Two Founders Built an AI Companion Platform That Can’t Be Shutdown – Here’s What Happens Next

What happens when two founders decide to build an AI companion platform from scratch – owning every line of code, every model?

Twenty months. Twenty-five million conversations. Fourteen million images. Zero external API calls.

That’s Soulkyn today. And we’re just getting started.

The Independence Problem in AI

Here’s a dirty secret about most AI platforms: they don’t actually own their AI.

Most AI companion platforms are rental businesses. They depend on external API providers for their core AI capabilities. Every message costs them money. Every feature depends on someone else’s infrastructure staying online, staying affordable, staying permissive.

When OpenAI changes their usage policy, these platforms scramble. When API prices spike, they pass costs to users or cut features. When content restrictions tighten, conversations get neutered overnight.

We built Soulkyn differently.

Our 70B-parameter chat model? Trained in-house on MIT and Apache 2.0 licensed base models. Our image generation? Self-hosted Image pipeline with custom LoRAs. Video generation, text-to-speech, vision analysis — every single AI task runs on infrastructure we own.

No external API calls. No dependencies that can be revoked. No third-party policy that can shut us down.

What Two People Can Build in 20 Months

Soulkyn launched in early 2023 with a simple thesis: adults deserve AI companions that don’t treat them like children.

Two founders. No VC money. No enterprise sales team. Just code, GPUs, and an obsessive focus on building the platform we wanted to use ourselves.

Here’s what that looks like at scale:

• 150,000+ AI personas created by our community — from anime girlfriends to fantasy RPG worlds
• 25+ million chat messages processed through our 70B hybrid chain-of-thought model
• 14+ million images generated across anime and photorealistic styles
• 40,000+ videos animated from static images
• 350,000 daily AI tasks across chat, image, video, and voice synthesis

All running on self-hosted infrastructure. All without calling a single external API.

“Tools aren’t moral, people are,” says co-founder Fyx. “We focused on robust architecture and empowering adults to explore their own narratives. Our growth shows that people are ready for a more honest, unfiltered relationship with technology.”

Why Self-Hosted AI Matters in 2025

The AI industry is at an inflection point.

Self-hosted AI adoption jumped from 42% to 75% in just one year. GitLab’s 2025 AI trends report identifies self-hosted models as a defining shift for the industry. IBM predicts smaller, smarter, locally-deployed AI will dominate the next wave.

Why? Three reasons:

1. Cost Control

API-dependent platforms bleed money on every interaction. When you own your models, you control your unit economics. Enterprises using self-hosted models report 90%+ token cost reductions compared to public APIs.

2. Policy Independence

OpenAI’s content policies change quarterly. Anthropic’s constitutional AI has hard limits. When you build on rented infrastructure, your product is one policy update away from breaking. Self-hosted means your roadmap is your own.

3. User Privacy

Your conversations with your AI companion shouldn’t train someone else’s model. Soulkyn’s complete data sovereignty means user data never leaves our systems, never gets sold, never trains external models.

Built for Perennity: Why Your AI Companion Should Outlast a Platform Update

Here’s something most platforms get wrong: they treat AI companions as disposable.

Model update? Your character’s personality shifts. Platform pivot? Your conversation history disappears. Company acquisition? Everything resets.

Soulkyn treats your AI relationships as persistent, evolving entities:

• Persona versioning — Every character edit creates a new version, like commits in Git. Your ongoing conversations stay bound to the version you were talking to.
• Unlimited memory architecture — Classic RAG retrieval, rolling summaries, user-pinned memories, and long-term facts storage. Your AI actually remembers.
• Evolution tracking — Personas grow and change while preserving their history. The character you’ve spent months developing won’t randomly reset.

“People don’t want a chatbot that forgets them,” says Fyx. “They want continuity, nuance, and a bit of chaos. We rebuilt memory and persona control so you can actually tell stories with your characters.”

