{"id":"fdf8b7a1-8ad2-46f1-abd2-db3d80ebea41","arxiv_id":"2607.23651","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":4,"one_line_summary":"BLUSE autonomously upchannelises MeerKAT F-engine data to ~1 Hz, forms 64 coherent beams per pointing, and has searched >1.2 million pointings for narrowband technosignatures since 2022.","lead":"Breakthrough Listen built BLUSE, an automated system that rides along on MeerKAT observations and has already searched more than 1.2 million sky pointings for alien technosignatures. It shows how array telescopes can multiply SETI search speed at low extra cost by sharing the same data stream.","discovery_kind":"new_method","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The end-to-end validation rests on a single S0/4k observation; borrowed TelState calibration is assumed, not shown, to hold across the other bands, F-engine modes, and subarray configurations that make up most of the 1.2M pointings.","rationale":"I agree with the reader's identification of the weakest assumption and with the ACCEPT verdict. The paper's central claim is an engineering one — that multicast F-engine access plus real-time upchannelisation, beamforming, and Taylor-tree search can autonomously process survey-scale data volumes — and that claim is directly supported by the described measurements, the end-to-end JWST detection (including the stamp file in Fig. 11), unusually open code (seticore, hpguppi_daq, commensal-automator, bfr5 generator, targets-minimal), and three years of operational statistics. The single-band/single-mode efficiency validation is the one place where a number measured once is being implicitly generalized across the full operating envelope, and that generalization underwrites the effective sensitivity of the archived 1.2M pointings rather than the throughput claim itself. Because science results and hit distributions are explicitly deferred, the correct remedy is the proposed multi-band/multi-mode efficiency characterization, reported alongside the forthcoming science papers — not a downgrade of this instrumentation paper. Minor secondary observations, none load-bearing: the abstract's \"1.2 million individual pointings\" actually corresponds to ~360k unique objects with repeats (§6 is transparent about this), and hit-level sensitivity vs. drift rate is not characterized, though the ±10 Hz/s range and SNR-6 threshold are clearly stated as tunable operational choices.","tokens_in":14788,"tokens_out":2040,"duration_ms":88848,"concrete_test":"Repeat the Eq. 2 coherent/incoherent efficiency measurement across the operational envelope: record raw-voltage pointings (as done for JWST in §5) in UHF and L bands and in 1k and 32k modes, ideally on bright calibrators or further JWST-type transits, with varying subarray sizes. For each, compute η, and additionally compute η in sliding ~30 s windows across the full 290 s recording to measure decorrelation from solution staleness, noting the TelState solution timestamp vs. recording time. If η < ~0.7 in any band/mode, or drops materially within a single recording, the sensitivity of that share of the 1.2M pointings must be revised in the forthcoming survey papers; if η stays ≥ ~0.8 throughout, the calibration assumption is confirmed and the concern closes.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader correctly identified the soft spot. Section 2.1 states BLUSE \"can retrieve and rely upon calibration solutions produced for the primary observer\" — i.e., phase/delay solutions derived for someone else's calibrator schedule, at some prior time, possibly on a different antenna subset. The only quantitative check that these solutions actually yield good coherent beams is the JWST test in Section 5: one S0-band, 4k-mode, 62-antenna observation giving η = 0.849 via Eq. 2 (0.92 by the Rajwade et al. 2022 method). Nothing in the paper quantifies η for UHF or L band, for 1k/32k modes, for small or split subarrays, or — critically — as a function of solution age. Phase solutions drift on minute timescales; the bfr5 generator applies retrieved solutions plus geometric delays at 1 s cadence over 290 s, but any residual antenna phase error present at solution-retrieval time (or accumulated since the primary observer's last calibration scan) directly decorrelates the coherent sum. Since SEFD scales inversely with beamforming efficiency, a silently degraded η in some configurations means an unknown fraction of the 1.2M pointings were searched at lower sensitivity than the system is capable of. I stress this does not defeat the paper's central engineering claim: 85–92% efficiency at S-band is already a credible demonstration that the approach works, throughput statistics are operational facts, and sensitivity