NASA fills Roman's first science queue with 118 programs chosen before launch

The mirror is 2.4 metres, the same class as Hubble's, and that number does more work than it looks like it does ^[4][3]. Seeing about a hundred times more sky in a single exposure is not seeing a hundred times deeper ^[6]. Roman buys area, not sensitivity per pixel, so its natural currency is population statistics: how many of a rare thing exist, and at which cosmic epochs.

The data figures at least hang together. 1.4 terabytes a day across 365 days is 511 terabytes, which sits comfortably against the roughly 500 terabytes a year NASA officials quote ^[7][8][1]. Dividing that annual figure by the 35-plus years of Hubble output it is said to match puts Hubble's long-run average near 14 terabytes a year, making Roman's rate about 35 times that ^[2][3]. Hubble's output was never constant, so read that as an average over a career, not an instrument-to-instrument comparison.

What the volume does not tell you is whether anyone can use it. For a General Investigator team, the scarce resource is unlikely to be photons; it is cross-matching, calibration and source detection at the depth their proposal assumed. NASA's own framing calls the wide field both a unique capability and a data management problem ^[6][16]. Telescope time was the thing you competed for with Hubble. With Roman, the competition looks more like software staffing inside each of the 118 groups.

The little red dot program is the sort of design I like, because it can lose. Vasily Kokorev's team at UT Austin starts its search two billion years after a big bang dated to 13.8 billion years ago, asking whether these objects genuinely fade with cosmic time or were simply missed ^[11]. The two hypotheses make different predictions: if the dots harbour the huge stars or black holes thought to seed globular clusters, their numbers should hold steady across cosmic history; if they are black hole stars feeding inside dense gas, the counts should fall as the fuel disperses ^[12]. JWST cannot settle a counting question over large volumes, which is what Kokorev means when he likens the attempt to using enormous binoculars to read a book held in front of your face ^[13].

Steven Finkelstein's program takes the complementary route, hunting very bright distant galaxies that other observatories can then dissect ^[14]. GN-z11, found by Hubble in 2016 at 400 million years after the big bang, is the reference case: JWST has since seen further, but GN-z11's luminosity is what makes chemical follow-up practical ^[15].

One ambiguity sits in the source itself. The 118 are described as observer programs, while Kokorev's is described as working from Roman's data, and many of the selected programs are expected to capitalise on the survey stream rather than command their own pointings ^[1][2]. Until that split is published, I would call this agenda queued rather than locked.