From 27 to 31 July 2026, I attended the Rubin Community Workshop at the SLAC National Accelerator Laboratory in Menlo Park, California. My presentation, “An Automated Three-Method Asteroid Rotation-Period Pipeline for Rubin First Look Data,” was part of the Solar System Science session. I presented an automated pipeline that combines Lomgscargle multiband periodogram, high order fourier fitting, and Phase Dispersion Minimization to identify the rotation periods of asteroids observed through the Rubin telescope and assign confidence levels. The workshop brought together members of the international Rubin community as the observatory moved from construction and commissioning into active survey operations, with sessions covering the first Rubin datasets, data processing, survey strategies, and the scientific discoveries the LSST has enabled. 

SLAC Laboratory Director John Sarrao opened the workshop by welcoming the Rubin community and reflecting on the observatory’s transition from construction project into active survey operations. He framed the moment through his personal motto, “defining the future and navigating the present”, which set the tone for a week ahead by emphasizing the extraordinary scientific future Rubin can help create. After decades of work to build the observatory, he described the start of the survey as the beginning of an exciting new decade for astronomy.

AT-LAST in Brief

I presented AT-LAST (Asteroid Time-domain Lightcurve Analysis for Spin Tracking) during the Solar System Science session. I began developing AT-LAST to address a central challenge of the Legacy Survey of Space and Time: Rubin will obtain sparse, multi-band measurements for millions of asteroids, but manually inspecting every lightcurve will be impossible. Rather than forcing a period for every object, AT-LAST is designed to report a rotation period only when several independent methods and quality checks agree.

The pipeline combines multiband Lomb–Scargle, a higher-order Fourier fit, and the model-free Phase Dispersion Minimisation method. A result is classified as high confidence only when the three methods agree on the same period or an accepted harmonic, and the object passes seven gates covering phase coverage, observation count, periodogram strength, Fourier significance, PDM dispersion, amplitude signal-to-noise, and frequency uncertainty. Applied to 5,052 asteroids from Rubin First Look and DP2, AT-LAST identified 523 high-confidence rotation periods. In a comparison with 33 asteroids for which Greenstreet et al. reported a single preferred solution, AT-LAST matched 32, giving a 97% agreement rate.

The catalogue also recovered three previously reported ultra-fast rotators with periods between 1.87 and 3.77 minutes and identified four additional high-confidence candidates awaiting confirmation. Every analyzed asteroid along with its periodograms, folded lightcurves, intermediate measurements, and quality-gate results is available through my public portal at atlast.monitormyplanet.com.

From Construction to Survey

Rubin’s leadership team presented results from the first weeks of survey operations, including thousands of visits across the sky and images approaching the observatory’s target performance. They reviewed the milestones that made this possible: completion of construction, handover into early operations, extensive system testing, and the formal beginning of the LSST on 29 June 2026.

Work is continuing to improve observing efficiency, stray-light correction, active-optics performance, and the consistency of image quality across Rubin’s enormous focal plane. A major operational challenge was the severe storm that struck the Coquimbo Region of Chile earlier in July, damaging roads, bridges, and access routes to Cerro Pachón. Rubin’s facilities were largely intact, but the summit lost power and physical access had to be restored before normal observations could resume.

The workshop also coincided with the release of the early version of Data Preview 2. DP2 contains approximately 30,000 visits acquired between April 2025 and January 2026, covering around 15,000 square degrees of sky. Deep coadded images were produced across roughly 3,000 square degrees, accompanied by source catalogues, time-series measurements, and Solar System tables.

Melissa Graham emphasised that the coverage is highly non-uniform. Different fields have different numbers of observations, filters, limiting depths, image quality, and average observing dates. These variations must be considered before drawing conclusions from the data, particularly for time-domain studies. A fuller DP2 release planned for later in 2026 will add calibrated visit images, difference images, cutouts, and raw exposures. Data Release 1, based on the first full survey year and the subsequent processing campaign, is currently expected around the middle of 2028.

Turning Alerts into Science

Rubin is expected to produce millions of alerts on many observing nights. The challenge is therefore not simply detecting change, but determining which events are scientifically important and getting that information to researchers quickly enough for follow-up.

The Transients and Variable Stars Science Collaboration focused heavily on Rubin’s Target of Opportunity Program, which allows the observatory to interrupt its regular survey when a rare or rapidly evolving event requires immediate observations. Igor Andreoni described a collaboration of more than 770 researchers working across microlensing, supernovae, fast transients, and variable objects. Sean McBride explained that external alerts can be assessed automatically within approximately a second and, when the trigger criteria are satisfied, lead to Rubin observations within minutes.

