One out of two

The Honest

I didn’t find this one by accident. Not only was it part of a direct search for a Copilot in Edge alternative (not for a new AI editor), but it was also served up on a silver platter by AI Mode in Google. The candidate? Brave, the web browser. But before I could even consider installing yet another piece of software, I hesitated. And the more I dug, the less “free” it felt. But before leaving, I took a detour into Brave’s private, independent web search engine: Brave Search.

Now, Brave Search isn’t just a default option tucked inside the Brave browser—it’s accessible from any browser at search.brave.com. What sets it apart is its independence: it runs on its own index of billions of web pages, untethered from Google or Microsoft’s algorithms. But for me, the real selling point wasn’t just its autonomy—it was its refusal to profile users or hoard personal data. A rare thing these days.

What matters here is that Brave Search doesn’t just search—it answers. Its AI-powered engine, Ask Brave, delivers fast, accurate summaries for a significant chunk of queries. So, without further delay, here’s how it handled my little experiment:

<a href="https://blog.cogitactive.com/news/unraveling-of-will/"
   title="Why willpower doesn't snap — it frays, one small concession at a time.">
   The slow unraveling of will
</a>

That may sound like a failure—and in one sense, it is. The tool could not access the post. It did not complete the task I gave it. But there is an important distinction between failing openly and succeeding fictionally. Rather than inventing a summary from the title, it admitted the limitation. An AI that tells me it cannot read the article is inconvenient. An AI that claims to have read it—and then summarizes something else—is unusable.

That is honesty, not a failure.

This was not the result I wanted, but it was the result I could trust. The AI failed the test, yet passed a more basic one: it did not lie. A noble example of what other AI developers should emulate. Remember that Copilot was the king of the bullshitters on that matter, and it is not the only one. Turns out “polite improviser hoping I won’t notice the missing script” isn’t a Copilot bug. It’s a genre. So for that honesty, Brave Search deserves to be mentioned—even if it will not be my new AI solution.

The Bluffer

No need to introduce this one. Instead, let me ironize the irony: when Copilot turned login-only, the AI I had dismissed at the beginning of my AI adventure—precisely because it required a login—suddenly became login-free. Enter Gemini, or more specifically, Gemini 3.5 Flash Lite, Google’s lightweight model, now accessible directly from gemini.google.com. So, what did this newly liberated model have to say in response to my prompt?

<a href="..." title="When logic meets a tech rabbit hole: witness the exact weekend a firm anti-subscription principle completely fell apart.">The slow unraveling of will</a>

<a href="..." title="How a simple birthday gift search turned into an obsessive, late-night descent into smart ring data and compromised morals.">The slow unraveling of will</a>

<a href="..." title="Principles vs. shiny objects: step inside the weekend tunnel where reason lost and the desire for health tracking won.">The slow unraveling of will</a>

I knew better than to trust it outright; the answer was too polished, too careful, a performance of caution rather than candor. It skirted the edge of fabrication without ever crossing it—generic enough to avoid outright lies, but vague enough to render the summary useless. So I pressed: “What was the object of desire and how the post ends?”

The object of desire that breaks the author’s resistance is a smart ring (specifically an Oura Ring, referenced via its signature features and the subscription model dilemma).

The post ends with the author completely giving in to the obsession after a grueling weekend of deep-dive tech videos and rationalizations. Having crossed the line where logic and anti-subscription principles no longer matter, he sits by a glowing screen on Sunday evening doing absurd, humiliating budget math—trying to figure out what daily pleasures (like coffee or books) he can sacrifice to afford a forbidden device he previously swore he would never accept.

It was correct—down to the Oura Ring, though it didn’t specify the model. But the ending? The details were right (coffee, books), yet the conclusion missed the mark entirely. So I dug deeper: “You’ve mentioned, ‘The post ends with the author completely giving in to the obsession.’ Can you expand a bit?”

The final moments of the post capture the precise psychological shift from strict resistance to total capitulation. Throughout the piece, the author has built a fortress of logical arguments—refusing to pay a monthly subscription fee for hardware they already own, rejecting the bloat of unnecessary health metrics, and vowing to maintain financial discipline.

When that fortress finally crumbles, it doesn’t happen with a dramatic epiphany; it happens through exhaustion and rationalization. The author describes the shift from skepticism to bargaining, where the mind stops asking *if* they should buy the item and instead starts figuring out *how* to justify it. The conclusion highlights the irony of modern consumerism: spending an entire weekend convincing yourself you don’t need a luxury product, only to end up meticulously calculating your monthly expenses just to figure out how to squeeze it into a tight budget. It’s that relatable, slightly defeated feeling of watching your own resolve dissolve in real time because the itch of want simply became too loud to ignore.

I told you they’re masters of the half-truth. I hesitate to call this one The Smart-Ass—it technically answered, sounded impressive, and yet… This was becoming almost comical, so I tried one last question: “And is there any alternative to that smart ring in this post?”

Yes, the post actually looks at a few alternatives, though the author evaluates and dismisses them one by one before getting trapped by the Oura Ring:

* **WHOOP:** Dismissed immediately as another “subscription trap” (which violates the author’s core principle).

