Augmented Intelligence for Actual Humans

While Silicon Valley races to replace humans,
we’re on a mission to elevate them.

Your brain's not broken.
Your tools are.

We design augmented intelligence that works for neurodivergent minds, not against them—because what works for the edges, often benefits us all.

The Problem


We're witnessing the fastest cognitive acceleration in human history. In 18 months, AI has evolved from writing simple emails to replacing entire job categories—but the tools we're given to adapt were designed for a world moving at human speed.

18
Months from ChatGPT launch to AGI-level reasoning models
68%
of knowledge workers will use AI daily by end of 2025
84%
of AI tools have experienced data breaches

The Acceleration Crisis:

  • AI capabilities are doubling every 6 months, but human adaptation takes years
  • Big Tech optimises for engagement and data extraction, not human flourishing
  • Neurodivergent minds—6.8% of the population—are canaries in the coal mine for cognitive overwhelm
  • We're 24-36 months from AI systems that outperform humans in most cognitive tasks

The Critical Window: Why Human-Centred AI Must Win Now

"We're not just building better tools. We're defending the future of human agency in an AI-first world."

The current moment represents a unique inflection point in technological history. Unlike previous revolutions that unfolded over decades, AI capabilities are advancing exponentially while the window for shaping their development rapidly narrows. The choices we make in the next 24 months will determine whether AI augments human potential or replaces it entirely.

The Exponential Reality

In November 2022, ChatGPT could barely write a coherent essay. By January 2025, AI systems demonstrate reasoning capabilities that match or exceed human performance in mathematics, coding, and complex problem-solving. This represents a 100x improvement in capability in just 24 months.

Conservative projections suggest we're 24-36 months from Artificial General Intelligence (AGI)—systems that outperform humans in virtually all cognitive tasks. Once this threshold is crossed, the window for shaping AI development closes permanently.

The Human Lag

While AI capabilities double every 6 months, human institutions adapt on 5-10 year cycles. Educational systems still teach skills that AI already surpasses. Workplace policies lag years behind technological realities. Most critically, the cognitive tools we use daily were designed for a pre-AI world.

This creates a dangerous gap: humans are being asked to collaborate with superhuman AI using tools designed for human-speed thinking.

Neurodivergent Minds: Canaries in the Cognitive Coal Mine

The 6.8% of the population with neurodivergent cognitive patterns aren't just struggling with current AI tools—they're revealing the fundamental flaws in how we're approaching human-AI collaboration.

Attention Fragmentation
ADHD minds, already vulnerable to dopamine exploitation by social media, face existential threat from AI systems optimised for engagement over understanding. What overwhelms neurodivergent minds today will overwhelm everyone tomorrow.
Cognitive Mismatch
Tools designed for neurotypical processing patterns fail 67% of neurodivergent users. As AI accelerates information density, this mismatch becomes a preview of universal cognitive overwhelm.
Privacy Vulnerability
Neurodivergent individuals are more likely to share sensitive information with AI due to enhanced emotional processing and trust patterns. This reveals the exploitation potential of current AI architectures.
Adaptation Advantage
When tools actually work with neurodivergent cognitive patterns, the results are extraordinary: hyperfocus capabilities, pattern recognition, and innovative thinking that outperforms neurotypical approaches.

The Surveillance Capitalism Endgame

Big Tech's AI strategy isn't just about building better assistants—it's about capturing the final frontier of human cognition. Every thought you share with ChatGPT, every prompt you give Claude, every conversation with AI becomes training data for systems designed to predict and influence human behaviour at unprecedented scale.

The Data Reality:

  • 84% of AI tools have experienced data breaches, yet users continue sharing intimate thoughts and creative processes
  • 8.5% of business prompts accidentally expose sensitive company information to systems that explicitly use data for training
  • 68% of employees bypass corporate AI policies by using personal accounts, creating massive security blind spots
  • 100% of "free" AI services monetise user data—when you're not paying for the product, you ARE the product

The stakes couldn't be higher: we're deciding whether AI will amplify human potential or harvest it for corporate profit.

The Alternative Path: Augmented Intelligence for Actual Humans

Heumans represents a fundamentally different approach to AI development—one that starts with human dignity rather than data extraction, cognitive diversity rather than algorithmic conformity, and community ownership rather than corporate control.

