Special Edition · The Week Ahead · Monday, Wednesday and Friday, 28 September to 2 October 2026
Next week, The Signal goes deep again. Three long reads from the frontier, Monday, Wednesday and Friday, each one a story we could not stop pulling at. Below is only the setup for each. What we found, and why it changes things, is in the piece.
COMING THIS WEEK
Mon · AI: One AI Just Read 146 Diseases From a Single CT Scan
Wed · Energy: China Just Broke Ground on Fusion Nobody Else Will Touch
Fri · Biotech: Where Two Drug Giants Failed, This Startup Just Won
One AI Just Read 146 Diseases From a Single CT Scan
AI · Drops Monday
A radiologist reading an abdominal CT is hunting through eighteen organs for one of dozens of diseases, and most AI tools are built to find exactly one of them. On 18 September, Alibaba's research arm published a model trained to find 146 at once, tested against a panel of 26 radiologists, then gave the code away for free.
It never saw a doctor point at a single tumor. It learned from the reports hospitals already write and were about to file away.
It held up on nearly 40,000 real scans, but almost none of them came from the hospitals a global rollout would actually need.
China Just Broke Ground on Fusion Nobody Else Will Touch
Energy · Drops Wednesday
Every major fusion company racing to a working reactor this decade is chasing the same reaction, deuterium and tritium, fused at 150 million degrees. On 15 September, China's ENN Group broke ground on a platform built to chase a much harder one instead, the one reaction that produces no stray neutron at all.
The reward is a reactor with no neutron problem. The price is a plasma twenty to thirty times hotter than anyone else is attempting.
Nothing has fused yet. The number that matters is not the groundbreaking, it is a temperature nobody has proven anyone can hold.
Where Two Drug Giants Failed, This Startup Just Won
Biotech · Drops Friday
Biogen and Johnson and Johnson each spent years chasing the same gene therapy for an inherited form of blindness, and each one missed its main goal. On 21 September, an 85-person company called Beacon Therapeutics said its version hit the mark.
All three companies used almost the same toolkit. The honest question is whether Beacon built a better drug, or just picked an easier finish line.
Investors reacted immediately. The one number that would settle the argument, how many patients were actually in each arm, still has not been released.
Three long reads a week, Monday, Wednesday and Friday. Frontier Signal explains frontier technology in plain English.