Breaking Up with Google Search

In June 2025, I recorded a conversation with Bill Fitzgerald about artificial intelligence. I ended up getting an earful from him about immorality within Big Tech and some of the aspiring oligarchs within the so-called “ed tech” realm. 

That discussion stuck with me. 

Bill drove home that AI is a form of surveillance. If you use ChatGPT or a similar LLM frequently, go into its settings and check its “memories” and see what it has retained on you. Then check the terms of service or EULA. What terms have they set around data retention?  Who can they sell or share your data with? What’s their policy around warrantless law enforcement data requests?

In some ways, that episode was a continuation of a conversation I've been having about AI and online privacy for over seven years. One of the final episodes of the podcast I recorded before moving overseas in 2019 was called Mommy, My Teacher Got Replaced by a Robot. In that interview and subsequent ones, I’ve noted the non-consensual nature of AI deployment. The major tech companies are hell bent on integrating AI into places where no one has asked for it and then retaining all the data from all the interactions you have with the models.

People aren’t asking for this. 

Specifically, when it came to AI in my day-to-day, I was bothered by the declining quality of Google's Search results and the insertion of AI slop into my queries and finally I had enough (see the example below, bleck!)

Put up or shut up, those were my choices. In June of 2025, I dropped Google Search and Chrome. I experimented with different browsers and search engines in various combinations trying to find what works best for me. My current set-up is Brave as my browser with DuckDuckGo as my default search engine. I landed on DuckDuckGo only recently (May 2026) after experimenting with some of the options described below.

A bit of context: Google’s Chrome browser is based on Chromium, which is an open source software program. Google adds a proprietary layer on top and packages it as “Chrome.” But since Chromium is open source, anyone can take that foundation and build their own browser — as Brave, Ecosia, DuckDuckGo, and many others have done.

Brave is a privacy-focused browser with some rigorous ad and tracker blocking built-in. They have a scorecard when you open the app showing the number of ads & trackers that have been blocked and how much time & data that that’s saved you.

Besides the privacy features my favorite feature of Brave is their cross device sync. Most browsers sync through an account you create, usually tied to your email address, and thus your personal info. You sign in, and the company stores (hopefully) encrypted copies of your bookmarks, history, passwords, and settings on its servers. Brave doesn’t require you to create an account. Instead, it uses a system called Sync Chain. When you create a Sync Chain, Brave generates a 25 word sync code, a set of cryptographic keys like a Bitcoin wallet, that allow you to sync across devices.

Brave also makes a search engine that I am less fond of. It’s really not great — but that’s the point. I’m not wedded or hostage to one platform or eco-system, subjected to the whims of its devs and management.

Ecosia’s offers a search engine and a Chromium based browser. You don’t need to use both, again you can mix & match. They lean into a social-capital gimmick: they pledge to plant trees based on the number of searches conducted. 

Ecosia doesn't run its own search index the way Google does. Instead, it acts as a search layer that gets most of its results from other providers. As of 2026, Ecosia says its search results come from three main sources:

• Microsoft Bing

• Google

• EUSP (European Search Perspective), a joint search index developed by Ecosia and the French search engine Qwant (more on Qwant in moment)

So it’s a do-gooder search engine that searches other search engines, providing you with the results.

Qwant is just a no frills search engine, with no AI integrations and a pledge to keep it that way. Qwant like Brave is privacy focused. It does not create detailed advertising profiles of users and does not track individuals across the web for targeted ads. It also positions itself as a European alternative to Google, an effort to break the continent’s reliance on US tech companies. To me, using Qwant feels how Google Search felt in the 2010s.

I expected some kind of drop-off in search quality when I moved away from Google, but it was the opposite. Google Search has degraded so much over the last few years that it’s nearly useless. The enshittification happened gradually, but it’s obvious to anyone who’s tried to look up something specific lately that Google Search is broken. I don’t remember who said it first, but it’s stuck with me: Google Search now basically operates in one of two modes. If it can serve you a volley of ads based on your query, it will. If it can’t, it hands you AI slop from Gemini.

