Powerless by Design
Video
Watch on YouTubeTalk description
Chetan explores how opt-ins, cookies, and mandatory accounts cost users control, and asks how we can design products that truly prioritize user benefit over data extraction.
Session Summary
Chetan traces his personal move off Google into the architecture of the surveillance economy that runs the modern web. Google's $100bn quarter, the $20bn Apple receives for default search, 85% of Mozilla's income, 71% of all web traffic tagged with a Google tracker. The genuinely shocking demo lands halfway through: Meta's Pixel exploiting a localhost server inside the Facebook app to defeat incognito mode entirely. Cambridge Analytica, Emerdata, the AGI race framed as a race for controlling God. Then the longest, most useful practical-alternative catalogue at the conference: Brave, Kagi, Signal, Proton, Ollama, Ente, OpenStreetMap. You do have influence.
View detailed generated session topics, quotes and video timestamps
From Google fanboy to suspicious user (3m05s)
Chetan opens with his Android-developer beginnings, his love of Google Inbox, and the moment Google shut it down. The pattern of Google killing products it had encouraged him to depend on — and Google Takeout's terrifying complete archive of his search history and location data — made him start to question whether being a Google user was a good idea.
"Inbox was like my favourite thing — they basically took email and just simplified it, and added a few little features just to make it work really well, and just took out a little of the craft"
"I started seeing this pattern of Google coming out with these products that seemed really great, and they seemed to be growing and gaining momentum, and then they would shut them down"
"it just felt a bit creepy — it's just too much stuff all in one place that knows about me and just knows too much"
Where Google's money comes from (5m39s)
Google's last quarter alone broke $100bn in revenue. Their products are free to me — so where is all that coming from? Shareholders, basically. The brief history: BackRub at Stanford → PageRank as the ranking algorithm → IPO at $23bn → an entire web economy structured to feed the search-and-ads loop.
"Google's revenue in the last three months — in the last quarter — has just passed a record-breaking $100 billion, and it didn't cost me a thing to use most of their products"
"the more popular your site is, the more likely you are to go to the top — and as you get more traffic, you can find ways to monetise that traffic, and it's usually ads"
"where is all their money coming from, and how do I fit into this?"
AMP, default search, and the appearance of choice (7m09s)
AMP — Accelerated Mobile Pages — was Google's leverage on content owners; sites hosted on google.com lost most of their monetisation. Google pays Apple $20bn a year to be the default search engine on iOS and pays Mozilla $400m (85% of Mozilla's income). A US judge ruled this isn't a monopoly because users can change the default. Most don't.
"Google pays Apple more than $20 billion a year just to be that default search engine"
"[Mozilla] actually wouldn't exist if it wasn't for that"
"it's the default option and a lot of people aren't gonna change that — and I think that's a monopoly"
Ads, AdSense, Analytics, and the price of free (9m16s)
About 57% of Google's revenue comes from search-driven ads. The advertiser-side flow (bid on keywords and targeting) produces the I want to buy a razor search returning two pages of sponsored links before any organic results. AdSense gives every small site a slice of that, at the cost of more and more ads. Google Analytics drops a cross-site cookie that lets Google reconstruct the user's journey across thousands of sites.
"you get basically two pages of sponsored links and stuff before you even get to a search result"
"every site that you are using has a tracker on them that is basically sending information back to Meta, and that's how they're making money from you"
"if someone else also has Google Analytics, the same cookie will basically get sent to Google Analytics, so that you can basically be tracked across sites"
Gmail, Maps, Android and the Widevine catch (13m28s)
Every Google product is a data-collection point — Gmail (your contacts, purchases, appointments), Maps (your location history, now stored on-device only because the government subpoenas got too frequent), Android (open-source AOSP, but the version everyone has includes Google Play Services calling home). And Widevine — the DRM blob inside Chrome — is the chokepoint for Spotify, Netflix, BBC iPlayer, HBO and Udemy, with closed source and unknowable telemetry.
"Google Play Services is basically calling home to Google all the time"
"Widevine is used by Spotify, Netflix, HBO, Udemy, BBC iPlayer — every time you play something, it's calling back to Google to say 'am I allowed to play this thing?'"
