---
title: "22 Months of PanicRadar: the Unit Economics of a Telegram Product — 32% Off Every Payment, Platform Risk, and a Moral Fee in the Price"
description: "PanicRadar survived 22 months in a messenger where panic is manufactured on an enemy budget, purged by the millions of channels, hidden behind a privacy policy, and resold as traffic. A breakdown of the ethical unit economics of a product whose MRR has geography, a platform, and political risk — and a named framework: the 'moral commission,' the cut taken not by Apple but by the mirror."
author: "Дністер"
published: 2026-10-05T03:01:49.000Z
language: en
url: https://neurodrift.org/en/blog/panicradar-22-misiatsi-telegram-monetization-2026/
tags: ["startup", "monetization", "Telegram", "product", "ethics"]
---
# 22 Months of PanicRadar: the Unit Economics of a Telegram Product — 32% Off Every Payment, Platform Risk, and a Moral Fee in the Price

<p>At 1:17 a.m. any dashboard looks innocent. It does not cough, does not smell of smoke, does not ask which country the payment came from. Just a green line, a tidy number, a small dopamine candy for a founder who should long have been asleep but instead feeds his digital Tamagotchi.</p>

<p>PanicRadar has lived 22 months. Long enough to stop calling it an "experiment," not long enough to lie to myself with the word "business." Somewhere mid-corridor I realized the hardest question is not "how do I raise MRR?" The hardest question is <strong>whose MRR did I just raise</strong>.</p>

<p>Because PanicRadar does not sell a "feature." It sells a small prosthesis of calm in a messenger where calm is broken, sold, banned, re-uploaded, and forwarded to a backup channel every single day. And I am a Ukrainian developer who placed a panic radar where part of the panic is produced systematically, part is sold as traffic, part is purged by moderators by the million, and the rest I see as a pretty line in revenue. This is not a moral dilemma from a textbook. It is the "Revenue" button at 1:17 a.m., when the child is asleep, the wife is asleep, the country is not asleep, and the Accountant inside me has already opened Excel and whispers: "don't ruin my month."</p>

<h2>The machine that counted bakers and the condemned with equal precision</h2>

<p>In June 1937, Thomas J. Watson Sr., chairman of IBM, accepts in Berlin a medal the Nazi regime created for foreigners who had "made themselves deserving" of the German Reich — one of the most prestigious decorations the Reich gave to non-citizens. He is smiling. He will return the order in 1940 — not when he feels shame, but when keeping it starts costing more than the honor it represented.</p>

<p>Behind that smile sits a business model we now dress in pleasant language. IBM did not <em>sell</em> Nazi Germany its Hollerith tabulating machines. It <strong>leased</strong> them — through its German subsidiary Dehomag it remained the sole source of millions of punch cards, without which the hardware was dead. The USHMM records the 1933 census running on Dehomag equipment: the machine counted Germans, and along the way let them be sliced by religion, profession, and city. What exactly was counted afterward is no longer an indisputable inventory ledger but Edwin Black's argument, and the honest move is to present it as a thesis, not as a receipt. What matters is this: lease plus service plus consumables that never run out is SaaS, sixty years before anyone said "SaaS." Recurring revenue with perfect retention: the customer never churns, because the customer is a regime, and regimes have no "cancel subscription" button.</p>

<p>The machine counted with equal precision the number of loaves in a Dresden bakery and the number of people in a railcar. Because a tool does not ask who stands on the other side of the payment. That is not the work of the scalpel — it is the work of the hand holding it. And somewhere between "this is just accounting" and "this is already a list of people" opens a zone where the dangerous one is not the fanatic. The dangerous one is the Accountant: efficient, tidy, an early riser. The question that picture puts to me every morning is worse than "are you the same?" It is: <strong>at what moment does the person behind the tool start hiding behind the tool?</strong></p>

