Tinfoil Hat White Paper

A personal AI of your own

AI is becoming part of everyday life. It helps us think, write, learn, and solve problems. As those conversations become more personal, a simple question matters more:

Who controls your AI?

Tinfoil Hat is being built for people who want that control themselves. A personal assistant running on your own computer, with conversations and memory that stay with you, no per-token inference fees, and an approach centered on helping you pursue your goals.

Our aim is to make local AI practical for everyday users. You should not need to become a machine-learning engineer to have an assistant of your own.

Privacy comes first

A personal assistant becomes more useful as it understands your preferences, projects, and priorities. You should be able to benefit from that context without handing your private conversations to a hosted AI provider.

Tinfoil Hat’s core experience uses local processing. Your computer runs the model and stores your conversations and personal memory. You decide what your assistant remembers, and you can review, edit, or delete that information.

Privacy is the foundation of the product.

A cooperative advocate

Tinfoil Hat is designed to work with you: helping you explore ideas, solve problems, and accomplish what you set out to do.

We plan to use abliterated/uncensored models, with additional fine-tuning as needed, to create a cooperative personal advocate.

Abliteration modifies a model to reduce learned refusal behavior. Uncensored, in this paper, describes models selected and adapted for greater freedom of interaction and fewer content-based restrictions.

The goal is an assistant that respects your choices and focuses on useful help without unnecessary lectures or moralizing.

Local power. Remote access.

Use it without a meter

Local AI changes the economics of using an assistant. Once you have the software and suitable hardware, there is no inference provider charging for each message or token.

Run your local AI all day for just the cost of power.

Experiment, revise, ask follow-up questions, and work through longer projects without watching a usage balance.

Software purchase, optional paid services, and upgrades are separate. Performance depends on your computer and the model you choose.

The initial product

The initial release is planned around a focused set of capabilities:

• Two local model options: A lightweight model for more modest computers and a more capable model of approximately 9 billion parameters for suitable hardware.
• Personal conversations and memory: An assistant that can retain useful context, with controls to review, edit, and delete what it remembers.
• Windows desktop application: A consumer-friendly interface for working with your local AI.
• Remote Android access: A companion application that connects to the assistant running on your computer.

Remote access will use a user-managed Tailscale connection. Your computer remains the home of your assistant and must be powered on and connected for remote use. Tailscale is a separate service with its own availability and terms.

This initial scope gives Tinfoil Hat a useful foundation that can grow with customer interest and available resources.

Ownership and payment

Own your version

Purchasing Tinfoil Hat gives you permanent use of the software version you acquire. Continuing to use that version does not require an ongoing Tinfoil Hat subscription.

An optional update subscription provides access to download subsequent versions while the subscription is active. Versions acquired during that period remain yours to use after the subscription ends.

If your subscription lapses and a new version is released during the gap, accessing newer versions requires another purchase at the entry price. If no new version is released during the gap, you can resume the update subscription without another entry purchase.

Your existing acquired versions remain usable either way.

The Tinfoil Hat token

For the initial release and the immediate future, the Tinfoil Hat token on Solana will be the only accepted payment method for purchasing the product.

The token connects the project’s community with a concrete use: purchasing access to the software.

Payment grants a software license. Simply holding tokens does not activate the product or grant access to paid versions.

Token allocation and vesting

The planned allocation reserves 18% of total supply for founder compensation, outreach, and testing.

¹ Token quantities assume a total supply of 1 billion.

These figures describe the intended allocation. Final holdings and launch distribution will be documented through the actual on-chain transactions.

Founder and development compensation: 12%

This allocation supports the founder’s work on Tinfoil Hat, including delivery of the initial two-model product and remote access.

The full founder allocation is planned to be placed in an on-chain vesting lock with 24 equal monthly unlocks.

Each monthly installment releases:

• 0.5% of total supply
• 5 million tokens, at a total supply of 1 billion
• One twenty-fourth of the founder allocation

The remaining installments stay locked until their scheduled release dates. Unlocked tokens may be sold to support the founder’s work and living expenses.

The schedule releases fixed quantities of tokens. Their dollar value depends on market conditions and available liquidity.

Once established, the lock address, beneficiary, release dates, and control terms will be published so anyone can verify the arrangement.

Outreach and participation

Advertising and outreach: 3%

A useful product needs to reach the people who can benefit from it.

This allocation supports advertising, promotional content, community campaigns, and relevant partnerships intended to bring Tinfoil Hat to prospective users.

Tokens may be distributed for these activities or sold to cover their costs.

Testing and community participation: 3%

Local AI needs to work across different computers and real-world situations. Community participation can help identify problems, improve usability, and guide practical improvements.

This allocation supports structured feedback, bug reporting, compatibility testing, and defined community testing programs.

Program details and eligibility will be announced as opportunities become available. Rewards will be tied to the terms of those programs rather than token ownership alone.

The advertising and testing reserves are separate from the founder’s vesting schedule. Their wallet addresses and any applicable locks or spending restrictions will be disclosed separately.

Building in public

Tinfoil Hat is currently in development. Our immediate focus is delivering the initial product described in this paper and showing how it works in practice.

As additional capabilities become ready to demonstrate or release, we will introduce them through project updates. This paper covers the announced scope rather than an extensive list of speculative future features.

We want progress to give people a reason to keep paying attention: working software, useful improvements, and a clearer picture of what personal AI can become when its owner is in control.

Scope and transparency

The initial features described here are planned capabilities until released. Additional development will be guided by adoption, feedback, technical feasibility, and available resources.

Local processing provides control over where the core AI runs and stores its information; device security and any optional external services also affect privacy.

The token’s stated purpose is payment for the software. It does not grant equity or revenue-sharing rights, and this paper makes no promise of token appreciation or market liquidity.