characterization is explicitly deferred to forthcoming science papers. It is a real gap, but one that tempers the future scientific yield of the survey rather than the architecture demonstration made here. Hence it warrants a flagged caveat, not a verdict change.","agreement_with_reader":"agree"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The manuscript describes BLUSE, Breakthrough Listen's user-supplied equipment at MeerKAT: an autonomous commensal system that subscribes to F-engine multicast streams, upchannelises to ~1 Hz resolution, synthesizes 64 coherent beams plus one incoherent beam per 290 s primary pointing, and runs a Taylor-tree de-Doppler search with stamp-file capture of per-antenna voltages around detections. The paper documents the processing pipeline (§2), the state-machine-based automation that handles heterogeneous subarray configurations (§3), the hardware (§4), an end-to-end validation using JWST's S-band telemetry downlink as a test source (§5), and three years of observing statistics totalling ~1.2 million viable pointings on ~360,000 unique objects (§6). The central claim is that multicast F-engine access plus real-time upchannelisation, beamforming, and drift search constitutes a working, cost-effective path to large-scale commensal technosignature surveys.","tokens_in":15209,"tokens_out":1985,"duration_ms":90935,"significance":"If the system performs as described, this is a substantial contribution to the field: ~29,000 new stars per month at ~1 Hz resolution dwarfs the throughput of prior targeted single-dish surveys, and the commensal model demonstrated here is directly relevant to survey design at MeerKAT, the VLA, and SKA-pathfinder arrays. The paper ships several concrete strengths that deserve explicit credit: (i) a genuine end-to-end validation against an independently known moving source (JWST ephemerides), recovering the expected transit in tiled offline beams, correlator images, and per-antenna stamp files; (ii) a quantitative beamforming efficiency (η = 0.849 via Eq. 2, ≈0.92 by the Rajwade et al. 2022 method) consistent with independent L-band literature; (iii) falsifiable operational statistics (pointing counts, unique-object counts, sky maps) rather than fitted parameters; and (iv) public code for essentially every pipeline component (seticore, hpguppi_daq, commensal-automator, bfr5 generator, targets-minimal). This is a system paper done the right way.","major_comments":[{"comment":"The one quantitative check that borrowed TelState calibration solutions yield good coherent beams is a single S0/4k, 62-antenna observation (η = 0.849). Yet §2.1 states BLUSE 'can retrieve and rely upon calibration solutions produced for the primary observer', and the 1.2M pointings of §6 span UHF/L/S bands, 1k/4k/32k modes, and varied subarray sizes. Phase solutions drift on minute timescales and were derived for the primary observer's calibrator schedule, possibly on a different antenna subset; residual phase error at retrieval time directly decorrelates the coherent sum, and SEFD scales inversely with η. This does not defeat the central engineering claim — 85–92% at S-band is a credible demonstration — but the manuscript should (a) state explicitly that η has been verified only for the one configuration tested, and (b) ideally provide a concrete check of solution robustness, e.g. η as","section":"§2.1 and §5"},{"comment":"The headline throughput numbers (1.5M beams processed, 1.2M viable, 360k unique objects, 29k/month) are given without a breakdown by band, F-engine mode, or subarray size. Since the reader is told in §3 that BLUSE operates across all configurations, a table or figure decomposing the pointing counts by band and mode would make the observing-progress claim auditable and would also support the calibration-uniformity question above. This is a modest addition from data the authors necessarily already have.","section":"§6"}],"minor_comments":[{"comment":"The target scoring S'_band = S_band + t×b×n adds quantities with mixed dimensions (seconds × subband segments × antennas). Presumably this is an ad hoc heuristic score, which is fine, but a sentence stating that the score is dimensionless by construction (and what b counts exactly for the 5-subband S-band case) would help.","section":"§3.5, Eq. (1)"},{"comment":"The efficiency from Eq. (2) is reported as 84.9% from a single measurement; no uncertainty is quoted. Even a rough error budget (thermal noise on P_coh/P_incoh, antenna-flagging fraction) would strengthen the comparison with Rajwade et al. (2022).","section":"§5"},{"comment":"The SNR threshold of 6 and drift range ±10 Hz/s are stated as operational choices; a brief note on the resulting expected hit rate per pointing (and how the