We were able to go deeper than what we expected.
Across a sliver of sky no wider than three full moons, humanity has now cataloged nearly 800,000 galaxies from the universe's earliest chapters — ten times more than anyone expected to find. The James Webb Space Telescope's COSMOS-Web survey, assembled from over 10,000 exposures and more than 200 hours of observation, represents not just a feat of engineering but a reckoning with how little we understood about the infant cosmos. Released freely to researchers worldwide, this archive invites a generation of scientists to ask questions its creators have not yet imagined.
- Astronomers expected a modest haul of early-universe galaxies — instead, JWST returned ten times that number, upending assumptions about how populated the young cosmos truly was.
- Building the mosaic demanded two years of painstaking data processing, as a brand-new telescope introduced noise, distortions, and artifacts that had to be untangled image by image across 10,000+ exposures.
- The team developed new techniques to measure how light spreads across galaxies, allowing sharp space-based images to be combined with blurrier ground observations without sacrificing precision.
- An interactive, machine-learning-equipped catalog has been released free to any researcher on Earth, democratizing access to what may be JWST's largest contiguous image ever taken.
- The survey's deepest ambition — mapping the Reionization Era's hydrogen-clearing bubbles from 13 billion years ago — remains an open frontier, with the dataset now positioned as the instrument to pursue it.
On Thursday, astronomers unveiled the James Webb Space Telescope's most expansive view of the cosmos: a catalog of nearly 800,000 galaxies spanning a patch of sky no wider than three full moons side by side. The COSMOS-Web project consumed more observation time in JWST's first year than any other effort — over 200 hours — and required stitching more than 10,000 individual exposures into a single contiguous image by a global team of nearly 50 researchers.
The results immediately defied expectations. Scientists had anticipated finding a modest number of early-universe galaxies within the survey area. They found ten times as many. "It was incredible to reveal galaxies that were previously invisible at other wavelengths," said postdoctoral researcher Maximilien Franco of the University of Hertfordshire. Project lead Jeyhan Kartaltepe of the Rochester Institute of Technology suggested JWST may never survey an area this large again, calling it a reference dataset that researchers will draw on for years.
What elevates the project beyond its scale is its openness. The team released an interactive catalog — free to any institution — equipped with machine learning tools to extract each galaxy's size, brightness, shape, and surrounding environment. That last detail matters: galaxies do not evolve in isolation, and mapping nearly 800,000 of them allows astronomers to study how cosmic neighborhoods shape galactic growth at a previously impossible scale.
The data took two years to process. As a new observatory, JWST introduced unfamiliar noise patterns and distortions requiring careful correction. The team also developed new brightness-measurement techniques, modeling how light spreads across a galaxy rather than simply summing it within a fixed boundary — a precision that allows space and ground-based observations to be combined without losing detail.
The survey's central ambition points toward the Reionization Era, more than 13 billion years ago, when the first galaxies began burning through the thick hydrogen fog of the early universe. The team plans to use the catalog to measure the size of "reionization bubbles" — vast clearings carved by ancient starlight. That work is ongoing. Already, several papers have emerged from the dataset, including one tracing how the brightest galaxies at the hearts of galaxy clusters have changed over 12 billion years.
Perhaps most telling: JWST outperformed its own pre-launch models. "We were able to go deeper than what we expected," Kartaltepe said. The nearly 800,000 galaxies now sit in a public database, waiting for researchers around the world to pose questions the COSMOS-Web team has not yet thought to ask.
On Thursday, astronomers released what may be the James Webb Space Telescope's most ambitious single view of the cosmos: a catalog of nearly 800,000 galaxies stretching across a patch of sky no wider than three full moons placed edge to edge. The image came from COSMOS-Web, a project that consumed more observation time in JWST's first year than any other undertaking—over 200 hours awarded to a global team of nearly 50 researchers. To assemble it, the team stitched together more than 10,000 individual exposures into what stands as the telescope's largest contiguous image to date.
The sheer breadth of the survey yielded a surprise. Astronomers had anticipated finding a certain number of galaxies in the early universe from this observing area. Instead, they found ten times as many. "It was incredible to reveal galaxies that were previously invisible at other wavelengths," said Maximilien Franco, a postdoctoral researcher at the University of Hertfordshire, describing the moment the data began to resolve on computer screens. The abundance caught even seasoned scientists off guard. Jeyhan Kartaltepe, the Rochester Institute of Technology astrophysicist leading the effort, told observers that the telescope may never again survey an area this large. "I think it'll be a good reference and a good data set that people will use for many years," she said.
