ProjectNew
A Multimodal, Polyglot, and Local AI Agent for In-Depth Research.
Designed to mitigate IP risks and data leaks, ProjectNew utilizes a closed-feedback Recursive Research Framework to securely process literature, images, and code without running out of context or losing your progress.
About ProjectNew
When we create a folder, the default name is often "New Folder". The name of this Agent comes from the same concept, but with the hope that every time people use it, they don't just create a new project or experiment, but discover new, innovate new, and invent new.
Solving the AI Research Dilemma:
- •The IP Risk: Prevents proprietary chemical formulas, patient data, and patent-pending R&D from being fed into models that "learn" and "replicate" for others.
- •The Cloud: Avoids tethering to external servers where every query is a potential data leak.
- •The Crawl Concern: Ensures your research doesn't become "training data" for a competitor’s future model.
- •The References: Eliminates the issue of reading too much literature causing a model to run out of tokens, or the researcher running out of time.
Recursive Research Framework
- •For every run started by the user, previous iterations' summaries, or previous periods' summaries, are attached alongside a continuous project lifecycle summary so everything remains in context.
- •Necessary literature summaries, images, codes, and formulas are fed by the user to make the agent more accurate.
- •This cyclic nature and close-feedback loop provide more robustness than standard chatbots that might hallucinate with increasing tokens for numerical values, code, and natural language.
Why ProjectNew?
Multimodal
- • LLMs for documents, texts, and codes.
- • VLMs for images.
- • Ex: Gemma3-27b, PaliGemma, Code Llama
Local
- • Operates entirely on single workstations, distributed systems, and servers.
User Control
- • User decides what model to use and what to do with the code.
- • No autonomous decision-making.
Polyglot
- • Seamless transition between Python logic, R statistics, and natural language research notes.
Data Privacy
- • Manage your own database.
- • 100% No Cloud.
Recursive
- • Continue exactly where you left off without worrying about token limits or loss of context.
Use Cases
Literatures
AI speeds up literature review by summarizing all reference materials collected by the user in minutes—a task that takes humans hours for a single paper.
Images
Large amounts of imagery like spectrums, cell photographs, and medical scans can be analyzed rapidly for the user to cross-verify.
Analysis
Based on theory, codes can be created and orchestrated by the agent for rapid analysis. Researchers with no prior programming knowledge benefit immensely.
Documentation
After every iteration, period, or phase of an experiment, automated documentations and summaries are prepared to keep things concise and together.
Plans
| Feature | Individual | College / University | Enterprise |
|---|---|---|---|
| Cost | Free | TBD | TBD |
| Access To | Personal researcher license | Department & campus licensing | Custom deployment & SLAs |
| Deployment | Local install, single user | Managed instances for labs | Dedicated support & integrations |
| Support | Community support | Priority support | 24/7 enterprise support |