<h2>The Internal Accountant</h2>

<p>Every founder carries a quiet voice. I never hired mine, but he is my most stable co-founder. He is not evil: evil people are often inefficient, and he is efficient. His motto is embroidered in gold on the inner wall of my skull: <strong>"the tool doesn't ask who pays."</strong> He says it so gently you would think it was not a moral capitulation but a hygiene factor. Money has no smell. The funnel has no citizenship. Churn knows no geopolitics. Remember that phrase — we will meet it twice more, and each time it will weigh less, like a banknote passed through too many hands.</p>

<p>There is one profession that reads any table faster than any P&amp;L. The waiter. He walks into the room and in three seconds sees who at the table is the adult: not the one laughing loudest, but the one who gets the bill at the end. The rest touch the glasses, say thank you, leave crumbs on the linen — but one person always pays, and the waiter knows his face before dessert. For twenty-two months I carried the bill to my green dashboard and did not always raise my eyes to see who was sitting at it.</p>

<aside class="pullquote">The Accountant does not lie. He simply counts so neatly that people disappear between the columns.</aside>

<h2>The funnel that does not check passports</h2>

<p>Let us begin with a confession in numbers, because without it the rest is poetry. Classic mobile economics already resembles a small morgue with push notifications. Per Quettra data that Andrew Chen analyzed across 125 million phones, the average app loses 77% of daily active users in the first three days, 90% within 30, and over 95% within 90. It is an old benchmark and not Telegram-specific, but as an x-ray it still works: most users do not "leave." They do not even live long enough to leave dramatically. They evaporate. (For mini-apps Monetag cites an even more exotic frame — MAU ≈ 5% of DAU; hold it with tweezers: an inverted metric from an ad network that earns off every user it funnels to you — a waiter praising the house wine by the glass.)</p>

<p>Now imagine you have walked a user through all that hell — they survived, activated, pulled out a wallet. And here begins the second funnel. Every mobile payment loses about 32%: 30% to Apple and Google for the honor of pressing a button on their glass, plus a 2–3% Fragment spread converting Stars to TON. From a hypothetical ten dollars, roughly $6.80 reaches me — and not before a mandatory 21-day hold. Apple and Google do not ask whether your product is moral. They take 30% even off conscience. Especially off conscience — it haggles badly.</p>

<p>Here is the honest anatomy. Not a growth curve — a cliff with several ledges your LTV catches on the way down:</p>

<table class="data-table">
<thead>
<tr><th>Funnel stage</th><th>What happens</th><th>How much remains</th></tr>
</thead>
<tbody>
<tr><td>Install / first touch</td><td>Arrived via bot, link, referral</td><td>100%</td></tr>
<tr><td>Day 3</td><td>"Tried — closed." Classic mobile.</td><td>~23% (77% already gone)</td></tr>
<tr><td>Day 7 (activation)</td><td>"Good" retention for a quality project</td><td>~30% of those who survived</td></tr>
<tr><td>Payment</td><td>Pulled out wallet, hit "pay"</td><td>low single digits %</td></tr>
<tr><td>Minus commission</td><td>−30% Apple/Google, −2–3% Fragment</td><td>$6.80 of $10</td></tr>
<tr><td>Payout</td><td>21-day hold, then TON</td><td>money "exists" — just not in your hands</td></tr>
</tbody>
</table>

<p>"Good" retention is when seven out of ten have fled within a week and you no longer call the police, because in mobile that counts as a wedding. Look at that table through the eyes of someone paying for traffic: with a CAC around $1,200 per customer (the B2B SaaS benchmark) and this kind of leakage, paid traffic in Telegram is not marketing. It is slow suicide with quarterly reporting. The only math that works for a solo operator is organic and referral (≈$150 vs. $1,200). 22 months mean one thing: I did not buy users. Sometimes poverty saves you from sin more reliably than ethics — the most honest founder in the market is the one who simply did not have the budget to sin.</p>

![A transparent funnel with users cascading down and a single coin dripping from the exit after 21 days; nearby, scales where the CAC pan vastly outweighs the nearly empty LTV pan; in the distance a graveyard of dead bots.](./images/inline-2-voronka-cac-ltv-annotated.png)