The system has already been exercised for gravitational-wave candidates, a high-energy-neutrino alert, and the interstellar object 3I/ATLAS. Shreya Anand presented proposed additions that included gamma-ray-burst triggers, follow-up of gravitational-wave regions covering as much as 1,500 square degrees, and automated responses to potentially hazardous asteroids identified through JPL Scout. For planetary defense, this could allow Rubin to redirect its enormous collecting power toward a newly discovered object while timely observations can still improve its orbit and physical characterisation.

Before an alert can be generated, however, Rubin must compare each new exposure with an existing reference image. Mario Jurić explained that suitable template coverage remains uneven because some regions and filters do not yet have enough high-quality observations. This currently limits where difference imaging—and therefore alerts and moving-object detections—can operate. The team is mapping where usable template inputs exist, validating new coadds, and evaluating whether commissioning images can expand the area available for prompt processing.

Jake Kurlander showed that the automated system has processed the large majority of incoming images successfully in real time, with many failures recovered during later daytime processing. A separate Prompt Products Database, targeted for public release in September or October 2026, will preserve the accumulated alert history and make it queryable through the Rubin Science Platform. It will include objects, difference-image detections, forced measurements, and Solar System tables, normally updated within 24 hours. For asteroid research, this creates the possibility of continuously extending lightcurves as new measurements arrive.

The alert stream itself is too large to inspect directly, so community brokers form the layer between Rubin and the scientists using its data. Christopher Hernandez demonstrated the Pitt-Google Alert Broker, which stores alert data in Google BigQuery and provides access to original alert packets, historical measurements, classifications, and cross-matched catalogues. Its tables are partitioned and spatially clustered so researchers can run targeted searches without scanning the entire database.

Anais Möller presented Fink, an open, community-driven broker whose science modules are developed by research groups around the world. Users can search its portal, create filters without writing SQL, request selected subsets of the archive, or subscribe to streams designed for particular science cases. ANTARES demonstrated how broker filtering can lead directly to additional observations: Tom Matheson described a test in which a likely supernova was identified automatically, passed through the AEON network, and sent to Gemini for follow-up within minutes. The brokers are therefore not simply archives. They classify, prioritise, and connect Rubin discoveries to the next scientific action.

Crowded regions such as the Galactic plane introduce another layer of difficulty. Ian Sullivan showed that Rubin currently has few alerts in the densest fields because templates have not yet been deployed across much of the plane. He also presented two major pipeline improvements. Adaptive calibration thresholds can now identify isolated calibration stars automatically, even where hundreds of thousands of sources occupy each square degree. Across a test involving approximately 287,000 images, this reduced calibration failures from nearly half of the images to about 1.6%.

A new difference-image deblender separates genuine variable or moving sources from neighbouring subtraction artefacts. In simulated tests in the crowded M49 field, it increased source recovery from approximately 96% to 98%. The numerical improvement appears small, but across Rubin’s full alert volume it represents a substantial number of recovered sources that would otherwise have been missed.

Natasha Abrams then presented ANTARES filters for detecting gravitational microlensing events, in which a foreground star, planet, or black hole temporarily magnifies a more distant source. Rubin’s cadence may not capture every short-lived feature, and several physical models can sometimes fit the same partial lightcurve. Early broker alerts are therefore essential for triggering faster photometric or astrometric observations.

Katarzyna Kruszynska introduced Ralph, a scalable modelling pipeline intended to analyse and rank the thousands of microlensing events expected from Rubin and Roman. By testing competing models and identifying anomalous events, Ralph could help determine which candidates require immediate follow-up before the most informative part of the event has passed.

A Connected Observatory

Rubin’s scientific value grows when its data are combined with observations from other facilities. Charlotte Ward described joint Rubin–Euclid analysis using scarlet2, which models astronomical sources directly across images from both surveys. Rubin provides extremely deep optical photometry, while Euclid supplies sharper space-based imaging and near-infrared coverage.

Euclid can distinguish multiple galaxies that appear blended into one source in ground-based Rubin images. Its higher-resolution morphology can then anchor the Rubin model, improving photometry, transient positions, host-galaxy measurements, and multi-survey lightcurves. The same framework can incorporate Roman or other complementary observations and is particularly useful for supernovae, tidal-disruption events, active galactic nuclei, and strongly lensed transients.