What makes the dataset extraordinary is not just its size but its accessibility. The team released an interactive catalog free to any researcher at any institution, complete with machine learning tools designed to extract physical properties from the galaxies—their size, brightness, shape, and the cosmic environments in which they formed. This matters because galaxies do not evolve in isolation. Whether a galaxy sits alone or nestled in a crowded region shapes how it grows, how it forms stars, and what it becomes. By mapping nearly 800,000 of them across a wide swath of sky, astronomers can now study these environmental influences at a scale previously impossible.
The work required more than raw observation time. The team spent two years processing the data, a meticulous effort to align and clean more than 10,000 images. As a brand-new observatory, JWST introduced unexpected challenges—noise patterns, distortions, and artifacts that had to be carefully corrected before the data could be trusted. The researchers also developed new techniques to measure galaxy brightness more accurately, modeling how light spreads across a galaxy rather than simply summing it within a fixed boundary. This precision allows them to combine JWST's sharp images with blurrier ground-based observations without losing critical detail.
One of the project's central goals was to map the Reionization Era, a period more than 13 billion years in the past when the first galaxies ignited and began clearing the thick hydrogen fog that filled the early universe. The team plans to use the nearly 800,000 galaxies as tracers, measuring the size of "reionization bubbles"—vast regions where starlight carved clearings in the primordial haze. "That's not something we finished yet," Kartaltepe acknowledged, but it remains the driving ambition behind the survey. Already, the team has published several scientific papers exploring the data, including one that examines the brightest galaxies at the centers of galaxy groups, tracing how their structure and star-forming activity have evolved over the past 12 billion years.
Perhaps most striking is that JWST performed better than pre-launch models predicted. Engineers had estimated how faint and distant the telescope could see. The reality exceeded those projections. "We were able to go deeper than what we expected," Kartaltepe said. The catalog holds what she calls "incredible potential," though she is careful to note the limits of current knowledge. "There's still so much we don't know." For now, the nearly 800,000 galaxies sit in a publicly available database, waiting for researchers around the world to ask questions the COSMOS-Web team has not yet thought to pose.
Notable Quotes
I don't know if the James Webb Space Telescope will ever cover an area of this size again, and so I think it'll be a good reference and a good data set that people will use for many years.— Jeyhan Kartaltepe, Rochester Institute of Technology, lead researcher of COSMOS-Web
It was incredible to reveal galaxies that were previously invisible at other wavelengths, and very gratifying to finally see them appear on our computers.— Maximilien Franco, postdoctoral researcher, University of Hertfordshire
The Hearth Conversation Another angle on the story
Why does it matter that they found ten times more galaxies than expected?
It means our models of the early universe were incomplete. We thought we understood how many galaxies existed in a given volume of space at certain distances. Finding ten times as many suggests either the universe was denser with galaxies than we calculated, or our detection methods were missing things. Either way, it changes the baseline.
And the reionization bubbles—what are those actually doing?
In the first few hundred million years after the Big Bang, the universe was filled with neutral hydrogen, a kind of fog. When the first stars and galaxies formed, their light ionized that hydrogen, clearing bubbles of transparent space. By mapping where those bubbles are and how big they grew, we can understand when and how the universe became transparent. It's like reading the fossil record of light itself.
Why release all this data for free instead of keeping it proprietary?
Science moves faster when everyone can see the same data. A researcher in Tokyo might ask a question about these galaxies that the COSMOS-Web team never thought to ask. By opening it up, they multiply the number of minds working on the problem. That's the whole point—Kartaltepe said she hopes anyone at any institution will use it for their own work.
The machine learning tools they developed—what do those actually do?
They automate the process of measuring galaxy properties from the images. Instead of a human having to estimate the brightness or size of each galaxy individually, the algorithm does it across hundreds of thousands. It's faster and more consistent, though it still needs human verification. They also developed a better way to measure brightness by modeling how light spreads rather than just adding it up in a box.
Will they ever use JWST to map an area this large again?
Kartaltepe said she doesn't think so. The observation time required is enormous—over 200 hours in a single year. That's a huge commitment for one project. So this dataset may be unique in that sense, which is partly why they're treating it as a long-term reference that will be mined for discoveries for years to come.