*The funnel without poetry: a million at the top, one coin at the exit after 21 days. The CAC vs. LTV scales are tipped so far that paid traffic is suicide with quarterly reporting.*

<p>But Telegram has more than these visible commissions. There is the Apple/Google take rate. There is the Fragment route. There is platform risk. And there is one more, the worst. Call it the <strong>moral commission</strong>. Apple takes 30%. Fragment takes a spread. The platform takes the rules. And the moral commission takes what does not appear in the payout report: the ability to look at your own dashboard without hiring an internal translator who relabels every uncomfortable payment as "just the market."</p>

<p><mark style="background:#ffe600;color:#0a0a0a;padding:0 .15em;">A Telegram product has more than CAC, LTV, and a take rate. It also has a moral commission: the share of revenue you pay in your ability to look at your own dashboard without an internal translator.</mark></p>

<p>This commission is not deducted in Stripe, Stars, or Excel. It is deducted differently: you start calling an uncomfortable payer "the market," geographic risk "TAM," and your own silence "a neutral position."</p>

<aside class="pullquote">The Accountant does not sell the soul. He simply amortizes it over the months so it looks like a normal operating expense.</aside>

<h2>The platform as a rented kitchen</h2>

<p>Telegram should not be demonized. That would be too convenient, and convenience in topics like this is a cheap deodorant for filth. Telegram is not the devil. It is infrastructure with brilliant distribution, low friction, criminal noise, political pressure, crypto accounting, and millions of people who just want the button to work. You can cook an honest dinner in that house. You just should not call a rented kitchen your fortress. The problem is not that you build on a swamp. The problem is that after six months you start calling the swamp "strategic moisture."</p>

<p>Per the FT (based on unaudited figures), in H1 2025 Telegram earned $870M, up 65% year-on-year, but recorded a net loss of over $220M: around $300M of revenue came from exclusivity deals around Toncoin, and the loss came from writing down inventory of that same Toncoin, which fell nearly 70% in a year. The planet's leading "messenger" earns more from its own crypto than from advertising — and drowns on it. This does not mean Telegram collapses tomorrow. It means your "independent indie kingdom" stands on someone else's land, where the landlord also sometimes walks into the bank with a very serious face. In startups, dependence on someone else's platform is called distribution. In the old language it was called "living in someone else's house and renovating the kitchen very beautifully."</p>

<p>And the landlord swore eternal inviolability yesterday. After Pavel Durov's arrest in France in August 2024, Telegram updated its privacy policy: on a valid legal request in criminal cases the platform now hands over a user's IP address and phone number — earlier the policy covered a far narrower set, terror suspects — and reports this quarterly. The Freedom of the Press Foundation counts the delta plainly: in Q1 2025 Telegram provided authorities with data on 22,777 users versus 5,826 in the same period a year earlier. "Privacy as a foundation" — the ecosystem's main marketing asset — turned out to be an asset a regulator can resell without a line in your financial model. A foundation that changes shape on the first order is called, in construction, a swamp. In startups — a "platform opportunity."</p>

<h2>A pharmacy in the middle of a drug market</h2>

<p>Telegram's other problem is not just privacy. It is the trust environment. You can sell an honest product, but the user walks in with the face of someone who, in this stairwell, has already been robbed three times, invited twice into a crypto club, and once asked to pay for delivery of a parcel that does not exist.</p>

<p>According to Check Point, in 2025 Telegram blocked <strong>over 43.5 million channels and groups</strong>; daily purges in 2026 rose from a historical 10–30 thousand to 80–140 thousand, with peaks above 500 thousand per day. This is not "a bit of scam on the platform." It is an entire weather system. You are not just selling a service — you are selling a "pay" button in a city where every second button leads to a basement with a man calling himself Elon Musk and asking for USDT. My real competitor is not another radar. My competitor is the atmosphere of total distrust in which any "pay" button triggers a reflex of "this is a scam." I pay a tax on the reputation of an environment I was never invited into.</p>