This is what non-consensual AI deployment looks like. I didn’t ask for these garbage Gemini results, and I didn’t ask for my searches to be flooded with barely readable articles churned out by AI content farms. It is a death spiral for the open internet, and I want no part of it.

I want to be clear here, this isn't some sort of big moral crusade. But we have to start asking ourselves at what point do we start pushing back against the monopolistic practices of big tech and should we really trust them with all the personal, medical, and financial data we do? You’d be shocked what some people put into Chat GPT.

Meaningful regulation isn’t coming in the near term, so it’s up to us to help ourselves.

* A version of this article originally appeared in my newsletter Takes & Typos

The Feds Were Asleep at the Switch for Crypto, AI Could Be Much Worse

An annoying part about getting older is that you realize how cyclical things are and that much of life is the same hustles and hassles with new names and labels. Whatever your thoughts about cryptocurrency (I'm a skeptic but also own a small amount), it's indisputable the sector is rife with scams, rugpulls, and unregulated securities. As a result, billions were lost in crypto hacks and scams over the last two years (see table below). This isn’t counting the firms like FTX, Celsius, and Voyager that went under, wiping out an additional 200 billion dollars from retail investors and people who trusted the shadow banks called crypto exchanges.

The Securities and Exchange Commission (SEC) finally woke up from its multi-decade slumber this month and issued massive fines against several crypto exchanges and the influencers shilling for them. Retired NBA star Paul Pierce was fined 1.4 million dollars for promoting a token called Ethereum Max. EMax was a pump & dump scheme where new investors were served up as exit-liquidity for the founders. EMax currently trades at $0.00000000099 per worthless token. Kim Kardashian was fined a similar amount for touting the same token in December. Regulators have stablecoins in their sights. Crypto exchange Kraken agreed to a 30 million dollar fine and agreed to shut down their staking program (if you don’t know what crypto staking is, you’re probably better off that way). Kraken’s rival Binance is expecting a massive fine as well.

2022 was a record year for hackers in crypto - Source: Decrypt, data from Chainalysis

The handwriting was on the wall for years about crypto scams but federal authorities waited over a decade before acting. In the interim millions of people were harmed. A regulatory framework for crypto exists and has existed since before the first token came to market. We can't afford to wait fourteen or even four years for the government to set the ground rules for AI. The potential for society wide harm is incalculably larger. 

After-the-fact debates are biased toward the expansion, rather than limitation, of a practice. In 2016, the city of Dallas had to reckon with the question of whether they would allow police to kill a barricaded shooter with a remote operated drone. They decided "yes" this is an acceptable use of force. Unfortunately, they decided it retroactively, months after the police had killed their target (with a Remotec Androx Mark V A-1, manufactured by Northrup Grumman). The police department refused to release documents related to the decision to use the robot and the city absolved the chief and the entire chain of command after the incident. The police chief, David Brown, now runs the police department in Chicago and there’s now a precedent regarding the use remote operated robots to kill people.

As venture capital abandons crypto projects, opting to fund cowboy AI projects, policymakers can't be passive. The potential harm from AI in journalism, financial markets, deep-fake aided scams, and law enforcement use of force have the potential to do far more damage than the crypto bros. Letting the DARPA and Boston Dynamic chips (see video) fall where they may is societal malpractice. If my concerns here seem alarmist to you, imagine trying to explain in-flight wifi to someone in 1994.

We can't wait to make the decision about the limits we will put on the use of AI until after they’re deployed. If you don't think there's people in law enforcement salivating to deploy AI robots and drones in low-income and Black neighborhoods, you don't know American history.

Another Take on that Silly Chat Bot (or Death to the High School Essay)

I hold heterodox views on teaching writing. In brief, the way we teach writing in K-12 largely prepares students for further academia and not actual life. For example, I don’t use MLA (or APA) citations in my everyday life; you likely don’t either. I haven’t used them since grad school. When I want to cite a source, I use a hyperlink, occasionally a footnote.