Our Approach:

  • Privacy by Architecture: 100% local processing means your thoughts never leave your device—not because we promise to protect them, but because we literally cannot access them
  • Neurodiversity as Design Principle: Built by neurodivergent minds for neurodivergent needs, creating tools that work better for everyone
  • Research-Backed Development: 50+ pages of ADHD memory research informing every design decision, not growth hacking for engagement
  • Community-Driven Innovation: Users own their data, their insights, and their contributions to the platform's development

The Narrowing Window

We estimate 18-24 months remain before the AI landscape crystallises around Big Tech solutions. Once network effects and data advantages solidify, alternative approaches become economically unviable.

This isn't just about building better productivity software. It's about proving that human-centred AI can deliver superior outcomes while preserving cognitive autonomy, personal privacy, and neurodivergent dignity.

What Success Looks Like:

  • Demonstrable proof that privacy-first AI delivers superior cognitive augmentation
  • Measurable economic returns from neurodivergent-inclusive design principles
  • Viable alternative to surveillance capitalism in the AI economy
  • Foundation for the next generation of human-AI collaboration tools

Personal Stakes: Why I'm Building This

My late-life ADHD diagnosis at 40 coincided with ChatGPT's launch. I experienced firsthand how AI could transform neurodivergent cognitive patterns from liabilities into advantages—but only when the AI truly understood how my brain worked.

After nearly losing everything when a previous co-founder exploited technical vulnerabilities to steal my company, I understand the stakes of building technology that genuinely serves users rather than exploiting them.

Through my work with Conductor Software on psychological safety measurement, I've seen how small changes in system design can produce massive improvements in human outcomes. The same principles that create psychological safety in teams can create cognitive safety in AI interactions.

We have one shot to get this right. The next 24 months will determine whether AI amplifies human potential or replaces it, whether neurodivergent minds are empowered or exploited, whether privacy becomes a luxury good or a fundamental right. Heumans exists to ensure we choose correctly.

The Heuman Alternative


Almost every major tech platform follows the same playbook: capture user attention, extract valuable data, optimise for engagement over understanding.

Big Tech

Data as Product: “Free” services that monetise your thoughts and relationships

Engagement Optimisation: Designed to capture and hold attention

One-Size-Fits-All: Mass market solutions for neurotypical patterns

Surveillance First: Privacy as afterthought, if at all

Heumans

Privacy by Design: Your thoughts never leave your device

Education Over Engagement: Learn about the transformative power of AI, with zero ads or notifications

Neurodivergent-Led: AI that adapts to how you think

Dignity First: Cognitive augmentation, not replacement

Meet Penny


Finally, an AI that understands that just because it can do something, doesn’t mean it should. Consider Penny your personal guide to the future of Human-AI interaction

Cognitive Pattern Recognition

Learn how your brain processes information

Discover your natural thinking patterns, attention cycles, and memory preferences. Knowledge that helps you work with AI tools more effectively.

Multi Dimensional Memory

Beyond simple note-taking

Captures temporal, emotional, relational, and procedural context—how your brain naturally processes information. Not just what you said, but when, why, and how it connects.

Adaptive Intelligence

AI that recognises cognitive states

Detects post-hyperfocus crashes, energy shifts, and attention patterns. Provides support that actually helps instead of overwhelming.

100% Local Processing

Your thoughts stay yours, always

Zero cloud dependency. No data transmission. No corporate surveillance. We literally cannot access your information—by design, not just by promise.

Infrequently Asked Questions

  • No! Penny runs 100% locally on your computer with zero cloud dependency.

    Once installed, it works completely offline. This is a core feature - your thoughts never leave your device.

  • Absolutely not. Penny uses its own fine-tuned, local AI models that run entirely on your machine. No external AI service accounts required.

  • Penny understands context and makes connections between your thoughts over time.

    It's like having a thinking partner who remembers everything and helps you see patterns you'd miss.

  • Penny is built with Tauri (auditable open architecture) and has no network connections to external servers.

    Your firewall logs will show zero outbound traffic from Penny, because we designed it to be yours alone.

