Streaming Science: replacing the paper with an AI you can query
Deep-tech fund Quantonation spotlights William Zeng's proposal to stream research into AI models instead of publishing papers.
In an essay relayed by the investor Quantonation, W. J. Zeng argues research could move past the paper: feeding code, data and notes into shared AI models that others query directly. He calls it 'Streaming Science'.
Why it matters
How a field shares knowledge shapes how fast it advances. Coming from an investor and researcher rooted in quantum, this questions whether the paper, and centralized repositories like arXiv, are still the right containers for discovery.
Key takeaways
- Zeng sketches successive eras of science, from correspondence and journals to today's 'augmented' phase and an emerging 'streaming' one.
- The core idea: stop reformatting work into papers, and instead stream raw artifacts (code, data, notes) into AI models that others interrogate.
- He suggests measuring a contribution by the 'compression' it adds to a shared model, leaning on DeSci tools (IPFS, zero-knowledge proofs, Sigstore) for provenance.
- It remains a vision: authorship, attribution and replacing centralized trust are open problems.
Spotted through a post by the deep-tech fund Quantonation, this essay by W. J. Zeng, a familiar name in quantum computing, imagines what comes after the research paper. Its title: From Augmented Science to Streaming Science.
His diagnosis is that packaging results into publications carries a heavy overhead, a storytelling and engineering tax that forces researchers to compress messy, living work into tidy written documents mostly for other humans to read. That friction, he argues, slows discovery.
The proposal: let subcommunities stream their artifacts (code, raw data, lab notes) into shared AI models, which others then query rather than read. A conference could ask the model for speakers or a tailored agenda; a contribution might be judged by how much it improves the model's ability to compress what's known. To handle trust and provenance without a central authority, he points to decentralized-science building blocks like IPFS, content hashing and zero-knowledge proofs.
The honest reading: this is a provocation, not a finished system. Attribution, authorship and whether decentralized infrastructure can replace centralized trust are all unresolved. But for fields like quantum, where progress is bottlenecked as much by how knowledge circulates as by hardware, it is a signal worth tracking.
Technology
Organization
Country / Region
Keep exploring
France's quantum ecosystem is maturing, says Challenges
The magazine argues French quantum is moving from speculative research to industrial champions, with local investors scaling up.
According to Challenges, quantum computing in France is shifting from a speculative field to one producing industrial champions, with companies such as Pasqal, Quobly and Multiverse Computing, and a local investor, Quantonation, helping structure the ecosystem.
Twelve CNRS quantum start-ups pitch to investors at the MEDEF Tech Evening
On 1 October 2026 in Paris, the fifth MEDEF Tech Evening is devoted to quantum, with the CNRS and Quantonation as partners.
The fifth MEDEF Tech Evening, on 1 October 2026 at the CNRS in Paris, is devoted to quantum technologies. Twelve start-ups born in CNRS laboratories present their work to investors, large companies and public bodies.
La French Tech·10 Sept 2026
A quantum and AI hackathon for defence, at the Grand Palais in December
The French Defence Innovation Agency and Artefact bring engineering students together on 2 to 4 December 2026, with teams selected in October and November.
The Defence Innovation Agency (AID) and Artefact organise a Quantum Defence Hackathon from 2 to 4 December 2026 at the Grand Palais in Paris, during the Adopt AI Summit. Engineering students work in teams of four to eight on quantum computing and artificial intelligence applications for defence.
Agence de l'innovation de défense·5 Sept 2026