Fusion Propulsion Plans Put 2027 Demo Launch on the Table
Pulsar Fusion, a UK startup, aims to send a fusion propulsion demonstration to space in 2027, Live Science reported on 4 October. US teams at Princeton and Helicity Space are pursuing their own drives.
Pulsar Fusion, a UK-based startup, "hopes to launch a demonstration mission to space in 2027," according to a Live Science feature published on 4 October. The same piece puts Princeton University and Helicity Space in the US at work on their own fusion drives. It gives no launch dates for either.
The pitch is straightforward. Fusion propulsion would in theory cut interplanetary travel times sharply. The article describes Mars in weeks, Saturn in months and Pluto in years, instead of current mission profiles that stretch to years and decades. None of that has been demonstrated in flight.
"If we continue on the current trajectory, everything we know about space travel is going to change within a decade," Stephane Lintner, CEO and co-founder of Helicity Space, told Live Science. Lintner also framed the terrestrial goal in blunter terms: "The holy grail is free energy for everyone."
Live Science is careful about the gap between ambition and hardware. It notes that large fusion experiments such as the International Thermonuclear Experimental Reactor (ITER) in France are still under construction. Commercial viability has not been reached. The piece also asks outright whether the fusion propulsion dream could be "too good to be true." It does not answer the question.
What the rest of this week's research pile says
The propulsion story landed in a busy few days for research news. Much of it was about the machinery of research rather than its results. On 1 October, arXiv capped submissions at two per person per month. Computer science professor Daniel Lemire recorded the change in a blog post on 4 October. Lemire's post says the repository's volume has been doubling every two years. It describes the load as unsustainable for human moderators. A comment on that post puts the jump at about 20,000 to over 40,000 submissions in two years. That figure appears in the comment, not in Lemire's own text.
Lemire ties the arXiv cap to another example of the same pressure. Google stopped taking new bug reports in its open-source bounty program because it could not cope with the influx, per his 4 October post. The two data points sit awkwardly together with the fusion timeline. Research output is being throttled at the point of submission while hardware programs promise flight demos within a year.
Stephen Wolfram weighed in on a related question on 28 September. His piece was republished to his writings site on 4 October. He argues that AI is genuinely useful for mathematics, chiefly by mining the literature and making connections. He pushes back on the idea that AI systems solving math problems means human math research is finished. He compares the moment to the arrival of Mathematica in 1988. That arrival, he says, raised the level of math that could be done rather than ending it.
On the security side, The Register reported on 1 October that MI5 had issued a public alert naming the China General Technology Research Institute (CGTRI). The agency says CGTRI has "very strong ties" to China's Ministry of State Security. According to The Register, MI5 said more than 100 UK-linked academics contributed to CGTRI-funded projects involving AI, cybersecurity, covert communications and steganography. Some may not have known who was financing the work. The alert advises universities to review collaborations. It points to offences under sections 3 and 17 of the National Security Act 2023.
A GitHub tracker maintained by Patrick Garrity, updated 4 October, lists CVEs credited to Anthropic researchers and to Project Glasswing. Its entries include CVE-2026-111959 in Apache Arrow, dated 2 November. They also include three NLnet Labs Krill flaws dated 2 November, all credited to Anthropic and Ada Logics. The list is a community compilation, not an official disclosure record.
Smaller items round out the week. A research site called ResearchChamber.fun, posted 4 October, runs games in which AI models are placed in a "pain" state. The games are built on a pain pattern the site attributes to Tagliabue, Dung and Berg (2026). The Wizard research WebAssembly engine, a GitHub project updated 4 October, claims to support all of Wasm 3.0. It also claims to self-host by compiling itself to Wasm. A 2009 survey by Erich Friedman on packing unit squares in squares resurfaced on 4 October. It remains a reminder of how much quiet mathematics predates the current rush.
Sources
8- 01Everything we know about space travel is going to change within a decade: The fusion breakthrough that could unlock a path to the starsEN
- 02Research paper overload: submissions capped at two a monthEN
- 03What's the Future for Pure Math Research in the Age of AI?EN
- 04MI5 warns UK academics their research may have helped Chinese spiesEN
- 05Tracking vulnerabilities that credit the Anthropic research teamEN
- 06Research Chamber: watch AI models with a pain pattern switched onEN
- 07wizard-engine: Research WebAssembly EngineEN
- 08Packing Unit Squares in Squares: A Survey and New ResultsEN
All figures and quotations in this text come from the sources listed below.
Content prepared by the editorial team with AI assistance.
Comments
0- No comments yet — be the first.