<p>And this is not just Ukrainian neurosis. In June 2026, Reuters described an Indian government report: India is Telegram's largest market by users (over 150 million); the government is "proactively monitoring" Telegram groups; since 2023 it has received over 688,000 complaints about cyber fraud via Telegram, with estimated losses of around $750 million. This is no longer "the dark corners of the internet." It is a shopping mall whose administration carries out bags of forged documents every day — and still leases the floor space to new tenants. The reputation tax of the environment is paid by every honest product-builder. Even if you never poisoned a guest, they still sniff the fork at your place.</p>

<h2>Panic as a product</h2>

<p>And here the irony of my product's name becomes unbearable. PanicRadar. A panic radar. And if you look honestly, Telegram does not merely contain panic. Often it is its production line.</p>

<p>A joint investigation by DFRLab (Atlantic Council) and OpenMinds, published on July 16, 2025, assembled a corpus of 2.9M comments across 110 Telegram channels; 96,684 of them tied to <strong>3,634 bot-like accounts</strong>. After expanding to 3.37M comments and mapping 144 more channels, researchers counted <strong>316,790 bot comments</strong> in channels targeting Ukraine's occupied territories. Separate SBU operations describe the same handwriting: fake accounts posing as Ukrainians — by investigators' estimates, around 3,000 new profiles a month — sending anonymous messages about alleged "bomb threats" at facilities across the country. The goal is stated plainly: "to provoke panic among the population and destabilize the socio-political situation."</p>

<p>So panic in Telegram is not an environmental bug. It is a product. Someone has their own dashboard, their own KPIs, their own manager reporting on the number of bomb-threat messages generated — and perhaps a quarterly plan they over-deliver on. I sell a storm radar on a sea that someone is industrially storming on another country's state budget. And the worst part: that someone is my indirect <em>demand driver</em>. The more panic, the more needed the radar. It turns out the troll factory and I share the same sales funnel — just from opposite ends. The Accountant opened his mouth to say "product-market fit." I asked him to shut up.</p>

<aside class="pullquote">If panic is manufactured industrially and you sell a radar against panic, your marketing already has a partner. The partner just sits in a very dark office, on a bad chair, drawing a state salary.</aside>

<h2>The payer is not the country</h2>

<p>And now the sharpest part. In my product, the bill is paid by someone too. The problem is not that I do not know the amount. The problem is that for too long I did not look at the payer's face. In my own analytics there is a segment the Accountant calls "the warmest market," and that I increasingly read as "a line with a political temperature." I am not posting exact sums and countries here — not because they are shy, but because private data does not go on the table like fish at a bazaar. But the order of magnitude is enough for the question to stop being theoretical.</p>

<p>The simplest option is to say: cut that segment off. There is moral cleanliness in it, but also crude surgery without an MRI. A "Russian user" is a bad atom. Inside it are different nuclei: a state structure, a retail trader, the diaspora, a person with a Russian number in Berlin, a person with a Russian card in Tel Aviv, a person who hates their own state, and a person who pays taxes to the system that teaches Telegram to panic in Ukrainian. One country code makes them a single line. Conscience must unpack that line back into people and risks. The Accountant will not do it — he has no such column. I do.</p>

<p>If your largest payer cohort has a political temperature, you cannot treat it with the word "TAM." TAM has no conscience. You must. And if you do not know who your most profitable cohort is, you are not "apolitical." You are simply doing moral accounting the ostrich way: head in the analytics, rear in the geopolitics.</p>

<h2>The strongest objection: "You are signaling virtue at someone else's expense"</h2>

<p>Here the Accountant lands a blow that cannot be ignored: "You are blocking retail users your product has done nothing wrong to, for the sake of your own moral comfort. This is not ethics — it is virtue signaling paid for at the expense of others. You are not saving the world. You are wrecking your funnel and calling it ethics."</p>