  • Beyond purchasing, you can contribute through the research lab by sharing anonymised insights about how you prefer to work / collaborate with Penny, participating in user testing, helping to onboard new users, or providing crucial feedback that shapes development.

    Alternatively, if you’re interested in investing or sponsoring this project, email murray [@] heumans [dot] com

  • Absolutely, but always maintaining the privacy-first approach.

    Teams will soon be able to share insights, but never raw personal data.

  • The core Penny app will always be available as a one-time purchase.

    Additional features (ie model upgrades, character agents) may have separate pricing down the track but for now — One price covers you for everything.

  • Penny analyses your conversations across five dimensions (emotional, temporal, relational, procedural, contextual) and builds a knowledge graph showing how your thoughts connect.

    Over time this knowledge graph expands, providing ever more valuable insights about the way your brain operates and develops relationships between ideas.

  • 1. We believe in accessible innovation.

    Early supporters like you make this possible. In return, you get lifetime access and opportunities to share in the profits - when we succeed, you succeed.

    2. It’s early days.

    We know we’ve got a long road ahead, so it seems only fair to massively reduce the price for people willing to adopt and support early.

  • No. Since Penny doesn't connect to the internet and stores everything locally, there's no remote attack surface.

    It's like having a private diary that only exists on your computer (but can also think and help you work through ideas like no other diary in history).

  • Definitely not. We are focused on learning how every brain works, in order to build better products for everyone.

    Whether you're neurodivergent, neurotypical, or have no idea what any of that means - if you're excited about thinking differently and want to be part of building something new, you belong here.

  • Yes! Penny has a "Community Contribution Model" where you can earn access through valuable contributions like user testing, community moderation, knowledge sharing, or accountability partnerships instead of direct payment.

    We’ll be bringing this online as quickly as possible - stay tuned.

  • v1 is Mac only, sorry.

    You'll need at least 8GB RAM and a decent processor. The AI models require some computing power, but most computers from the last 5 years should handle it fine.

    The next version of Penny will work on Windows, Mac, and Linux.

  • Very little. Your memories are stored as lightweight JSON files.

    Even years of conversations typically use less than 100MB - about the size of a few photos.

  • Major updates quarterly, with smaller improvements monthly (if not more often).

    The roadmap is community-driven based on user feedback, so the more we hear back from you, the more likely we’ll implement your preferred features.

  • Right now Penny works through text conversation. Document upload and file analysis features are planned for future releases.

  • One-time purchases fund development, community contributions reduce barriers, and future features (character agents marketplace) create ongoing revenue without subscriptions.

    We think that’s super sustainable.

  • Absolutely. Your data lives in simple files on your computer that you can copy, backup, or export anytime.

    No vendor lock-in.

  • Mobile companion apps are planned for 2026.

    The main processing will still happen on desktop for privacy and performance.

  • Only the information you explicitly choose to share in conversations.

    No telemetry, no usage tracking, no analytics.

    The only data generated is your memories and conversation history, which are stored locally and nowhere else.

  • Just copy your Penny data folder to your new machine and you're good to go.

    All your memories and conversation history transfer perfectly.

  • Penny only uses resources when you're actively using it.

    When minimised or closed, it uses virtually no CPU or memory.

  • Not yet, but multi-user support is planned for late 2025. Each person will get their own private memory space.

  • Yes. Just delete the Penny folder and everything disappears forever.

    No cloud backups or hidden copies exist anywhere.

  • Currently Penny works through direct conversation, but calendar/email integration is on the roadmap for late 2025.

  • Not yet, but community sharing features are coming. You'll be able to share specific insights while keeping personal details private.

  • Community support through the Heumans community, plus direct access to Murray and the development team during the early adopter phase.

  • Yes! Early adopters get direct input on development priorities. Your feedback literally shapes what gets built next.

  • Just opt in during setup to contribute anonymised insights about how your mind works.

    Help advance neurodiversity research while improving Penny for everyone.

  • Then we’ve let you (and ourselves) down.

    Penny was built to adapt to every kind of mind, so if that’s genuinely how you feel, please let us know what we can do better.

  • Since everything runs locally, Penny continues working forever.

    Plus we're committed to open-sourcing core components if we ever shut down.

  • Professional collaboration features are on the roadmap.

    Therapists could use Penny to help clients understand their own thinking patterns.