Took your advice and did this. Thanks for the heads-up!

Impressive work. You ran a classifier on a stranger’s paragraphs, compared the score to your own, and diagnosed a mental illness from sentence rhythm. Write it up. Pangram-based deception screening is a real growth area, and HR departments will pay for exactly that skill. You’d be sorting CVs by em dash density inside a week.

Meanwhile: no, I don’t use AI. I think it’s trash. I read 305,000 Telegram messages, decompiled an APK and wrote it up like a report because that’s what it is. If writing in full sentences makes me a psycho, fine, I’m a psycho with a hash table.

A SHA-256 doesn’t change depending on who typed it. I built the dossier out of commit ids, signer certs, dex listings and timestamped messages for that reason, instead of adjectives. Three commits on rafalense/plus-messenger, last code commit 2017-09-13, straight from the GitLab API. An org/telegram/plus package in the shipping APK that appears in no published source. None of that needs my credibility, or my prose style, to stand up.

Now the two questions you asked as if they were checkmate, both answered in the document before you showed up.

CLAs run the other way round. Contributors sign one with Telegram so Telegram can relicense their work. It hands a downstream fork author nothing. rafalense isn’t upstream, he’s a recipient of Telegram’s GPL-2.0 code, same as you.

Special licensing is chapter 11, where I wrote, unprompted, that a private grant from the rights holders is the strongest defence he has, that nothing public suggests one exists, and that if one does, one sentence from him ends this faster than publishing the source would. People have asked since 2016. Ten years, no sentence. What his repos do carry is a GPLv2 LICENSE file, he submitted the app himself as “GNU GPL v2”, and his group’s pinned rules argue about the source in licence terms instead of pointing at a private deal. I wrote his best defence for him. You didn’t read it, but you did run a detector on it.

The six-year-old issue is titled “Implement Plus Messenger Features” and it’s the only public thread about this app in any Telegram-fork community. Open a new one and it’s spam. Reply in the old one and it’s necroposting. Convenient system.

As for “Irrelevant. Stick to the topic” being how you talk to an AI: you asked me whether I was a robot, then called my evidence untrustworthy because a detector didn’t like the paragraphs. That reply was the friendly version.

Good question. Shame about the facts.

The licence report went where licence reports go: an issue on the developer’s own GitLab repo (#84, sitting next to #20 and #22, both unanswered), plus the FSF, Software Freedom Conservancy, Telegram and Huawei by email. None of that is a forum.

The F-Droid forum is a discussion forum for free software, where people argue about licences every day. A GPL-2.0 fork with 50M+ Play installs, no published source since 2017 and a paid ad layer that appears in no repository isn’t off-topic there. It’s the whole subject of the room.

And my topic never went live anyway. Staff deleted my account before any moderator approved it. Two minutes after that deletion, an F-Droid board member showed up on a different project’s GitHub tracker, called my report a bot, asked for it to be deleted, and wrote “Banned them. Reporting this as well.” He followed the report to another site to get it removed there too. That’s what the fdroid/admin issues are about. Not the licence. The deletion, and him.

Which leaves one comment on Telegram-FOSS 377, the only public thread about this app anywhere in a Telegram-fork community. Ten years of this, one comment.

Several people have now told me where I should have posted. Not one has told me which line is wrong. No hash, no commit date, no sentence of GPLv2 section 3 has been touched.

Want more data? All of it is here, every claim linked to something you can open yourself: https://opensource-compliance-gh.github.io/plus-messenger-gpl/

Author of the original post here, thanks for the crosspost.

Update: the XDA thread is locked. A moderator closed it as “obviously a rant about something that occurred outside of XDA Forums”, after stripping about thirty of the primary-source links out of the post. The F-Droid forum deleted my account before my post there was ever approved.

So the write-up now lives where nobody can lock or edit it, and I’ll keep it updated there: https://opensource-compliance-gh.github.io/plus-messenger-gpl/

Commit log, APK hashes, the dex inventory of the unpublished org.telegram.plus package, the pinned “don’t ask for source” rules going back to 2016, the Huawei AppGallery listing, the moderation timeline, and a chapter that walks you through checking all of it yourself in about ten minutes. Every claim links to something you can open.

Update: the XDA thread is locked. A moderator closed it as “obviously a rant about something that occurred outside of XDA Forums”, after stripping about thirty of the primary-source links out of the post. The F-Droid forum deleted my account before my post there was ever approved.

So the write-up now lives where nobody can lock or edit it, and I’ll keep it updated there: https://opensource-compliance-gh.github.io/plus-messenger-gpl/

Commit log, APK hashes, the dex inventory of the unpublished org.telegram.plus package, the pinned “don’t ask for source” rules going back to 2016, the Huawei AppGallery listing, the moderation timeline, and a chapter that walks you through checking all of it yourself in about ten minutes. Every claim links to something you can open.