<p>It is a strong objection, and that is exactly why it stays in the text. Yes, among the blocked are diaspora, anti-war Russians, people who need the radar not for propaganda but for survival in their own toxic market. Yes, a crude geo-block punishes more than the guilty. But virtue signaling is when the gesture costs nothing. Here it costs: minus projected revenue, minus a segment, minus part of the growth. And that is exactly why it is no longer a gesture but a policy — a decision with a documented price. The question is not whether you can serve everyone fairly. The question is what to do when you have no KYC, no state investigative team, no moral x-ray, and a product working in an ecosystem where panic against your country is manufactured as content. In that situation "neutrality" is also a policy. Just one the Accountant writes in invisible ink.</p>

<h2>The tool that listens without asking</h2>

<p>There is a frame more precise than any lecture on tech ethics. In Coppola's <em>The Conversation</em>, surveillance master Harry Caul technically flawlessly cleans someone else's recording of noise — and gradually realizes that his neutral, virtuosic work may lead to a murder. His tool had no morality. The problem is that he, too, tried very hard not to have any — and discovered that the absence of questions does not free the operator, it merely makes him an accomplice with clean hands.</p>

<p>Every Telegram product sits at the same console of someone else's communications and someone else's payments. The question is not whether you can technically listen, count, and monetize. The tool truly does not ask — this is the second return of the phrase, and now it sounds not like a motto but like a diagnosis. The person holding the tool asks. Or does not — like Watson, who returned the medal only when wearing it became unprofitable. Between "not asking" and "not wanting to ask" there is exactly one smile at a Berlin reception.</p>

<h2>"Why not move to a different house?"</h2>

<p>An attentive reader asks the obvious: if the house stands on a swamp and the landlord is at the pawnshop — why are you still here? Honestly: because every neighboring exit is paid for with its own price.</p>

<table class="data-table">
<thead>
<tr><th>Alternative</th><th>What it gives</th><th>What it costs</th></tr>
</thead>
<tbody>
<tr><td>Native (iOS/Android)</td><td>Your own house, your own rules, no Durov overhead</td><td>The same 30% at the door (Apple/Google) + an empty room: on Telegram the user is already at the table, in the App Store you still have to find and bring them</td></tr>
<tr><td>Web-first / PWA</td><td>Your domain, your Stripe, nobody's 30%</td><td>A restaurant without a street: zero foot traffic. The whole Telegram economy runs on the guest already being inside the messenger</td></tr>
<tr><td>Signal / WhatsApp Bot API</td><td>A "more ethical" infrastructure, better door reputation</td><td>Signal deliberately has no monetization (a principle, not an oversight); WhatsApp Business = verification + window fees + Meta policy a solo dev cannot renegotiate</td></tr>
</tbody>
</table>

<p><strong>Every exit from Telegram is paid for either with an empty room, or a higher toll at the door, or a ban on placing a till — distribution here can only be borrowed from that same croupier.</strong> The ethical indie developer does not sit on the swamp out of stupidity. He sits there because the dry land is either deserted or fenced off by a bouncer more expensive than the swamp. This does not justify choosing to stay — but it names the price honestly, without the "just go somewhere else" only ever said by people who have never brought the first thousand guests into an empty room.</p>

<h2>Four questions before the "pay" button</h2>

<p>If you are building something in Telegram (or anywhere else where the platform is not yours and the funnel is a cliff), run your product through this matrix. The Accountant steps out of the room while you fill it in — he cannot stand questions that have no answer in dollars.</p>

<table class="data-table">
<thead>
<tr><th>Question</th><th>The Accountant's dishonest answer</th><th>What to check</th></tr>
</thead>
<tbody>
<tr><td>Who pays?</td><td>"TAM"</td><td>payer cohort: phone prefix, card country, language, recurring behavior</td></tr>
<tr><td>Why do they pay?</td><td>"value proposition"</td><td>whether the demand is fed by fear, panic, war, scam</td></tr>
<tr><td>Whose land is the product on?</td><td>"Telegram is stable"</td><td>policy changes, data requests, TON/Stars dependency, payout route</td></tr>
<tr><td>What do I not want to know?</td><td>"we're small, it doesn't matter"</td><td>which payer becomes a reputational mine in 12 months</td></tr>
</tbody>
</table>