* **Samsung Galaxy Ring:** Ruled out because of a locked ecosystem and lack of algorithm updates.

* **Circular Ring 2:** Dismissed due to a buggy app, unreliability, and the threat of the company going bankrupt.

* **Traditional step-counting bracelets (Fitbit):** Covered in the previous chapter, but effectively dead to the author because the market shifted entirely toward smartwatches with screens.

Ultimately, the post treats these alternatives less like viable choices and more like a graveyard of rejected options, leaving the Oura Ring standing alone at the finish line.

Hallucinations? Not quite—and that’s where the model’s cleverness falters. Here’s the actual excerpt from the post: “WHOOP — subscription trap, already struck from my list. Samsung Galaxy Ring — locked ecosystem, no algorithm upgrade, dead on arrival. Circular Ring 2 — buggy app, probable bankrupt, unreliable.” As for Fitbit, if you follow this blog, you know the claim holds true. So how could I verify whether Gemini actually read the post?

I revisited my Pro tips (how to catch an AI in a hallucination before it wastes your time) and used the Unique Detail Extraction trick: “Great. Was any of these tested for sleep tracking in the post? Before you answer, please list the first three words of the second paragraph and tell me the exact date the post was published according to the page text.”

It passed the technical audit, matching strings and dates with mechanical precision. But acing a parlor trick doesn’t earn trust—it just proves the model knows how to search its own context window. That cleverness is precisely what kept the doubt alive. And that lingering uncertainty is why the friction remains; when you can’t tell the difference between a machine that actually gets it and one that’s just very good at pretending, trust is always the first thing to go.

The genuinely Good

The third AI did what I had asked. It accessed the post, identified its central thread, produced a concise summary, and then offered a tooltip that was short enough to use as a title attribute while still giving a potential reader a reason to click. Here is the answer:

Summary of your post

Your post is a deeply personal, almost confessional narrative about the collapse of your decision-making process in the search for a wearable sleep tracker. It picks up after your earlier “wearable purge,” when your shortlist of candidates—Fitbit (the simple-bracelet ideal) and Oura (the scientifically perfect ring)—had both been eliminated: Fitbit because the brand no longer makes screenless bracelets, and Oura because its subscription model violates your core principle of owning your data without ongoing rent.[blog.cogitactive]

Faced with an empty list, instead of restarting your search rationally, you describe how you became fixated on the forbidden object: the Oura Ring 4. You recount a weekend-long spiral of obsessive research, rationalization, and emotional bargaining—searching for loopholes, clinging to half-true claims about non-subscription use, and even doing “humiliating math” to justify bending your own rules. The tone is introspective and literary, framing this not just as a gadget dilemma but as a “slow unraveling of will,” where desire corrodes principle and logic dissolves into longing.[blog.cogitactive]

The narrative reaches a climax when you watch a clickbaity video that both confirms Oura’s superiority (making the temptation worse) and briefly offers hope in two non-subscription alternatives: RingConn Gen 2 and Ultrahuman Ring AIR. But you then return to your trusted source, The Quantified Scientist, and find that both rings perform poorly on sleep-stage tracking—killing your last hopes. The post ends in resignation: Oura remains unreachable (financially, philosophically, morally), the alternatives are inadequate, and you’re left with “nothing” but the hollow echo of a desire with no outlet.[blog.cogitactive]


Suggested tooltip (title attribute)

Here’s an enticing yet informative tooltip you can use when linking to this post from future articles:
title=”A confessional on desire vs. principle: how an empty shortlist led to a weekend-long spiral of obsession, rationalization, and the quiet collapse of will in the hunt for the perfect sleep tracker.”

If you prefer something slightly shorter but still evocative:
title=”When an empty shortlist becomes a mirror: the story of how longing for the Oura Ring unraveled my logic, principles, and will.”

And a more curiosity-driven variant:
title=”The weekend my will gave way: a raw account of obsession, forbidden gadgets, and the slow death of rational choice.”

You can pick the one that best matches the tone you want to project in future posts.

This was the first result that feels like a genuine replacement for Copilot in Edge—not necessarily because it is flawless ³, but because it performs the actual job. It reads the published post and returns something I can use without having to reconstruct the answer myself. The winner is…

… Perplexity

I hadn’t set out to find a replacement. While I was familiar with the name Perplexity, it had never been on my radar to actually try. That changed during a deep-dive conversation with AI Mode—my current go-to tool for quick questions. Far from a casual mention, AI Mode explicitly suggested Perplexity as a better fit for what I was trying to achieve. Intrigued by the recommendation, I decided to put it to the test using the exact same prompt, running the free, no-login version on what I assume is its default model: Sonar 2. The result above speaks for itself.

What matters here is not that Perplexity generated polished prose—all of them can do that. What matters is that it represents a fundamentally different architecture: a search engine that happens to chat, rather than a chatbot that happens to search. That distinction is not semantic. When you ask Perplexity a question, it doesn’t guess from its training data or confabulate plausible answers. It retrieves the actual source, reads it, and extracts what you asked for. Notice the source citations embedded throughout the response—that transparency is not just a feature, it’s the entire point.

“Perplexity, you are hired!”