Good questions.
Clear answers.

We’re building CASP to develop designs, software, scientific proposals and business improvements. Bring a goal, an idea or something you’ve built. Investigate different approaches, guide the work and continue developing the same project as your needs change. The intended outputs are editable so you can carry them into your existing tools.

Continuum is the research and learning engine behind CASP. We’re developing it to bring continuous research and accumulated knowledge into every project, helping CASP discover new possibilities and develop stronger ideas over time.

CASP is being built around a continuing project: investigating unfamiliar routes, developing the work and carrying it forward as requirements and evidence change. The aim is to bring that experience together, in CASP or through supported tools. The commercial experience is still in development.

Deeppoint is our scientific research direction. We’re developing it to investigate prior work, explore potential contributions and develop scientific proposals and proposed experiments. Proposed findings still need the relevant expert review and experimental validation.

The research engine exists. The commercial experience is still in development, and general availability has not been announced. You can explore the planned access options and register your interest.

The Discover and Improve examples are prepared, interactive illustrations of how a project could develop. They use real sources, but do not run a live search or generate new research.

We’re developing a standalone workspace, access through supported assistants, an API for products and an organizational experience. The intended experience lets you continue the same project and developed work across supported interfaces, with editable outputs you can use in your existing tools. Availability and pricing have not yet been announced.

We’re exploring commissioned research and technical training and evaluation datasets for AI companies. These would be developed to a buyer’s requirements, including original problems, verified solutions, supporting sources and scoring criteria. Contact us to discuss the work you need.

The commercial service and its privacy controls are still in development. Our intended policy is to use project information for the work you authorize; broader use would require a separate agreement. We’ll explain the applicable terms before you share private project material. The early-access form asks for your email and a short description so we can reply. Please leave confidential details out of that message.

WANDR evaluates research that must search widely and substantiate findings across sources. Soft F1 measures research quality and coverage. Hard F1 requires a credited result to meet all its evidence requirements.

CASP recorded 0.583 Soft F1 and 0.493 Hard F1 across 45 tasks. These are the means of its best preserved official evaluator results collected during development. Other systems’ scores are separately published references, not a controlled head-to-head test. See the results and methodology or read the WANDR paper (opens in a new tab).

Caspian Labs is being built by Atrin Farnamfar. His experience includes AI research with Professor Katri Nousiainen, customer discovery through Princeton-led NSF I-Corps and building and shipping software products. Get in touch with Atrin.

Choose “Request early access” on the homepage and send your email and what you would like to create or improve. You’ll receive confirmation on the website when your request has been submitted. We’ll review your goal and contact you about early access and next steps.

You can also email atrin@caspianlabs.co directly. Registering interest does not create an account or commit you to a purchase.