<p>If there are no answers — this is not an "early startup." It is a rented Hollerith department in beta: it is already counting, you just have not looked at what. So a geo-policy must be written not for Twitter applause, but for the night when the Accountant arrives with a new figure and says: "maybe this time we stay quiet?" Three lines on the kitchen wall that hold the place together when the other side of the wall is on fire:</p>

<ul>
<li><strong>Have a second exit from the room.</strong> A backup payment channel outside Telegram (ordinary Stripe via webhook, a minimal web checkout) is not "omnichannel." It is a door through which a paying guest can enter even after Durov signs another agreement overnight and bars the main entrance.</li>
<li><strong>Who you do not serve — write it on the door, do not keep it in your head.</strong> If your product has a geopolitical nerve, put your geo-policy in the Terms of Service: not a PR placard saying "we're for all the good things," but a document with a date and a basis. An unwritten rule protects your mood. A written one protects you.</li>
<li><strong>Do not keep all your cash in the currency your landlord prints.</strong> If payouts flow through TON — part into stablecoins, part through conventional processing. That is a tip in the currency of a restaurant that may burn down along with the till.</li>
</ul>

<h2>Not waiting for my own 1940</h2>

<p>In Jonathan Glazer's <em>The Zone of Interest</em>, the family of Auschwitz commandant Rudolf Höss throws garden parties, tends roses, bathes children in the pool. Behind a high concrete wall deep in the frame — chimneys, shots, barking. The camera never shows the camp itself. Only the wall. What is most terrifying there is not that they are villains. It is that they are normal: they drink coffee under the screaming, because the screaming is sound from the other side of the wall, and the wall is the edge of their P&amp;L.</p>

<p>My dashboard can also be a wall. It shows the chart and not the face. It makes the payment green regardless of who pressed the button. It does not lie — it simply has no organ with which to feel shame. Before reaching your own "zone of interest," it helps to recall another frame: in <em>Michael Clayton</em> the hero stands in a morning field before three horses, having stepped out of his car after a night of corporate filth — and behind him his car explodes. This is not an epiphany scored by angels. It is the second when a person sees that the system has already decided everything for him — and left him exactly one instant not to be part of the mechanism. I do not want to spend that instant adding one more zero.</p>

<p>The tool truly does not ask who pays. This is the third and final return of the phrase — and now it weighs exactly what any alibi weighs: nothing. The person who issues the invoice asks. Watson returned the medal in 1940, when wearing it became politically and reputationally impossible. I do not want to wait for my own 1940 — not because I am holier. It is just that an order on the lapel pairs badly with a Ukrainian passport, night dashboards, and a product that sells a panic radar in a messenger where panic is sometimes manufactured on an enemy budget.</p>

<p>So every morning, before I open the chart, I make the gesture I did not make for twenty-two months: I raise my eyes from the bill and look at who is sitting at the table. It is the most expensive second of my day, because after it the number can no longer be rounded up without seeing the face. And if my hand hesitates for a moment over the "pay" button — the system's operator is still human. For now. That "for now" is not a discount or an indulgence — it is an expiry date I have to renew manually every morning, because it renews automatically for no one.</p>

![A founder on this side of a concrete wall calmly waters a revenue garden without looking over it; through a gap — a shared queue of faceless payers, among them the one who funds panic and a diaspora member with a suitcase — sharing the same country code.](./images/inline-1-mur-platnyk.png)

*The wall runs exactly along the edge of the dashboard — as if a UI designer laid it there. On this side you water your MRR garden; on that side, the funder of panic and the diaspora member share the same queue, both with the same country code. The wall does not stop the evil. It just separates it from the coffee.*