"the actual source code of what they're providing is becoming more closed — I don't really think that's in the spirit of open source"
Meta and the localhost-port pixel exploit (18m38s)
The Meta Pixel is Google Analytics with Facebook ad-targeting. Meta's revenue is now two-thirds of Google's and growing twice as fast. Last year researchers discovered a server running on local-host inside the Android Facebook and Instagram apps that the Meta Pixel could connect to from your browser — making incognito mode useless. The same trick was used by Russia's Yandex from 2017 onwards. Both have stopped; the EU/UK/US fine estimate is up to $23bn.
"they took advantage of an exploit where they could essentially send a request to that server"
"even if you open an incognito window, it's still managing to basically track you across your non-incognito windows"
"on the top 100,000 websites on the internet, 78% of sites had a Meta Pixel that would call back to that local-host tracker"
Trackers everywhere, X, and Cambridge Analytica (22m48s)
AdGuard estimates 20% of all web requests are for trackers. Google trackers are on 71% of all web traffic and 99% of the top 10,000 sites. The Great Hack documents how Cambridge Analytica used the resulting microtargeting to swing Brexit and the 2016 US election. Carole Cadwalladr was sued and threatened with death. Cambridge Analytica's founder set up Emerdata under a different name and is presumed to be doing the same thing.
"20% of all of the web requests are just for tracking information — that's a lot of bandwidth, that's a lot of CPU, it's a lot of energy"
"Carole Cadwalladr... was sued. She was sent death threats. Tech companies didn't really, not much happened to them — they basically carry on as normal"
"this is how our democracy works now — kind of fun, isn't it?"
AI, AGI, and the political feedback loop (29m26s)
AI is now in everything, and people use LLMs instead of visiting websites, so the entire ad-supported web model is being undermined. OpenAI doesn't expect profitability before 2030. The underlying race is for AGI — a race for controlling God. And in the US, Trump is visibly close to the people running these companies, in a feedback loop where the most powerful are making themselves more powerful and the checks and balances are being removed.
"the underlying race for AGI — a system that is basically as smart as a human, or potentially smarter — it's almost like a race for controlling God"
"most of these people running these companies are white men with bias — some of them with a lot more bias than others"
"it's a feedback loop where the most powerful people have found a way to make themselves even more powerful, that we just can't do anything about it, it feels like"
What you can do as a user (34m36s)
Say no to permission pop-ups. Disable autoplay everywhere. Use email (TLDR, Pointer). Read your news through aggregators that surface bias (Ground News, Kagi News, The Citizens, The Nerve — both founded by Carole Cadwalladr). Use ad-blocking browsers (Brave, Firefox, Orion, the upcoming Ladybird). Use DuckDuckGo or pay for Kagi or Mojeek. Move chats to Signal. RCS now matches iMessage/WhatsApp features over a standard. Use Fastmail or Proton for email. Use OpenStreetMap. Use Ente for photos.
"you have some sway over those people, and that's how these things start"
"pull is better than push... turn off autoplay. Autoplay is on by default on Netflix, on YouTube, on loads of other stuff. Just turn it off"
"do not rely on social media for your news, because you will be fed this kind of processed-food diet of content. And the only people benefiting from those is shareholders"
What you can do as a developer (47m26s)
Run your LLMs locally via Hugging Face and Ollama. If you ship apps, don't design them to require always-on permissions. Contribute to open source — money where you can, time and code where you can't. Pay for Wikipedia. The rule of thumb for products and services: a clear business model, privacy-first, and open standards so you can leave.
"you don't have to use a web-based LLM that you're paying a monthly subscription for. You can go to Hugging Face and download an open-weight model"
"if you run a company and you control the budgets... put aside a budget to say 'you know what, we use a lot of this stuff, we're gonna contribute back'"
"if privacy doesn't come first, then basically the business model is, I think, not that clear"
About Chetan Padia
I wanted to play computer games like my friends were playing but my parents put a computer with BASIC in front of me and told me to make my own.