BluckerTV: five months, one subscriber

BluckerTV was a Brazilian general purpose video on demand platform. Not a gameplay site: video infrastructure, built on PeerTube, with its own interface, its own plugins, its own delivery stack, a registered trademark, a registered company, and an app published on Google Play.

The idea came from a real loss. Videolog, a Brazilian platform that ran from 2004 to 2015, shut down and took its archive with it. BluckerTV promised the opposite: what you publish does not disappear. That was the differentiator, and as you will see, it was also the problem.

All of it was built by one person. Development, infrastructure, DevOps, product, legal, compliance, moderation, and support.

What ran underneath

PeerTube 8.2.3 pinned to the 8.2.x line, with a patches-only policy. Node 22 LTS, PostgreSQL 16, Redis 7, nginx, FFmpeg, Ubuntu Server 24.04 on a KVM4 VPS. Delivery over HLS with WebRTC peer to peer, and transcoding kept deliberately lean: two rungs only, 360p30 and 1080p60. No 1440p, no 4K, because every extra rung multiplies CPU and storage cost without multiplying the audience.

At the edge, Bunny CDN with an origin shield in Chicago and an 88.8% cache hit rate, backed by Backblaze B2 for object storage. Cloudflare on DNS, orange cloud at the apex, SSL Full Strict.

The Android app was a TWA wrapped around the PWA. iOS was evaluated and postponed on purpose, not forgotten.

Nine in-house plugins, and an architecture decision with a legal job

All in TypeScript, under the peertube-plugin-blucker-* namespace. Nothing in the PeerTube core was modified, and that was not just hygiene: PeerTube is AGPLv3, and touching the core opens the derivative work reading, which would require opening the plugin source. Doing everything through plugins defused that risk.

• blucker-copyright ran the full DMCA and trademark flow: notice, counter-notice within 10 business days, and a case dashboard. It started full of crons and automation and was later refactored to 100% manual, which deleted more than 1,500 lines of its own code. It ended with an automatic purge that anonymises the data five years after each case closes.
• blucker-monetizacao capped views per IP per video per day as an anti-fraud measure. The IP moved from plain text to a SHA-256 hash: keeps the anti-fraud function, stops retaining personal data.
• blucker-age-verification implemented Brazil’s Law 15.211/2025. The architecture was chosen carefully: the national ID never entered signup or login. The gate only fired when opening a sensitive video, once per account, stored as a hash.
• blucker-blucks was the short video feed, with its own ranking algorithm.
• blucker-ads was advertising with no cookies, no metrics, and no targeting: a static banner served from the platform’s own server, with display time bought by rotation.

Plus blucker-ganhos, blucker-voice-search, blucker-blust, and blucker-livechat, the last one discontinued along with live streaming.

The legal work, written from scratch

This is the least common part of a solo project, and the one that consumed the most time.

Terms of Use at 29,998 characters across 10 sections, compressed down from roughly 83,000 in the first draft, with every legal citation audited against an official source. A Privacy Policy of 9,988 characters, mapping the legal bases and data subject rights of Brazil’s data protection law.

The most interesting finding was an error in the text itself: the clause on liability for third party content still described the old regime of Brazil’s internet framework law. The Supreme Court had declared that provision partly unconstitutional, with the ruling handed down in June 2026. The text was legally out of date before the site even opened, and was rewritten to follow the interpretation the Court had fixed.

The trademark was filed in November 2024, class 41, and granted in June 2026, unopposed. It is held by a natural person, so it survived the company’s closure.

The numbers

Public snapshot from July 2026, taken from the platform’s own transparency page:

Metric	July 2026
Users	72
Videos	270
Views	3,361
Hosted	1.8 TB
Paying subscribers	1

The Blucker+ plan cost R$ 6.90 a month and raised the storage quota from 50 GB to 1 TB. A single subscriber signed up, on 31 May. The final balance in the payment gateway account, at shutdown, was R$ 6.18.

On the other side, total cost landed around R$ 8,100: roughly R$ 4,250 in development tooling, R$ 2,850 in server and domains, about 40 dollars a month in CDN and storage, plus a registered address and taxes.

How that spending clustered tells its own story: April alone was 75% of all tooling spend, and from April to July it dropped 96%. The build-out peak and the cooling off show up in the bank statement before they show up anywhere else.

Why it ended

It was not a technical failure. The platform was live, working, compliant, and had a published app. It ended on arithmetic.

The maths did not work. With marketing projected at around R$ 1,045 a month, it would take 145 subscriptions at R$ 6.90 just to break even. A single cancellation put the operation in the red.

And the differentiator was a liability. “What you publish does not disappear” plus a generous quota means paying for storage forever, for people who already cancelled. Even at 200 or 300 subscribers, a wave of cancellations would leave the operator paying to host the entire archive with no revenue at all. The promise that sold the product was the same one that sank it long term. That is why the unlimited quota idea died before it shipped.

The central question had no answer. There was never a good answer to “why would anyone use this instead of YouTube”. The original use case, archiving my own videos, had already been abandoned before that.

The risk was inverted. The platform was funded by revenue from the YouTube channel. So the project that existed as a hedge against demonetisation was consuming exactly the income it was meant to protect. It amplified the risk instead of spreading it.

The build order was wrong. Infrastructure, compliance, and trademark came before demand was validated. By the time marketing came up, there were five months of fixed cost on the board and no market signal at all.

What stopping early saved

The decision came before spending the R$ 500 to R$ 1,000 a month budgeted for ads. Over a year that would have added R$ 6,000 to R$ 12,000, with the same outcome. Stopping in August cost far less than stopping in December would have.

The shutdown was executed in full, not abandoned: company formally dissolved with a certificate, final tax declaration filed, bank account closed, registered address cancelled in writing, cards removed from every vendor, and auto-renewal turned off everywhere. The trademark was kept.

Real net loss came to about R$ 5,500, after discounting the server still usable until March 2027 and the share of taxes that went to social security.

What it left behind

Production incidents resolved, which is where the real learning happens:

• A database backup was left publicly reachable through the CDN. Found and removed.
• Broken DKIM selectors destroyed email deliverability for the whole domain and led to a Spamhaus listing. Fixed by realigning SPF, DKIM, and DMARC, ending at 9.5 out of 10 on mail-tester.
• An npm install fired in the wrong directory almost took the installation with it. Recovered from backup, with zero downtime.
• A storage lifecycle rule was left at 1 day after a test and stayed that way for over a month. Found and corrected, with the bonus discovery that the vendor’s web panel silently overwrites a field it does not even display.

Contributions to PeerTube, under the handle @lipfelipef: two issues resolved, five opened, and one pull request. One of them, about the channel statistics window, was picked up by another contributor with the maintainer’s blessing and became a full feature in the project, with a period selector and weekly and monthly aggregation.

A security report to Framasoft, PeerTube’s maintainer: an account verification flaw in version 8.2.3, medium severity at CVSS 4.3, sent through the official security channel. The technical details are left out of this text on purpose, because other instances run the same version.

The honest summary

Five months, a whole platform in production, nine plugins, a granted trademark, a company opened and closed properly, a published app, an accepted security report, and a contribution merged into a serious open source project.

And one subscriber.

Both halves are true at the same time, and that is exactly why this text exists. The technical half taught infrastructure, video, licensing, and regulation. The business half taught something harder: building it well is not the same as building something anyone wants, and the time to find that out is before the first invoice, not after the fifth.