29, 2026, 8:02am

@atmos I want to repeat a point I’ve made in other threads: local-only is a valuable model to have with real sovereignty benefits. However, Anytype was designed on local-first principles, which solves for a different set of problems.

I apologise if you feel misled. As mentioned in other threads, in places you feel the marketing is not accurate, I’m more than happy to update it. On the examples you used, I can understand why you hold your perspective from the local-only angle. However, I don’t think it’s a fair representation when you look at it from the other perspective (the majority of users wanting local-first sync).

The local-first dream, what Anytype was built on, is to serve the people who want the benefits of cloud software while minimising the sacrifices to your sovereignty. The marketing about ‘working off the grid’ and ‘holder of your keys’ is addressing this angle.

Think about it from the average user: they like Notion as a product but don’t like its lack of privacy, security, proper offline mode, etc. Many users come to Anytype to solve these specific issues. They do want to have their notes sync between their devices using the internet and not manage their own backups. They don’t want the app developers to hold keys to their accounts.

The problem the majority of users are looking to solve isn’t: ‘not use any server ever’.

If you read the Anytype marketing as local-first being a reaction to cloud-first services, this is exactly what is promised and (hopefully) delivered to users. The app should keep working even when you go off the grid; it’s not expecting you to never be connected to the grid ever. For you to be the only holder of your keys; it’s not expecting that your data never syncs with any server ever.

I feel you’re interpreting the marketing as being disingenuous because you’re evaluating it under your local-only desires. But when you change the desires to what many local-first users want, I think it’s a different story. I don’t feel the marketing was ‘carefully’ written to skirt what is defensible. It was written for a local-first world.

Again. This doesn’t mean local-only isn’t valuable or doesn’t have its place. It’s just that Anytype wasn’t built from the ground up to solve local-only problems, it was designed with e2ee sync.

We will have more to share on the next Town Hall, however I think this conversation is being blown a little beyond its scope. My initial comments on ‘local-only’ were on a different point—it was in response to the original posted saying Anytype is VCware when it prioritises chat features over local-only.

@Zak-from-Zork brings up a pertinent point: local-only users would need a very different business model, pricing strategy, go-to-market approach, etc. That’s all true. However, the bigger barrier is not on a technical level, it’s on a user level: we see that most people today don’t want local-only.

atmos · June 29, 2026, 1:24pm

I’ve made my decision. Stepping out of this thread and off Anytype. I want to explain why, because I think it’s relevant to the direction conversation.

The anyproto org README opens with “Trust our code, not our words” (.github/profile/README.md at main · anyproto/.github · GitHub). I took that seriously. The code says local-only is experimental and deprioritized. Your own posts say it generates too many support tickets. The homepage says “No server, no gatekeeper” and “Nobody is mediating the connection between your devices” (https://anytype.io/). The code, the team’s statements, and the marketing don’t line up, and I’m trusting the code.

On give and take. Zak framed it as “the eternal principle of giving and taking,” and your post made the same point. Users want free stuff, servers aren’t free, the team needs revenue. Fair. The team built software for Android, iOS, Mac, Windows, Linux. The engineering years are real, the servers cost money, and I’m not asking for free labor.

But there’s a line I want to name. Paying for a service that runs on someone else’s server is paying for costs. Paying to access data that lives on my own device is a different category. The first is exchange. The second feels predatory, especially in an age where we own less and less. Games get pulled from stores. Movies disappear from “purchased” libraries. Software goes subscription-only, and the “buy” button has been lying for fifteen years. Notes aren’t games or movies. Anytype didn’t write them. The notes are mine. Renting access to my own notes is a category I can’t accept, even if the encryption means the company can’t read them.

That’s why I moved to Obsidian. Steph Ango’s “file over app” essay (File over app — Steph Ango) puts it simply: apps are ephemeral, files can last. My notes are plain markdown now. No company can update, revoke, or paywall them.

On local-only’s future. You’ve pointed to the 99% who don’t use it. If that’s accurate, I think the honest move is to remove the mode. Keeping it as experimental, neglected, and support-ticket-generating serves no one. Not the team. Not the 1% who hit its bugs. Not the 99% whose features get deferred. Paywalling local-only support, as Zak suggested, would confuse the remaining users and accelerate their departure. Removing the mode cleanly, with honest marketing, would let Anytype be what it actually is: a Notion alternative with encryption and offline access. That’s a good product. It’s just not the product the homepage sells.

I do hope self-hosting stays. It’s the one mode that delivers real sovereignty for the technical user. The official setup is difficult, but a community member has built an easier bundle at GitHub - grishy/any-sync-bundle: Anytype Bundle: Prepackaged All-in-One Self-Hosting · GitHub. I haven’t verified its safety, but it points toward what a more accessible self-host experience could be. If local-only goes, self-hosting becomes the only sovereignty path, and it should get more accessible, not less.