When it comes to writing tasks, many teachers are obsessed with page and word counts. I try me best to avoid them. When I assign a task to a student, I tell them how many ideas or arguments they need to present, not how much to write. If a student can make a coherent argument for the abolition of the filibuster in the Senate, using two arguments, some evidence, and a counterclaim in 800 words, great! Oh, you need 1800 words? Fine. In the end, I care more about the ideas my students are interrogating than about the volume of writing they produce.

Writing instruction should ideally center on real-life use cases. They need opportunities to play with complex ideas rather than writing fewer, longer high-stakes pieces. I’ve been paid to write. Definitionally, I am a professional writer (don’t laugh). In all the occasions I have been paid to write something, it’s never been much more than 1000 words. If that’s good enough for Slate, it should be good enough for an IB/AP/A-Level comp class.

Now before you start erecting guillotines… Yes, students need to write and revise more often. Yes, they need to be taught to write for specific purposes and contexts (a wedding toast, a resume, a cover letter). But if a high school student can craft a coherent, thoughtful 1000 word essay, they’re in good shape. More isn’t better, it’s just more.

Lastly, almost all the writing my students do in class is hand-written, on-demand. I give them a prompt and some stimulus (a map, a data set, a passage from a primary source) and they go to town for the period. But in each class I teach, there’s usually also one longer, more formal essay each year where students are required to demonstrate more traditional essay skills. I really don’t enjoy reading or grading them, but I understand the exercise has some value.

The preceding was my philosophy on writing until last week when that silly chat bot rolled into room 157.

Real talk, I am never assigning another out of class essay as long as I live. Ain’t no way in hell I’m gonna throw away my evenings and Sunday afternoons trying to figure out if the essay I am reading is Charlie’s or a chatbot. Nope, nein, nada—that is for suckers. I ain’t no sucka.

But it’s bigger than that. The emergence of AI created content onto the mainstream of our society with essentially no public debate or government regulation is incredibly problematic. Even worse, Open AI, the creator of ChatGPT (this is the only time I will use the name of the bot in question because every time you mention them you’re advertising for them), was co-founded by the problematic richest man on the planet, Elon Musk. Even worse squared, another co-founder, Sam Altman, was behind a massive crypto scam, Worldcoin. It promised to provide a form of UBI by collecting iris scans from half a million people in developing states in exchange for a crypto token that now trades at $0.02221. I am not making this up—this is possibly the worst idea ever, carried out by the worst people possible.

To be clear:

I want nothing to do with it.

Burn it with fire.

Let it fall forever in the Mines of Moria with the Balrog that killed Gandalf.

If you think I am being extreme here, that’s okay. Most people I talk to about this topic say the same. I got called a luddite for this take on my own podcast Friday night.

Here’s the thing. Philosophically, when I am presented with a moral question, I assume the “most likely, worst case scenario” and work backwards in crafting my personal response and preferred public policy outcome. For example, should we arm teachers? Well, do you want a racist Karen teacher that “fears for her life” shooting a Black middle schooler? No? Me either. So that’s a rubbish idea. Next, do we want the coverage of the upcoming election to be a torrent of partisan AI crafted propaganda and foreign-funded AI disinformation? If your answer is no (and unless you’re a psycho or a libertarian tecno-triumphalist, the answer should be no), we have to ask ourselves how do we prevent this dystopian hellscape scenario from taking place?

That’s where my conversations about mainstreaming AI start. Some of these pieces coming out from teachers about how they plan to integrate the bots in their practice are the most naive non-sense I’ve read in my whole life. Obviously, AI and machine learning are coming and have a place in our future. But do we have to let some of the worst people on the planet implement it with literally no regulatory checks, foresight, and the absence of an inclusive societal discourse? That’s just silly but not as silly as assigning the same tired essay prompts in 2023.