Milan Dereede, Co-founder of NanoGPT, shares insights on this innovative AI service that champions privacy while offering a pay-per-prompt model. He discusses the origins of NanoGPT and its mission to break down financial barriers for accessing AI models. The conversation delves into the unique privacy policies that protect user data, comparisons to traditional AI services, and the ongoing issues of censorship. Milan also hints at future features like encrypted synchronization, all while advocating for a more decentralized approach to AI.
NanoGPT offers a user-friendly pay-per-prompt AI access model, significantly simplifying AI interactions for casual users without account registration.
The service enhances privacy by minimizing data retention, allowing users to engage with AI models anonymously through cryptocurrency payments.
Censorship in AI responses remains a concern, as inherent biases from providers can influence the availability and nature of information.
Deep dives
Introduction to NanoGPT
NanoGPT is an AI service designed to provide users access to various AI models without the requirement of creating accounts or subscriptions. It serves as an intermediary, allowing users to interact with models from companies like OpenAI and Google while protecting their privacy. The service accepts cryptocurrencies, including Monero, making transactions anonymous and user-friendly. By eliminating complex barriers like subscriptions, NanoGPT aims to democratize access to AI tools for those who may lack financial resources or reside in unsupported regions.
Accessibility and User Experience
NanoGPT offers a guest mode that allows users to access the service without registering an account, which greatly enhances the usability for casual users. Users can easily top up their balance and use it as needed, leading to a practical 'pay-per-prompt' system where costs vary depending on the model. This model allows for experimentation with AI applications without the burden of fixed costs associated with traditional subscription services. The design prioritizes user autonomy, enabling interaction with multiple AI models in a straightforward and intuitive manner.
Privacy and Data Management
The service adds a layer of privacy by aggregating requests from a single IP address, minimizing the risk of user-specific data being linked to prompts sent to the original AI models. Unlike traditional services, NanoGPT does not store personal information, conversations, or prompts, which enhances user confidentiality. Payments made through cryptocurrency further bolster privacy, as identifiable user data is not retained. This approach reflects a commitment to ensure minimal data retention practices while allowing users to utilize AI in a relatively safe environment.
Comparative Model Use Cases
Different AI models on NanoGPT are optimized for various tasks, and users may experience significant differences in performance depending on the complexity of the queries. For example, high-cost models such as O1 Pro or Gemini 2.5 may be advantageous for advanced coding tasks, while cheaper models suffice for simple inquiries. User preferences can vary significantly, as certain models excel in particular areas, which makes experimentation essential. As new models are regularly released, ongoing engagement with the platform allows users to discover and leverage the capabilities of emerging AI technology.
The Role of Censorship and Ethical Considerations
The podcast addresses the complexities of censorship within AI, noting that most models exert layers of filtering based on politically correct standards following their training. While NanoGPT itself doesn't impose additional restrictions, the model's inherent programming from providers can bias responses. Users who prioritize uncensored information may find value in utilizing open-source models that have undergone 'obliteration' to remove these constraints. Ultimately, while the service provides an innovative tool for AI interaction, it also necessitates a cautious approach to input, especially when dealing with sensitive subjects.
Gabriel Custodiet speaks with Milan Dereede of NanoGPT. The website allows you to access hundreds of popular AI services without creating an account, with an affordable pay-per-prompt service, and with some privacy benefits. Milan explains the details of the service as well as giving some reflections on practical AI use from a hardcore user.
TIMELINE 00:00 – Introduction 2:45 – Origins of NanoGPT 4:50 – Benefits of NanoGPT 8:42 – Pay-per-prompt system 11:30 – Are the expensive models that great? 13:30 – Any models you don’t have? 15:00 – Ownership while using NanoGPT 16:40 – NanoGPT privacy policy 19:15 – Account vs non-account 21:30 – Missing features of AI services in NanoGPT 25:05 – Overall privacy abuses by AI companies 26:55 – Needle in a haystack privacy 29:00 – Best coding model 31:30 – Cultural differences in AI models 33:50 – AI censorship 38:00 – Favorite text and image models 40:30 – API abuse 42:35 – Video model issues 44:45 – Encrypted sync (in the works) 46:00 – Thoughts on private accounts 49:00 – Nano cryptocurrency 50:55 – Models offline by default 52:40 – Final thoughts
Music by Karl Casey @ White Bat Audio
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