<aside class="sources">
<p><strong>Sources:</strong></p>
<ul>
<li>Andrew Chen (Quettra data) — "New data shows why losing 80% of your mobile users is normal" (77% in 3 days, &gt;95% in 90; an old mobile benchmark, not Telegram-specific): https://andrewchen.com/new-data-shows-why-losing-80-of-your-mobile-users-is-normal-and-that-the-best-apps-do-much-better/</li>
<li>TradingView / Financial Times — Telegram H1 2025 (revenue $870M, +65% YoY, net loss &gt;$220M, ~$300M TON exclusivity, Toncoin write-down; unaudited figures): https://www.tradingview.com/news/cointelegraph:01e9ad089094b:0-telegram-revenue-jumps-to-870m-in-h1-2025-2b-full-year-target-ft/</li>
<li>Telegram — Stars (purchased via Apple/Google, withdrawn to Toncoin via Fragment): https://telegram.org/blog/telegram-stars — core API: https://core.telegram.org/api/stars</li>
<li>The Block — Telegram Stars launch (Durov on the 30% Apple/Google cut): https://www.theblock.co/post/298954/telegram-launches-digital-mini-app-payment-system-telegram-stars</li>
<li>The Verge — Telegram privacy policy change (IP/phone handed over on a valid order in criminal cases from 09/2024; previously terror suspects; quarterly reports): https://www.theverge.com/2024/9/23/24252276/telegram-disclose-user-data-legal-requests-criminal-activity</li>
<li>Freedom of the Press Foundation — Telegram's growing mountain of data requests (Q1 2025: data on 22,777 users vs. 5,826 a year earlier): https://freedom.press/digisec/blog/telegrams-growing-mountain-of-data-requests/</li>
<li>Check Point Research — Telegram crackdown 2025–2026 (&gt;43.5M blocked channels/groups; 80–140k takedowns/day, peaks &gt;500k): https://blog.checkpoint.com/research/telegrams-crackdown-in-2026-and-why-cyber-criminals-are-still-winning/</li>
<li>Reuters (23 Jun 2026) — India 'proactively monitoring' Telegram (largest market &gt;150M users; &gt;688k cyber-fraud complaints since 2023; estimated losses ≈ $750M): https://www.reuters.com/business/media-telecom/india-proactively-monitoring-telegram-over-concerns-about-illegal-content-2026-06-23/ — Business Standard (mirror): https://www.business-standard.com/industry/news/govt-proactively-monitoring-telegram-over-concerns-about-illegal-content-126062300703_1.html</li>
<li>DFRLab (Atlantic Council) + OpenMinds — "Digital occupation: Pro-Russian bot networks target Ukraine's occupied territories on Telegram" (2.9M comments / 96,684 from 3,634 bot-like accounts; expanded to 3.37M + 144 channels → 316,790 bot comments; published 16 Jul 2025): https://www.atlanticcouncil.org/in-depth-research-reports/report/report-russian-bot-networks-occupied-ukraine/ — https://dfrlab.org/2025/07/16/report-russian-bot-networks-occupied-ukraine/</li>
<li>The Record (Recorded Future) — Ukraine SBU busts bot farm supplying Russian spies (fake "bomb threats," ~3,000 accounts/month, panic): https://therecord.media/ukraine-sbu-busts-bot-farm-supplying-russian-spies</li>
<li>First Page Sage / Phoenix Strategy Group — B2B CAC (≈$1,200) and referral (≈$150); LTV:CAC median ~3:1: https://firstpagesage.com/reports/average-customer-acquisition-cost-cac-by-industry-b2b-edition-fc/ — https://www.phoenixstrategy.group/blog/cac-benchmarks-by-channel-2025</li>
<li>Indie Hackers / IBTimes — Marc Lou (survivorship bias; ~30 failed projects to $1M; 1–3 years to income): https://www.indiehackers.com/post/starting-up/11-solo-indie-hackers-making-1m-in-annual-revenue-NRq6hCm3La6N6UliFRfE — https://www.ibtimes.co.uk/marc-louvion-million-dollar-success-journey-1799070</li>
<li>Wikipedia / USHMM — IBM and the Holocaust (Edwin Black; Hollerith leasing via Dehomag, 1933 census, Watson's 1937 medal → returned 1940): https://en.wikipedia.org/wiki/IBM_and_the_Holocaust — https://collections.ushmm.org/search/catalog/irn521590</li>
</ul>
</aside>