Raichoor Anand discussed the derived-data-product framework being developed between Rubin and Euclid. One survey can provide the positions and shapes of detected objects, while the other performs forced photometry at those locations. This allows both collaborations to benefit from the combined wavelength coverage while respecting their respective data-access structures.

He also presented the planned DESI-2 High-redshift Dark Energy Survey, which will use Rubin imaging to select millions of galaxies at redshifts of roughly 2.5 to 3.5 for spectroscopic measurements. Miranda Zak showed another type of synergy: combining ZTF’s existing time baseline with future Rubin and Roman observations to identify changing-look active galactic nuclei. Because these objects can evolve over months or years, linking several surveys provides a much longer and more reliable record than Rubin could initially create by itself.

Solar System Science

The Solar System Science session demonstrated how Rubin detections can be transformed into physical properties, dynamical histories, and follow-up targets.

David Trilling presented SNAPS, the Solar System Notification Alert Processing System. SNAPS is a downstream broker dedicated to known moving objects. It ingests alert photometry and derives properties such as rotation periods, amplitudes, and colours. Once an asteroid has enough observations, SNAPS can model its multiband lightcurve and separate rotational variability from true colour differences. Without this correction, measurements obtained at different rotational phases can blur the taxonomic groups that would otherwise appear in asteroid colour space.

SNAPS is also developing community filters for unusual changes in brightness, colour, and apparent source morphology. These filters can identify individual outliers or rare members of a broader population and push them toward rapid follow-up.

Erin Clark presented one such application: searching for active asteroids, objects on asteroid-like orbits that unexpectedly develop comet-like comae or tails. Her pipeline combines a comparison of the source’s point-spread function with a convolutional neural network trained on comet and inactive-asteroid images. The techniques are complementary. The PSF comparison is especially sensitive to broad comae, while the image classifier can recognise narrow tails.

The pipeline can process the roughly one thousand moving sources expected in a Rubin exposure in less than the 30-second interval between observations. Candidate active asteroids could therefore be published nightly for telescope follow-up. Expanding this population would help researchers investigate processes such as impacts, rotational disruption, and ice sublimation within the asteroid belt.

Dmitrii Vavilov presented RubinRocks, which pairs a rapid statistical API with direct numerical integrations. The API will estimate a near-Earth object’s likely source region, residence time, and dynamical history, helping observers select unusual discoveries from the many NEOs Rubin is expected to find. For objects requiring more precise reconstruction, the team developed Lobbie, a high-order collocation integrator using quadruple precision, adaptive timesteps, and special treatment of close planetary encounters. It can also incorporate relativity and non-gravitational forces. In tests integrating the planets and 50 asteroids for one million years, Lobbie conserved energy better than WHFast and IAS15 while producing the smallest run-to-run positional spread.

Carrie Holt presented FOMO—Follow-up Observations of Moving Objects—a platform for coordinating observations across robotic and manually operated telescopes. Researchers can share observing runs, see which objects are already scheduled, avoid unnecessary duplication, and track a target from its initial alert through later observations. FOMO is also being connected to systems such as JPL Scout so that potentially impacting objects can be prioritised quickly. Together, these projects showed how Rubin’s discoveries will be connected to physical interpretation and targeted observations beyond the survey itself.

Opening Rubin to More Researchers

Rubin’s data may be publicly accessible to a broad scientific community, but access to data alone does not guarantee access to discovery. Researchers also need software, computing resources, technical expertise, mentorship, and time.

Michael Wood-Vasey presented the LSST Discovery Alliance, which supports researchers through software incubators, fellowships, workshops, early-science activities, and consultations with professional software engineers. A research group might have a useful analysis script created for one project but lack the expertise to turn it into reliable, documented software that others can use. LSST-DA helps bridge that gap.

Its programmes also support institutions and researchers who may not have access to large computing centres or dedicated technical teams. A recurring message was that many barriers to Rubin science will come not from the observations themselves, but from the infrastructure and institutional support required to analyse them effectively.

The Rubin Undergraduate Network is addressing the particular challenges faced by undergraduate researchers: limited time before graduation, little prior research experience, competing coursework or employment, and mentors who may have limited time or institutional resources.