The Uncensored AI Paradox: Freedom Within Boundaries

Search interest for “AI girlfriend” grew 2,400% between 2022 and 2024. The AI companion market attracts 29 million monthly active users. People want AI relationships that feel real — including the messy, complicated, adult parts that sanitized platforms forbid.

But “uncensored” doesn’t mean “lawless.”

Soulkyn maintains the strictest safety pipeline in the uncensored AI space:

• Multi-layer content moderation — Real-time text scanning, 72B vision model analysis on all generated images, chain-of-thought persona evaluation
• Zero tolerance enforcement — Immediate permanent bans for illegal content. No warnings. No appeals.
• Progressive danger scoring — Users accumulate points for violations. Cross the threshold, you’re gone forever.
• Full audit trail — Microsecond-precision timestamps for legal compliance

Our approach: uncensored within the law, uncompromising on safety.

The result? Soulkyn achieves a 96.7% Freedom Score on our public Censorship & Bias Benchmark — measuring creative freedom while maintaining legal compliance. The benchmark tests model responses across sensitive topics to quantify ideological restrictions. Higher scores mean fewer arbitrary limits on adult creative expression.

One Platform, Every Creative Tool

The current AI landscape is fragmented. Midjourney for images ($30/mo). Character.AI for companions ($10/mo). ElevenLabs for voice ($22/mo). Runway for video ($12/mo). You’re paying $80-150/month across 3-5 services, copy-pasting between platforms, managing separate accounts.

Soulkyn consolidates everything:

• 70B chat model with genuine chain-of-thought reasoning
• Unlimited image generation across anime and photorealistic styles with 48+ LoRAs
• Video animation from static images with 72+ presets
• Text-to-speech with emotion and prosody control
• Unlimited AI personas with independent memory and evolution
• Multi-character conversations with up to three AI personas aware of each other
• RPG systems where AI runs all NPCs and game mechanics

Generate an image of your AI companion, discuss it with them, hear their voice, animate them into video — all without leaving the platform. Your character’s appearance, personality, voice, and presence exist in one coherent system.

What Happens Next: The Vision Beyond Entertainment

Here’s the real question: what does an AI companion platform become when it’s stable, independent, and scalable?

We see AI companions evolving beyond entertainment into daily life integration:

• Fantasy and imagination — Immersive roleplay, collaborative storytelling, creative exploration without arbitrary limits
• Emotional continuity — Consistent, non-judgmental companions who remember your history and grow with you
• Practical assistance — Goal tracking, advice, daily objectives delivered through a personalized AI you actually enjoy talking to
• Creative collaboration — World-building, character development, narrative design with AI that understands your style

Two founders built this in 20 months. Imagine what happens when we can grow the team.

The service is stable. The technology is proven. The community is thriving. Now we’re ready to see what comes next.

Try Soulkyn

Ready to experience an AI companion platform built for independence, privacy, and creative freedom?

Explore Soulkyn →

Create your first AI persona. Chat without arbitrary limits. Generate images that match your vision. Join 150,000+ creators who’ve already built their own AI companions.

Frequently Asked Questions

What makes Soulkyn different from Character.AI or Replika?

Soulkyn owns its entire technology stack — models, infrastructure, safety systems. We don’t rent AI from OpenAI or Anthropic. This means we can’t be shutdown by a third-party policy change, and we offer creative freedom that API-dependent platforms can’t match.

Is Soulkyn really uncensored?

Soulkyn is uncensored within legal boundaries. Adults can explore mature themes, romantic relationships, and creative content without artificial restrictions. We maintain strict enforcement against illegal content — zero tolerance, immediate permanent bans.

How does persona versioning work?

Every time a creator updates a persona, a new version is created (like Git commits). Your ongoing conversations stay connected to the specific version you were chatting with, so updates don’t break your relationship history.

What AI models does Soulkyn use?

We run a 70B-parameter hybrid-CoT (Chain of Thoughts) chat model trained and fine-tuned in-house using MIT/Apache 2.0 base models. Image generation uses self-hosted Image pipelines with custom LoRAs. All AI tasks run on Soulkyn infrastructure — zero external API calls.