stamp-file 'dial' of §2.4 interacts with it) would help readers assess data-product volumes.","section":"§2.3"},{"comment":"The fine-channel bandwidths (1.01/1.59/1.62 Hz) differ subtly from the '~1 Hz' used in the abstract and §2.2; worth stating once that the exact value depends on receiver/mode as tabulated.","section":"§2, Table 1"},{"comment":"The caption notes markers are not representative of beam size/shape; adding the synthesized beam FWHM at S0 for scale would make the tiled-beam demonstration more quantitative.","section":"Figure 7"},{"comment":"'Shared on request' is weak for a paper whose validation rests on one observation; depositing the JWST-test raw voltage recording and stamp files (modest volume) in a public archive would make the key validation fully reproducible.","section":"Data Availability"},{"comment":"Minor typographical issues: 'hpguppi daq' (§2) appears with a stray space; the ESDKB URL in footnote 8 is broken across lines; Figure 2's caption could spell out the FreeSubscribed/RecProc abbreviations on first use.","section":"Throughout"}],"recommendation":"minor_revision","confidential_remarks":"This is a solid system-description paper from the group that built the instrument; self-citations are appropriate to prior BL papers and software. The science yield is explicitly deferred to forthcoming papers, so the editor should treat this as an instrumentation/validation contribution — on those terms it meets the bar. The only real soft spot is single-configuration validation of borrowed calibration, which the authors can address with a caveat plus at most a small amount of additional analysis."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"This is the engineering write-up of BLUSE, the Breakthrough Listen backend that has been running commensally on MeerKAT since 2022. The headline numbers are real: full-band F-engine ingest, ~1 Hz upchannelisation, 64 coherent beams plus incoherent, Taylor-tree search, and >1.2 million pointings already processed. That scale is the new result. Commensal beamforming SETI already exists (COSMIC, ATA, MWA), but the open architecture, stamp-file intermediate product, and multi-year operating statistics on a southern-sky array are concrete and useful.\n\nWhat they did well is straightforward. The pipeline description is clear, the hardware and automation tables are complete, and the JWST S-band telemetry test is a proper end-to-end check: tiled beams track the moving source, stamp files show the signal on every antenna, and they measure 84.9 % coherent/incoherent efficiency (0.92 by the Rajwade method). Code links are public. Sky coverage and throughput growth year-over-year are documented without hype. Free parameters (SNR cut, drift range, beam count, ranking scores) are stated as operational knobs, not hidden.\n\nThe soft spot the stress-test flags is real and correctly sized. Calibration is borrowed from TelState solutions made for the primary observer; the only quantitative efficiency number is one S0/4k, 62-antenna JWST track. Nothing is shown for UHF/L, other F-engine modes, small or split subarrays, or solution age. Residual phase error would silently lower sensitivity on an unknown fraction of the 1.2 M pointings. That tempers the eventual science yield, not the architecture claim made here. Science results are deferred, which is appropriate for this paper.\n\nThis is for instrument builders, array operators, and anyone planning SKA-era commensal surveys. Citation pattern is normal for a systems paper. Math is elementary and correct; data rates and efficiency arithmetic check out. I would send it to referees without hesitation and would cite the architecture and throughput numbers myself.","headline":"Solid systems paper: a working full-bandwidth MeerKAT commensal SETI backend with 1.2M pointings already in the bag; the single-band JWST efficiency check is a real but proportionate caveat.","tokens_in":16176,"tokens_out":531,"would_cite":true,"duration_ms":11014,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"An automated MeerKAT backend has already searched more than 1.2 million coherent beams for technosignatures by riding along with ordinary observations.","keywords":["technosignatures","SETI","MeerKAT","commensal observing","beamforming","radio astronomy instrumentation","narrowband search"],"falsifier":"Measure coherent-beam efficiency and residual phase error on a grid of real survey pointings spanning UHF, L-band and all S-band sub-bands; if efficiency routinely falls well below the ~85 % JWST test value, the survey’s claimed sensitivity and localisation power do not hold.","tokens_in":15987,"feed_emoji":"📡","tokens_out":870,"duration_ms":17533,"temperature":0.7,"pith_summary":"This paper describes