The code is source-available, which is better than closed. But the freedom to fork, modify, and self-build is currently limited to the small percentage of users with the technical skill to compile and maintain a fork. That’s changing. AI coding tools are giving even non-technical users the ability to read, modify, and rebuild the software they use. Source-available matters more now than it did three years ago.

So: Anytype is a Notion alternative, not an Obsidian alternative. Notion but better. Encrypted, offline-capable, with optional self-hosting. For Notion users, clear win. For sovereignty purists who want no server, wrong product, and the marketing should say so plainly. The team built something genuinely difficult across five platforms. I just wasn’t the audience, and the marketing made me think I was.

I might lurk now and then to see how the project develops. I hope Anytype reaches self-sustainability and exits VC dependence. The local-first ecosystem needs more independent players, not fewer. Good luck with the Town Hall.

kaye · June 29, 2026, 1:38pm

I appreciate you chiming in and sharing your perspective. The conversation, lurking, and feedback is always welcome.

I do think one part is worth adding clarity on:

You don’t have to pay to access data that lives on your own device on Anytype. You can continue to use Anytype as long as you want without paying. Once you reach your free storage limit, files stop syncing but you can continue using the app (create objects, make edits, etc.) and you never lose access to your data. Obsidian’s model is that you pay for sync, which is effectively very similar—except their default is you get no free sync at all. Not trying to convince you of anything, I just want to ensure that was clear.

Additionally, we have no plans today to remove local-only or self-hosting.

ThatGuy · June 29, 2026, 3:38pm

A very well-written response - thank you. I’ve seen what I’d characterise as the “purist challenge” in many forms down the years. It always has the same general shape, which is a claim that the devs have departed from an ideal path as defined by the purist and that catering for (pandering to?) a wider range of user needs is selling out.

This is business, not charity or religion. The Any team has to find a way of making this product and the associated services economically viable while at the same time protecting some core principles. That means compromises. By all means question whether they’re the best compromises, but remember that my best isn’t necessarily the same as your best.

Think about it from the average user: they like Notion as a product but don’t like its lack of privacy, security, proper offline mode, etc. Many users come to Anytype to solve these specific issues. They dowant to have their notes sync between their devices using the internet and not manage their own backups. They don’t want the app developers to hold keys to their accounts.

The problem the majority of users are looking to solve isn’t: ‘not use any server ever’.

I think you need to keep saying this - I’ve been with Anytype from the start, and this debate doesn’t die

ThatGuy · June 29, 2026, 3:41pm

The correct title for this thread is:

Privacy, sovereignty and direction: how I interpret the Anytype code, the Balderton thread and the governance structure

As replies demonstrate, you’re presenting opinions, not facts.

Zak-from-Zork · June 29, 2026, 4:33pm

My intention was indeed not to have a paywall to throw people out of their own projects.

The idea for ‘local only’ or ‘on premise’ would be, that you can use it as long as you like on your own risk. However, to get updates and security patches you would pay for the service.

I think this scenario is getting more and more important. Think of absolutistic countries where democratic spirits could organise resistance.

Signal is used widely in such circumstances, however the need for intenet-connection may be a disadvantage.

Anytype could solve such security threads with the any-protocol and could add even more value to collaboration than pure chat.

Humain · June 30, 2026, 8:27pm

I think this conversation should be split off to start a discussion about the future of the Local-only and On-premise modes.

Personally, I think offering paid access to the Local-only mode would make sense if it were offered without a subscription. For example, free access to updates for one year; after that, a new license would be required if the user wants the latest update plus access for the following 12 months.
This would help achieve the goal of sustaining the business model and would remain fair and equitable. No holding users hostage.

But it would be hard to sell this idea to those who have been using it for free, so we’d have to require it for new users and offer lifetime “Local-only” access to current users; otherwise, there’s a risk of a backlash.

Looking forward to your feedback.

fleurcode · July 1, 2026, 6:38pm

This is good, strong and very informative. The titles do sound like AI sometimes which is funny to me, I guess this says more about the amount of reading I’ve done on Gemini than anything actually relevant here. Great job on the detailing and the gigantic amount of comprehensive information. Knowing the Encryption works already puts me at ease and makes me want to use Anytype further, even if I would love to see more privacy.

kaye · August 6, 2026, 2:04pm

Apologies, I was meant to respond but it slipped through my radar. Yes, if we want to really make local-only work, then we’d have to revisit our approach to the business model more fundamentally. Your suggestion is what I believe Sketch and few other apps did, which seems fair, but it’s fallen out of fashion.

There are a lot of possibilities with any-sync, indeed. Anytype is just one manifestation of the software that can be built. Hopefully more local-first and sovereignty forward software succeeds in the future.

FYI — We have updated our docs, specifically the section on local-only (among many others), to hopefully more accurately inform users that Anytype is built to be local-first, and local-only is not a mode we prioritise or encourage per se. We explain our position on the option clearly, even though I understand many local-only users may disagree.

Happy to hear any feedback or suggestions on any docs/materials.