Bionoculars: Early Access

We are excited to release the early access version of Bionoculars.

Bionoculars is a life science search engine designed to create a space for scientific exploration that builds on our users' experience and intuition.

Our main ambition: build a transparent, exhaustive AI-assisted research tool that empowers scientific reasoning rather than automating it away.

Why we built this

This project was born from four observations about the current state of life science research.

1. Life science has its own complexity. Anyone who has searched for a gene, a protein, or a disease in the literature knows the problem. The same concept can appear under many different names: a gene like "prion protein" can be written as "prnp", "Prn-p", "CD230", or "PrPL-P1-like" depending on the author and the context. A variant like "V600E" might also show up as "1799T>A" or "rs113488022". On top of naming, biological concepts are deeply interconnected: a gene relates to a pathway, which relates to a disease, which connects to a drug, and so on. Navigating this web of relationships manually across thousands of articles can lead to a loss of research time.

2. AI is introducing measurable biases into research. A landmark 2026 study in Nature, analyzing 41.3 million research papers, found that while AI adoption gives individual scientists a significant productivity boost (3x more papers, 4.8x more citations), it collectively narrows the scope of science, shrinking the volume of topics studied by 4.63% and reducing scientists' engagement with one another by 22% (Hao et al., 2026). The authors concluded that AI tends to automate established fields rather than explore new ones. This echoes earlier warnings about "illusions of understanding" where AI tools create scientific monocultures in which dominant methods and viewpoints crowd out alternatives (Messeri & Crockett, 2024). Even more concerning, recent work shows that LLMs systematically favor content generated by other LLMs, creating a self-reinforcing cycle when used in evaluation or selection roles (Laurito et al., 2025).

3. Current tools force a false choice. Researchers today are stuck between two extremes. Classical tools like PubMed are exhaustive (over 40 million indexed citations) but demand enormous manual effort. On the other end, AI-powered tools select at most a few dozen articles behind a black box, generate a summary, and ask you to trust it. One gives you everything but little help navigating it, the other gives you a shortcut but strips away your ability to explore.

4. Data sovereignty matters. American tech giants currently control more than 70% of Europe's cloud infrastructure. The question of where research data lives is gaining attention: in December 2025, the European Commission published a paper on enhancing data sovereignty for research, highlighting the need for European control over research data and infrastructure.

What Bionoculars does today

Even in early access, Bionoculars is already a useful tool. It searches across 80M+ life science articles and experiments with methods that try to address the problems above:

• Keyword groups replace the rigid boolean formulas of PubMed with semantic clusters powered by the UMLS Metathesaurus data. You see why articles are ranked the way they are, and you can edit that logic directly.

• AI actions take a deliberate step back from full automation. Instead of letting AI choose what you read, you choose, then ask the AI to help you synthesize. Every claim is cited. Every source is traceable.

• A knowledge graph maps real relationships between concepts, each link backed by actual articles improving over simple co-occurence based techniques.

Along with this comes a long-term ambition to one day build a reliable, article-sourced knowledge graph from scratch. But let's not get ahead of ourselves.

Why early access, and why now

There is always more left to do, and if we wait until everything is perfect, we will never release. So we decided to start with a limited early access offer.

This offer comes at a low price (€6/month or €60/year) because our first users are the foundation of what Bionoculars becomes. You will help shape the tool, uncover new ways to use AI for science, and push us to get better, faster. But also because we are honest: the product still has rough edges (potential bugs, incomplete index on some closed-access articles).

We are a self-financed, small team. Any participation in the project, whether financial through a subscription, or intellectual through feedback, is of great help. We want to make it worth your time and your money.

References