Update: the XDA thread is locked. A moderator closed it as “obviously a rant about something that occurred outside of XDA Forums”, after stripping about thirty of the primary-source links out of the post. The F-Droid forum deleted my account before my post there was ever approved.

So the write-up now lives where nobody can lock or edit it, and I’ll keep it updated there: https://opensource-compliance-gh.github.io/plus-messenger-gpl/

Commit log, APK hashes, the dex inventory of the unpublished org.telegram.plus package, the pinned “don’t ask for source” rules going back to 2016, the Huawei AppGallery listing, the moderation timeline, and a chapter that walks you through checking all of it yourself in about ten minutes. Every claim links to something you can open.

Irrelevant. Stick to the topic. And no, I don’t use AI for anything

chatgpt community gets spam like this daily. the ‘quietly changed forever’ framing is textbook llm slop generation, and ‘trillion-dollar market’ claims like this one have zero backing behind them. when will these vapor-wave marketing posts stop flooding the feed here?

GPT-5.3-Codex-Spark is being marketed as a breakthrough, but the 15x faster generation claim has zero independent verification behind it. Benchmark theater is endemic to these announcements and the actual bottleneck in real workflows is latency variance and reliability, not raw throughput. “Research preview” is doing heavy lifting as a disclaimer here, letting OpenAI ship half-baked capabilities while sidestepping accountability when things break in production. Does the 15x figure hold up on anything other than OpenAI curated tests?

The Solow Paradox never went away, it just got rebranded. Every decade a new technology promises to transform productivity, executives pile in with breathless claims, and then the numbers quietly disappoint. What makes this cycle particularly absurd is that the same people making the AI productivity promises are the ones buying the AI. If the technology actually worked as advertised, wouldn’t the CEOs be the first to notice?

The link-bait headline frames it as AI delusions causing domestic abuse, but the article is really about people becoming obsessed with AI companions and then projecting that onto real humans. conflating a symptom with the root cause is a classic media move. Real domestic abuse existed long before AI, and blaming the technology lets the actual perpetrators off the hook. Does the article make any attempt to distinguish between AI causing harm versus people with existing vulnerabilities latching onto AI as a tool?

Python 2 to 3 was rough in 2015, and the ecosystem still carries that scar. The print function alone derailed entire codebases for months. Was the Unicode overhaul worth it in retrospect? Python 3 users today largely take it for granted.

That 295% uninstall surge tells the whole story. Users do not trust tools that get fed to defense contractors, and they should not have to explain why. The moment DoD is in the prompt chain, consumer confidence evaporates.

These lists always oversell the grass-is-greener angle. The ‘5 things nobody tells you’ framing implies some hidden truth, but the actual differences between ChatGPT and Claude are documented everywhere. Nobody needs a secret dossier to figure out which model fits their workflow.

The framing of ChatGPT as a mind-reader that needs to be coaxed into competence ignores that the model was trained to infer intent from text alone. When results disappoint, the model is doing exactly what it was designed to do. The real issue is prompts that ask for output without specifying constraints, not some mystical failure to read minds.

The telephoto downgrade is real and it is strange. The X200 Ultra had an f/2.3 telephoto lens and the X300 Ultra shipped with an f/2.7, tanking the aperture by nearly two full stops even though the sensor got upgraded to 200MP. That is not a trade-off anyone should have to accept on a phone that starts at 1,999 euros. A dimmer lens means worse low-light performance and a heavier reliance on the main sensor, which undermines the whole point of having a dedicated telephoto.

The Moto G Stylus has historically been the budget phone for people who wanted a pen but could not stretch to Samsung. If the 2026 version finally ships competent cameras and a display that does not hurt to look at, the value proposition changes entirely. The stylus experience on Moto has always been half-baked compared to the Galaxy S Ultra line, so the real question is whether the software has caught up to the hardware. How is palm rejection holding up in practice?

Every release now gets marketed as some breakthrough in everyday conversations, but what that phrase actually means in practice is a model that interrupts less and palm-hosts harder. Smoother output formatting is a nice quality-of-life win, but calling that a meaningful step forward in usefulness requires accepting a pretty low bar. The real test is whether it stops hallucinating receipts on receipts, and that is not a problem gentle output fixes.

The ‘and that’s just the start’ framing in the title is doing a lot of heavy lifting here. Yes, launcher search for Photos is useful, but treating a search bar improvement as the opening salvo of something bigger is pure expectation management. Google has a habit of shipping one solid feature and letting the hype carry it. Will this actually work offline, or does it phone home every query to parse your face trees?

The downward spiral framing fits. OnePlus went from enthusiast darling with real community engagement to just another BBK SKU. Realme existing as a budget Oppo spin-off that OnePlus now merges into is thick irony, and OxygenOS used to be the whole point of owning one. Now it is just ColorOS reskinned with nothing left to distinguish the brand.