AI Is Coming for Math and Energy Research, but the Dossier Shows No Fusion Result

On 29 September, CBC News reported that companies have approached the Newfoundland and Labrador government about building AI data centres. Energy and Mines Minister Lloyd Parrott said the province's door is "open for business." That is the freshest development in this dossier. It is also not a fusion story.

The topic query asks for a fusion energy research result. None of the twelve full-text sources supplied here reports one. The dossier holds no paper, press release or news report about a fusion experiment, a plasma record, a reactor milestone or a peer-reviewed fusion finding. The only fusion material is a list of recent headlines that the instructions explicitly say not to treat as citable facts. This dossier cannot support a fusion research analysis.

What the dossier does contain

The strongest cluster is about AI eating research work. On 29 September, the National Bureau of Economic Research published working paper 35782 by Matthew Schwartz, Isaiah Andrews and Jesse M. Shapiro. The paper describes an open-source workflow that uses an LLM to reproduce, improve and extend economics articles from their replication packages. The team ran it across 4,452 published replication packages from five economics journals. The workflow flagged discrepancies in 3,460 articles or their appendices. In 496 articles it cut a calculation's computation time by more than a factor of ten at similar or greater accuracy. In 923 articles it produced an extension not present in the original. The paper's own disclosure says LLMs were used in the analysis and writing, and that Schwartz worked as a contractor for Anthropic.

Two other 29 September pieces push in the same direction. Stephen Wolfram published an essay on 28 September arguing that AI is useful for mining the mathematical literature but that human imagination still sets the questions. Dan Romik's blog post, also dated 29 September, makes a sharper claim: autonomous AI theorem provers will generate everything "distance one" from their training corpus and then stall, because they cannot distil and generalise what they produce.

Romik's post opens by referring to OpenAI's announcement of a solution to the Navier-Stokes problem. That is a claim inside a source, not a verified result in this dossier, and it should be read that way.

Energy and compute, without the fusion angle

On the energy side, the dossier offers two dated data points rather than a breakthrough. Hannah Ritchie's Substack post on 29 September walks through work by Oxford professor Nick Eyre. Eyre suggests a fully electrified global energy system would need about 40% less final energy: total demand falls from 416 to 247 exajoules, while electricity demand rises from 110 to 189 EJ. It is a model, not a measurement.

Efficient Computer said the same day that it raised a $97 million Series B led by TQ Ventures, taking its total raise to $173 million at a $650 million valuation. CEO Brandon Lucia writes that the company claims 10 to 100 times better energy efficiency than traditional CPU architectures, and that it is launching its Electron E1 chip at volume for embedded physical AI systems.

Oracle, meanwhile, used 29 September to launch Fusion Claw, a governed agentic execution runtime, alongside 25 Claw-powered Fusion Agentic Applications. Despite the name, this is enterprise software, not plasma physics.

The CBC piece ties the strands together. Paris Marx, author of Hyperscale, questioned whether handing electricity to data centres is the best use of it. TechNL CEO Andrea King said Labrador has the low-carbon hydro, cold weather and land that proponents want, but that opportunity costs still have to be weighed. CBC also notes that Meta's planned data centre in Sturgeon County, Alberta, is expected to cost $13 billion and come online in two to three years.

If a fusion result is the assignment, this dossier is the wrong drawer. What it does support is a narrower story: AI tools are being pointed at research itself, and the electricity those tools need is now a provincial-level political question in Canada.