Strategic frame

Social graph fragmentation

Investor / advisor · 2026-05-10

In 2026 the social graph is split across messengers and standalone consumer apps cannot rebuild it. This is the structural reason @appss runs cross-ecosystem-as-default rather than building a new graph from zero, and why our products live inside someone else's graph, never trying to replace it.

This is a constraint, not a problem. Most early-stage consumer-app theses fail the same way: they assume they can build a social graph from scratch. They can't, in 2026, and the failure is structural. Our entire cross-ecosystem doctrine is what you do once you accept this.

The constraint

Three forces compound:

Users don't install new apps. The cost of installing a new app vs. opening something inside a messenger they already use is high enough that consumer-side LTV/CAC almost never sustains standalone discovery products. App Store / Play Store are saturated with quality-poor inventory, and discovery in either is broken.
Users don't fill onboarding quizzes. In an era where every personal AI claims to know you already (ChatGPT remembers your chats, etc.), consumers are unwilling to teach a brand-new product about themselves from zero. Cold-start data is the price of entry, and the price keeps rising.
Users don't post content into a small new system. Reviews, check-ins, posts stay where the audience is, Instagram, Telegram, native messengers, not in a new product that invited them.

Together: the social graph and the user-context layer are split across messengers and superapps, with no neutral place to consolidate them. A standalone consumer app that depends on social-graph features is fighting a battle that nobody won in the past five years.

Evidence

Mark's first project, Placenger (10 years ago)

Open placenger.com, it grabs your location, turns nearby places into chats in messenger style, every photo / post about that place becomes a «message» from a stranger. Strong wow on first open. Real engagement on the first session.

Why it died: users wouldn't post regularly into the system. Their social graph stayed in external platforms, they had no reason to migrate the activity. Ten years later, the same shape of product still fails for the same reason.

A 2026 thought experiment from a recent advisor call

Mark and a friend were sketching «a Glovo for friend-recommended places», I want to follow Gleb's restaurant taste, get notified when he chimes in:

Google Places has the inventory and the reviews. But there is no social graph layer on top of it, and adding friends by email is friction nobody overcomes.
In Prague, migrants are on Telegram, locals are on WhatsApp, regular reviews are on Instagram, no neutral place that contains the relevant social graph.
Standalone option would require building «the giant system everyone onboards into», that is the graveyard.

This is why mini-app surfaces, Telegram, TikTok Minis (post-divestment), Discord Activities, LINE, are the rising shape. They paraphrase the constraint: don't make me install; I'm already here.

Implications for @appss

Cross-ecosystem-as-default

We don't build «the @appss app users come to». We go where the social graph already is:

Telegram (home). Live, deep.
TikTok Minis (post-divestment, 1.9B MAU), Tier 2 candidate.
Discord Activities, Tier 2.
LINE Taiwan / Thailand, Tier 3, year-2.
World, WeChat, Douyin, long-bets.

In each ecosystem, we are guests. The social graph is not ours, we don't pretend it could become ours. We surface on top.

Full multi-ecosystem doctrine in where to play next.

Apps Pro is creator-tool, not consumer-app

This is partly why the two-product framing works: we sell to creators (already motivated to build, already paying for tools), they ship consumer apps that live inside existing graphs (Telegram Mini App, Discord Activity, etc.). We never own the consumer acquisition moat. Creator + native viral mechanics of the host platform owns it. CAC of a creator >>> LTV of a creator. Model converges.

Every consumer-touchpoint must live inside an existing graph

@appss Mini App in Telegram ✓
Apps Pro Bot in Telegram ✓
A standalone iOS / Android app for creators? No, never. Creators work via web + Telegram. Sufficient.

Pattern across our AI features: parse the existing graph, don't build a new one

Wherever the product needs user context (Market Research AI personalises, Designer Studio takes brand cues, Builder reads existing repo), the principle is derive from what exists, not require new input. The product degrades to a generic alternative if no graph signal is available, the product doesn't refuse to work, but doesn't pretend to social-graph-as-data when there is none.

Apps Pro Bot is a live example

Not a separate place creators go. In their existing TG. Forwarding a Market-Research answer from one creator to another carries a footer (Research powered by Apps Pro), viral mechanic built on top of the already-existing graph between those two TG users. We don't own the graph. We ride it.

Tensions

Part of our pitch is data moat (121 production tables over 2 years). That is our internal analytics graph of creators and end-users, internal data ≠ external social graph. Tempting line to blur, important to keep separate.
Cross-ecosystem expansion is non-trivial per ecosystem. Engineering, partnerships, capital costs vary dramatically. The doctrine is right; the entry-cost model has to be honest per case (see multi-ecosystem).
We are not Pavel Durov. No monopoly distribution channel. Any consumer-discovery feature on @appss only works if it parasitises native discovery on the host platform (Telegram What's New, Stars-trending, TikTok For You, etc.).

Open questions

Which «parse the existing graph» mechanics are legal and durable on each target platform? Instagram closed its API; Telegram is largely open; Google Places has ToS constraints.
Where is the line between «riding someone else's graph» and «building our own creator-graph»? Apps Pro Academy / leaderboards / community, at what point are those «our graph»? How durable would it be without external anchors?
What changes with on-device personal AI (iOS 26 / Android equivalent)? If every user has an agent that already knows their graph from all sources, does our parsing layer still matter?