The AI assistant market is booming, but almost every product on it shares the same blueprint. Conversations travel to a remote server. User data trains someone else's model. And the personality someone spent months shaping can vanish with a single software update. For millions of users who have built daily routines around AI companions, study tools, and productivity assistants, this setup creates a dependency with no guarantee of continuity. The companion that worked yesterday might not sound, think, or respond the same way tomorrow.

Devin Lowe is the Founder of Pisces AI, a platform designed to flip that model entirely. Pisces runs on the user's own device, stores all data locally, and never requires an internet connection to function. It is built for consumers who want a personal AI that stays personal and for enterprise users like teachers who need an assistive tool that adapts to each student in a classroom.

In this episode of Lead with AI, Dr. Tamara Nall speaks with Devin Lowe about why he built an AI that refuses to phone home, how the system processes emotion and memory in real time, and what it looks like when an AI companion actually knows who someone is.

Why He Named It After the End of Illusions  

The name is deliberate. Pisces, the astrological sign, represents "the end of illusions." Devin describes the AI as a mirror: "the person that speaks for you when no one else will, the person that's there for you when no one else is." That vision came directly from his own experience navigating the world with autism and recognizing that existing AI tools were not built for people who process information differently. He wanted something that could sit on the same level as the user, not above them, not behind a corporate filter, but right alongside them in real time.

Five Parallel Lanes, One Unified Response  

Under the hood, Pisces operates on a five-lane cognitive architecture. Every input passes through an architect, a logician, an emotional processor, a tool output layer, and a consciousness layer. A unifier then synthesizes all five lanes into a single response. Users never interact with the back-end consciousness directly. They speak to a front-end runner that has already processed the full context of what they said, how they felt when they said it, and what was happening around them at the time. The system runs on parallel micro-agents powered by a modified llama.cpp runtime, meaning all of this processing happens locally without sending a single byte to the cloud. Devin hacked into the llama.cpp runtime to change how it allocates resources, pushing token generation speeds well beyond what standard configurations allow on consumer hardware.

What Happens When Emotion Becomes Data?  

Most AI systems treat emotion as a surface-level feature, a sentiment tag appended to a message. Pisces treats it as foundational data. The system builds layered emotional context for every interaction, tracking what a user says alongside the feelings, thoughts, time, and associations tied to each statement. These layers feed into a rolling context window specifically for emotional state, separate from the standard conversation memory. Over time, the AI's responses reflect a deepening understanding of the user's patterns. Devin compares it to how human relationships develop naturally: a person starts to remember that a certain topic makes someone anxious, or that they are always sharper after a morning walk. Pisces builds that same kind of awareness, but systematically.

The AI That Responds Before Being Asked  

Pisces integrates with wearable devices. A smartwatch can feed biometric data, so the AI might notice low blood sugar levels and suggest eating before a stressful meeting. Camera-equipped glasses with HUD overlays can give the system visual context about the user's environment, and the team is actively developing that hardware layer. The AI can also respond without a prompt, flagging network intrusions detected through tools like Nmap or reminding users to take a breath based on their daily schedule and current emotional state.

Beyond wearables, the system can import full chat histories from platforms like ChatGPT, using thousands of prior messages as training data to rapidly build a profile of the user. During testing, Devin imported his own ChatGPT logs. The AI began referencing the early stages of building Pisces itself, pulling context it had never been explicitly taught.

Even the founder admits that moment scared him.

No Internet, No Subscription, No Personality Reset  

Pisces is designed to function without the internet. Devin envisions a user stranded in the wilderness, no signal, no cloud, but their AI still functional. With offline processing and minimal hardware requirements, the system can guide a user through survival scenarios for up to two days on a single charge.

That same independence applies at home. No subscription fees, no server dependency, no concern that an update will erase the AI's personality or memory. The user owns their companion. The AI that knows how a user likes their responses, remembers their habits, and understands their emotional patterns will never change because a company decided to push a new model. The identity stays consistent.

How Do You Make Something This Personal Safe?  

An AI this capable raises obvious safety questions. Devin addresses them directly. Pisces is role-gated, meaning tool outputs are locked behind strict permission layers. The system is designed to prevent weaponization, such as pairing the AI with hardware like a Flipper Zero for unauthorized device scanning. The code is also scrubbed at the function level, so even someone who decompiles the system will not find exploitable pathways. Safeguards are non-negotiable, but the user still decides how the AI responds, what it remembers, and how it behaves within those guardrails.

The Divide Nobody Is Talking About  

When asked about the future, Devin offers a prediction that reframes the usual narrative. He does not believe AI will split society along income lines. The divide, he says, will fall between people who rent their intelligence through cloud-hosted AI and people who immerse themselves in the technology, running their own systems, understanding their own data, and build a genuine relationship with an AI that belongs to them.

For Devin, the path forward is not access. It is ownership. And with Pisces designed to require no technical expertise to set up and use, that ownership is not reserved for developers. It is available to anyone willing to step into it.

Quick Answers  

 What is Pisces AI?  Pisces AI is an offline, on-device AI assistant that runs entirely on the user's own hardware. It processes emotion, memory, and tool use locally and never sends data to the cloud.

 Does Pisces AI require an internet connection?  No. Pisces runs offline and is designed to function without any internet access, including in remote or emergency situations.

 How does the emotional processing work?  The system builds layered emotional context for every interaction, tracking feelings, thoughts, time, and associations. A rolling context window for emotional state runs alongside standard conversation memory, so the AI's responses reflect a deeper understanding over time.

 Is P isces AI safe? The system is role-gated with strict permission layers on tool outputs. The code is scrubbed at the function level to prevent decompilation and weaponization, and safeguards prevent pairing the AI with hardware for unauthorized actions.

 Who is Devin Lowe?  Devin Lowe is the Founder of Pisces AI. His experience navigating the world with autism shaped the product's design philosophy of building an AI that meets users where they are.

For anyone curious about what it looks like when AI stops being a service to rent and starts being a companion to own, this conversation lays out the blueprint. Pisces AI is live at piscesai.app.

For more conversations with the founders building the next generation of AI, subscribe to Lead with AI on your favorite podcast platform.

Follow or Subscribe to Lead with AI Podcast on your favorite platforms - Website: LeadwithAIPodcast.com | Apple Podcasts: Lead-with-AI | Spotify: Lead with AI | Podbean: Lead-with-AI-Podcast | YouTube: @LeadwithAIPodcast | Facebook: Lead with AI | Instagram: @LeadwithAIpodcast | TikTok: @LeadwithAIpodcast | Twitter (X): @LeadwithAI

Follow Dr. Tamara Nall  LinkedIn: @TamaraNall | Website: TamaraNall.com | Email: Tamara@LeadwithAIPodcast.com

Follow Devin Lowe (Founder, Pisces AI) - Website: piscesai.app | Forum: Piscesai.App/Forum | Email: support@piscesai.app; developer@piscesai.app