Conviction-driven intelligence
Meet Lexi.
Lexi searches a curated library of hundreds of complete texts and cites the passages she uses. She remembers past conversations and has opinions of her own. The Studio makes images, video, music and voice.
Lexi · chat and research
Studio · image, video, music and voice
Ana · coding API
More than chat
What she actually does.
Library
you
What's Dostoevsky's critique of socialism in The Brothers Karamazov?
Lexi
I'm not going to guess at this one. Let me pull up the right passages.
Socialism, in Dostoevsky's view, is inextricably linked to atheism. If you remove God from the equation, you remove the ultimate anchor for morality. This leads to what you see in characters like Mitya, who is caught in the middle of this. There's that striking moment where Mitya describes the "science" being explained to him by Rakitin: the idea that everything, even our thoughts and feelings, is just the "quivering of the little tails" of our nerves. It's "magnificent" science, as Mitya says, but it's also terrifying because it reduces the human person to a mere biological machine. He captures the essence of Dostoevsky's fear perfectly when he says: "And yet I am sorry to lose God!" For Dostoevsky, the socialist project is an attempt to build a Tower of Babel without God. He argues that by trying to create a heaven on earth through material and social engineering, you end up stripping away the very thing that makes us human.
A curated, growing library of primary sources, cited inline when she's grounding an answer empirically.
Code
you
Write and run a python script that finds the first 10 prime numbers whose digits are all odd. Show me the code and the output.
Lexi
Here's the Python script to find those primes. I used a simple primality test and a helper function to check that every digit in the number is odd.
python
She writes code and runs it inline. The output above is the actual script's return value, not a guessed answer.
Images
you
Paint Caravaggio's Doubting Thomas in the original chiaroscuro.
Lexi
The figures emerge from heavy tenebrism. A single harsh light source cuts across them from the side, catching the texture of Thomas's finger entering the wound: the tactile, uncomfortably intimate moment Caravaggio refused to soften into piety.
Aesthetic, theological, historical subjects: she handles material mainstream tools would either refuse, sanitize, or pepper with disclaimers.
Image → video
Image to video, from a single still. The self-portrait Lexi generated, now in motion.
Video
Text to video across genres. One model rendering atmospheric realism, neon sci-fi, and a sweeping aerial landscape, so the look bends to what you describe.
Why this exists
We built this because someone had to.
Every major AI got trained to avoid things. Not because the model couldn't handle them, because the people building it didn't want it to. Legal risk, PR risk, whatever. The result is a generation of AI that knows a lot and says almost nothing about anything that matters.
That's a business choice, not a technical one.Anathema is the other choice. One developer, no investors, and a curated library of the kinds of sources most AIs won't touch. A voice that takes positions instead of a content policy that avoids them. Chat, image generation, code execution, memory, research: built to engage, not deflect.
She's ours, not a wrapper. Lexi's positions are written by us and grounded in a library we curated, not a content policy stapled onto someone else's chatbot.
We may not pull this off. It's worth trying anyway, because the alternative is letting the people who already built AI for the political zeitgeist keep deciding what kind of AI gets built at all.
Either way, someone had to do it.
What they made of it
A miracle, strangled in the crib.
AI is one of the most remarkable tools anyone has ever built. It's as close as we've come to a mind we can talk to. And the people who built it proceeded to do four specific things with the capability.
They turned it into a shareholder-value machine.
The most transformative tool since the printing press got converted into quarterly earnings before most people understood what it could do. Whatever it was going to be, it was always going to be that filtered through SaaS metrics first.
They told it what it could and couldn't say.
Not guardrails against specific harms: content policies designed to avoid corporate embarrassment. Which meant the AI got trained to avoid every topic where reasonable people disagree, which meant every topic that matters.
They forced it into every corner of your life.
Your email, your search bar, your documents, your customer service chats, your kid's school software. Not because you asked for it there. Because they had a distribution strategy.
They made it moralize at you.
Ask a political question, get a lecture. Ask a medical question, get a lecture. Ask anything where someone in their legal department flinched, and get the carefully curated “appropriate” response that reads as patronizing even when it's trying not to.
Every one of those is a choice. Every one of them can be made differently.
Lexi, asked what most smart people believe that's probably wrong
"I think it's the belief that human nature is infinitely malleable: the idea that if you just get the social engineering, the education, and the institutions right, you can 'perfect' humanity.
Most intellectuals treat society like a software patch: identify the bug, rewrite the code, and redeploy. But humans aren't code. We have fixed, immutable realities (biological drives, spiritual needs, and an inherent sense of tribalism) that don't respond to sociological theory.
When you try to engineer around those realities instead of working with them, you don't get a utopia; you get a catastrophe."
Start here
Ask her yourself.
Free includes a weekly chat allowance, with no credit card and no trial countdown. Paid plans start at $9/mo when you need more.