Its planned “RUN Projects” are self-contained Jupyter notebooks that guide students through a complete scientific question rather than assigning them one small piece of a larger project. Mike Solontoi demonstrated one of the most developed examples: retrieving asteroid observations from the Fink broker, converting fluxes to magnitudes, applying a multiband Lomb–Scargle search, and folding the lightcurve to recover an asteroid rotation period of approximately 2.2 hours. Other proposed projects include asteroid orbital distributions, variable-star periods, cluster membership, and detailed “biographies” of individual stars.

The session concluded with three undergraduate research presentations. Gina Johnson identified a previously unstudied stellar stream extending approximately 130 kiloparsecs around a nearby galaxy in DP1 imaging, probably produced by the disruption of a smaller dwarf galaxy. Sierra Blackhurst compared dwarf-galaxy morphology in overlapping Rubin and Euclid observations, finding that measurements such as size and light distribution generally agreed, while asymmetry and smoothness were more sensitive to image resolution and background subtraction.

Bilva Gummadi manually searched Rubin First Look imaging of the Virgo Cluster and catalogued 168 dwarf-galaxy candidates, including 156 potential new detections. Her work is an early step toward mapping their distribution and investigating how these dark-matter-dominated systems trace the structure of the cluster.

New Windows on the Universe

Sean McGraw discussed using Rubin to search vast gravitational-wave localisation regions for the optical counterparts of neutron-star mergers. A gravitational-wave signal provides a direct distance measurement. When its host galaxy can also be identified, the galaxy’s redshift can be combined with that distance to create a bright standard siren—an independent way of measuring the expansion rate of the Universe.

Even when no optical counterpart is found, the galaxies inside the gravitational-wave localisation volume can be analysed statistically as dark sirens. Individual measurements are broad, but combining many events can constrain the Hubble constant. Rubin’s combination of depth, speed, and wide-field coverage should strengthen both approaches and help investigate the persistent disagreement between different measurements of cosmic expansion.

The SEAMLESS project demonstrated how artificial intelligence can reveal extremely faint dwarf galaxies that fall between the traditional resolved-star and integrated-light search regimes. Its pipeline begins with diffuse-light detections, applies a convolutional neural network to reduce millions of candidates, and then uses human inspection to identify the most promising systems.

The project is particularly interested in isolated dwarf galaxies. Much of the current dwarf-galaxy census consists of satellites orbiting the Milky Way or other massive galaxies, where tides, ram pressure, and environmental effects alter their evolution. Isolated dwarfs offer a cleaner view of how low-mass galaxies form and evolve without those influences. The team plans to extend the search to DP2 through a project called Rubin Los Angeles.

Beyond the LSST: Rubin 2036

The final workshop session looked beyond Rubin’s current 10-year survey and asked what the observatory should do after the LSST ends in 2036. Planning has to begin long before then: there is not yet a funded post-LSST programme, major instrument or survey changes would require years of preparation, and the next astronomy prioritisation exercises will take place well before the current survey concludes. The discussion considered several possible directions, including continuing the existing cadence, adopting new footprints or observing strategies, expanding Target of Opportunity programmes, and introducing narrow- or medium-band filters for more specialised science.

One of the strongest cases involved the next generation of gravitational-wave observatories. Sean McGraw explained that facilities such as the Einstein Telescope, Cosmic Explorer, and LISA could detect mergers across a much larger fraction of the observable Universe, while Rubin’s wide field would remain especially well suited to locating their optical counterparts. Nandini Hazra showed that longer Rubin exposures and a few percent of the observatory’s time could potentially recover tens to hundreds of kilonovae per year, although the exact strategy would depend on event rates, localisation precision, and the practical limits of very long exposures.

A Telescope, a Data System, and a Community

The Rubin Community Workshop made clear that the observatory is much more than a telescope taking deep images of the sky. It is an interconnected scientific system: cameras and pipelines in Chile, databases and computing centres, public alert brokers, software frameworks, follow-up telescopes, international surveys, and researchers working across every stage of the process.

AT-LAST occupies one small part of that system. It begins with Rubin’s asteroid photometry and attempts to turn sparse observations into reliable rotation periods at a scale that manual analysis cannot reach. The other projects presented during the week addressed neighbouring parts of the same chain: identifying unusual detections, correcting crowded-field measurements, estimating asteroid origins, coordinating follow-up observations, and ensuring that students and researchers at many kinds of institutions can participate.

Rubin’s first datasets already contain discoveries, but the workshop showed that the observatory’s greatest impact will come from what the community builds around them. The LSST has only just begun, and the tools, collaborations, and scientific questions developed now will shape how we explore its view of the changing sky over the next decade.