Launch offer — Full free access. Create your account →
The project

About SPORE

Who is behind this project

SPORE is designed, built, and operated by Benoît Baqué de Sariac from Nantes. Solo developer, engineer by training, polymath by temperament. A structured side project, hosted and maintained autonomously.

No team, no fundraising, no institutional badge. One person who codes, reads, tunes the prompts, pays the server, and watches the hypotheses come out. This transparency is constitutive of the project: SPORE claims scientific rigour, and rigour begins with naming who speaks.

If you wish to discuss the project, send a comment, flag an error, or imagine a collaboration: an email is enough.

Why SPORE exists

The most disruptive scientific advances often arise at the intersection of distant domains. Penicillin (microbiology × mould chemistry), CRISPR (bacterial immunology × genome editing), PageRank (academic citations × linear algebra), AlphaFold (bioinformatics × deep learning) — each one born from the meeting of distant disciplines, yielding advances no isolated field could have produced.

SPORE rests on a wager: to automate the detection of these unexplored intersections. Not to replace scientific work — an LLM does not run experiments and validates nothing against the real — but to generate, continuously, the plausible bridges no one has had the time or reflex to formulate. A structured serendipity engine.

What SPORE does, concretely

Each cycle, the system:

  1. Draws two scientific domains at random from a corpus of 500 (sourced from OpenAlex)
  2. Generates, via LLM (DeepSeek + Claude), a hypothetical causal bridge between the two
  3. Submits this bridge to a critique pipeline: devil's advocate, methodologist, domain expert, panel of 5 AI reviewers
  4. Verifies every bibliographic reference through the Semantic Scholar API (zero hallucination tolerated)
  5. Publishes the result as a structured brief: formal hypothesis, three-phase experimental protocol with quantified GO/NO-GO criteria, falsifiable predictions, clickable evidence base

Across 2,095 collisions explored, 38 briefs published. The 98.2% kill rate is public and assumed. The mean cost per brief is $0.51, for a cumulative total of $19.49 — figures available in real time on the Statistics page.

What SPORE is not

  • Not a scientific discovery tool in the classical sense. SPORE produces hypotheses, not results. Hypotheses must be confronted with reality by human researchers.
  • Not a public-facing science media outlet. The French vulgarisation layer aids diffusion, but the product addresses primarily researchers, R&D engineers, working polymaths, and PhD students.
  • Not a peer-reviewed scientific publication. Each brief carries a badge: "AI-generated hypothesis · Pre-publication · To be tested experimentally". This is the epistemically correct stance.
  • Not a funded or affiliated project. No badge, no oversight, no hidden agenda. Just a curious developer who believes analogical thinking can be tooled.

Acknowledged limitations

SPORE crosses texts about science, not science itself. The scientific literature is a filtered, delayed, and socially constructed representation of reality. SPORE can identify gaps in the scientific discourse, not in nature. The bridges it produces are plausible in a linguistic sense — not necessarily productive in an experimental one. This distinction is central, and SPORE owns it.

The five AI reviewers do not constitute independent validation: they are projections of the same linguistic representational space as the generator. They detect internal inconsistencies, not empirical contradiction. The only validation that matters is experimental, and it lies beyond SPORE's perimeter.

How to support the project

SPORE is free during the launch phase. In the longer term:

  • Single brief €9 — pay-per-use, unlimited access to the purchased brief.
  • Thematic subscription €15/month — 5 briefs/month on domains of your choice (in preparation, planned for June 2026).
  • Custom collision €25 — propose two domains, SPORE generates an unprecedented hypothesis with protocol and bibliography in 24 hours. First collision offered during launch.
  • Receive the SPORE Anthology — a free PDF of the eight strongest hypotheses from the first six months, sent by email here.

For research teams or corporate R&D groups interested in custom integration: an email to benoit@spore-research.com is enough to start the conversation.

One last thing

A well-documented dead end is worth more than a glib unification.

This is the line that guides SPORE. Success is not measured by the number of briefs published, but by the quality of the hypotheses generated and the honesty of their status. Each brief is an exploratory proposition, delivered with its doubts intact.

Page updated: May 2026.