BLUSE, a fully automated system that taps MeerKAT’s multicast F-engine data stream and runs a continuous technosignature search without needing dedicated telescope time. It upchannelises the full available bandwidth to roughly 1 Hz, forms dozens of coherent beams on stars and other objects inside each primary field of view, and runs a Taylor-tree de-Doppler search on every beam. End-to-end tests with the James Webb Space Telescope’s S-band telemetry confirm that the beamformer and search pipeline work as designed. Since mid-2022 the system has processed more than 1.2 million individual pointings, showing that array-based commensal observing can multiply the volume of SETI data by orders of magnitude at modest marginal cost.","feed_headline":"1.2 million beams searched for alien signals on the side","feed_subtitle":"MeerKAT’s multicast stream lets a SETI backend ride every observation at full bandwidth","key_machinery":"The BLUSE pipeline: multicast subscription to MeerKAT F-engine packets, real-time upchannelisation to ~1 Hz, coherent beamforming from primary-observer calibration solutions, and an in-line Taylor-tree de-Doppler search that writes hits and voltage “stamps.”","core_discovery":"BLUSE demonstrates that a user-supplied backend can autonomously ingest MeerKAT’s full-bandwidth F-engine multicast streams, form 64 coherent beams plus one incoherent beam per primary pointing, and search them for narrowband drifting signals at ~1 Hz resolution, accumulating more than 1.2 million processed pointings since 2022 and thereby establishing commensal array surveys as a rapid, cost-effective path to large-scale technosignature coverage.","pith_inferences":["If other multicast-capable arrays adopt the same pattern, the global SETI search rate could become limited by compute and storage rather than by allocated telescope hours.","The dependence on primary-observer calibrations suggests a natural next step: an independent real-time self-cal loop that would keep beamforming robust when the primary schedule is sparse or poorly calibrated.","Stamp-based archival strategy implies that the long-term science return will hinge as much on how aggressively stamps are kept as on how many beams are formed."],"forward_implications":["Technosignature search volume can grow by tens of thousands of new stars per month without competing for primary telescope time.","Voltage “stamp” files around each hit allow later re-beamforming anywhere in the primary field of view, turning detections into reusable data products.","The same multicast architecture can host additional search algorithms (imaging, pulsed searches, machine-learning detectors) in parallel with the existing Taylor-tree pipeline.","Sky maps of processed pointings already reveal dense coverage of the Galactic plane, Virgo Cluster and deep fields, ready for statistical analyses of hit rates versus frequency and sky position."],"fun_headline_variants":["1.2M MeerKAT pointings scanned commensally for technosignatures","BLUSE rides MeerKAT multicast to search 1.2M beams at 1 Hz","Autonomous SETI backend processes full-band MeerKAT F-engine data","Commensal BLUSE forms 64 beams per pointing across 1.2M targets","MeerKAT commensal survey hits 1.2M technosignature pointings since 2022"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The calibration solutions produced for the primary observer stay accurate enough for coherent beamforming across every band, subarray, and 290-second integration the survey actually uses.","fun_headline_variants_meta":{"raw":{"variants":["1.2M MeerKAT pointings scanned commensally for technosignatures","BLUSE rides MeerKAT multicast to search 1.2M beams at 1 Hz","Autonomous SETI backend processes full-band MeerKAT F-engine data","Commensal BLUSE forms 64 beams per pointing across 1.2M targets","MeerKAT commensal survey hits 1.2M technosignature pointings since 2022"]},"model":"grok-4.5","effort":"low","cost_usd":0.004013,"raw_usage":{"total_tokens":1328,"prompt_tokens":880,"num_sources_used":0,"completion_tokens":101,"cost_in_usd_ticks":40128000,"prompt_tokens_details":{"text_tokens":880,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":347,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":880,"tokens_out":101,"duration_ms":7031,"temperature":1.0,"reasoning_tokens":347,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-30T16:38:08.823780+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Measure coherent-beam efficiency and residual phase error on a grid of real survey pointings spanning UHF, L-band and all S-band sub-bands; if efficiency routinely falls well below the ~85 % JWST test value, the survey’s claimed sensitivity and localisation power do not hold.","supporting_citations":[],"review_version":1}