{"id":"9720a6f7-fe95-4c49-b5ec-532618aacc88","arxiv_id":"2608.00046","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A new high-proper-motion object, SMDET-1, has Spitzer color limits implying a Y dwarf within ~7.4 pc of the Sun (T_eff < 391 K).","lead":"A deep-learning program scanning ten years of WISE satellite images found a fast-moving, very red infrared object, SMDET-1, in a crowded part of the Milky Way. Its extreme Spitzer color suggests it is a cold Y-dwarf brown dwarf, likely within 7.4 parsecs of the Sun, awaiting confirmation by spectrum or parallax.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Headline numbers hinge on photometric calibrations at the red edge; the ch1-depth choice is not fatal because both interpretations still yield a Y-type color.","rationale":"The reader's weakest assumption (ch1 depth) is not the single most load-bearing issue. Even if the 3.4σ forced-photometry flux is real, ch1−ch2 becomes 3.09 mag, which is still redder than the adopted limit 2.81, and the resulting T_eff≈349 K, d≈5.6 pc remain inside the abstract's inequalities. The Y-dwarf phototype and the <7.4 pc distance are therefore robust to the ch1 interpretation. What is NOT robust is the mapping from a color limit of 2.81 to T_eff<391 K and d<7.4 pc. This mapping uses empirical polynomials calibrated on a small sample and applied at the extreme red edge, where there are few anchor objects. The paper's own caveat about metal-poor outliers makes this a genuine risk. A metal-poor brown dwarf can have unusually red ch1−ch2 at a given T_eff, so SMDET-1 could be warmer and more distant than the headline values while still explaining all photometry. A leave-one-out robustness test on the calibration, or a NIRSpec spectrum, would settle the issue. Because the paper labels SMDET-1 a candidate and transparently states these caveats, the CONDITIONAL verdict is appropriate; my concern does not move the verdict.","tokens_in":17002,"tokens_out":16291,"duration_ms":174782,"concrete_test":"Perform a leave-one-out re-fit of the §4.1 T_eff(ch1−ch2) and M_ch2(ch1−ch2) second-order polynomials using the JWST-based sample (Beiler et al. 2024; Rowland et al. 2024), recomputing the inferred limits at ch1−ch2=2.81 mag each time. If excluding any single object (especially WISE 0855−0714, the coldest anchor) shifts the T_eff limit by >50 K or the M_ch2 limit by >0.5 mag, the headline values are not robust and should be reported with wider uncertainty or as pure phototypes. If the re-fits are stable, the calibration concern is largely resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The discovery layer is solid: WISE + Spitzer astrometry over 12.4 yr with a resolvable 2012 Spitzer ch2 detection rules out artifacts, and an unphysical tangential velocity rules out a background source. The load-bearing step is not the choice between the §3.1 ch1 limit (17.38) and the §3.2 forced-photometry residual (17.66): either way ch1−ch2 ≳ 2.8, a Y-type phototype, and the quoted upper limits T_eff<391 K and d<7.4 pc remain valid upper limits. What would threaten the central claim is the conversion of that color into physical properties. The T_eff(ch1−ch2) and M_ch2(ch1−ch2) polynomials used in §4.1 are calibrated on known brown dwarfs and applied at the red edge; the paper itself notes (end of §4.1) that metal-poor objects can be very large outliers in the temperature–color plane (e.g., Faherty et al. 2025). If SMDET-1 is a metal-poor or young/dusty object, the same red Spitzer color could correspond to T_eff>391 K and a larger photometric distance. The high proper motion alone only bounds the distance to within tens of parsecs for any reasonable disk tangential velocity, not to <7.4 pc. Thus the qualitative discovery is robust, but the two headline physical numbers are conditional on the calibrations and on SMDET-1 not being an outlier.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the discovery of SMDET-1, a high proper motion (~1.3\"/yr) object identified through the SMDET pixel-level deep learning search on time-resolved unWISE coadds. The object is independently confirmed by a 2012 Spitzer/Deep GLIMPSE detection at 4.5 μm and by additional WISE detections spanning ~12.4 years. It is undetected in Spitzer ch1 and in near-infrared JHK imaging, yielding a Spitzer color limit ch1−ch2 > 2.81 mag. From this color limit the authors derive a photometric temperature limit T_eff < 391 K, a photometric distance limit d < 7.4 pc, and a Y dwarf phototype. The paper also presents early SPHEREx observations that are inconclusive. The central claim is that SMDET-1 is a newly discovered, very nearby Y dwarf candidate.","tokens_in":17220,"tokens_out":5031,"duration_ms":53290,"significance":"If the characterization holds, SMDET-1 is a valuable addition to the incomplete census of the coldest brown dwarfs within ~20 pc and demonstrates the utility of deep learning for discovering faint moving objects in crowded fields. The discovery layer is strong: the Spitzer ch2 detection at ~2'' resolution corroborates the WISE detections, the motion is consistent across multiple epochs, and the paper carefully distinguishes the 5σ ch1 limit from the sub-threshold forced-photometry residual. The authors also cross-check their temperature and color relations against multiple published polynomials, and they explicitly flag the possibility of metal-poor outliers. The main risk is that the physical limits (T_eff < 391 K, d < 7.4 pc) are derived from photometric calibrations at the red edge of the color–temperature and color–absolute magnitude relations, so the quantitative headline numbers are conditional on SMDET-1 being a normal solar-metallicity field object.","major_comments":[{"comment":"The T_eff(ch1−ch2) polynomial is fitted to JWST-based effective temperatures of a small sample (the black points in Fig. 3) and then evaluated at ch1−ch2 = 2.81, at or beyond the reddest calibration points. The paper notes that metal-poor objects can be very large outliers (Faherty et al. 2025) but does not quantify how this affects the T_eff < 391 K claim. Since this limit is a headline result, the authors should either (a) quantify the systematic scatter/extrapolation uncertainty using all published objects with ch1−ch2 > 2.5, including known outliers, or (b) explicitly present T_eff < 391 K as a solar-metallicity, field-age photometric estimate and modify the abstract accordingly. As written, the abstract implies a robust physical upper limit.","section":"§4.1, Fig. 3"},{"comment":"The paper adopts the 5σ ch1 limit (ch1 > 17.38) over the forced-photometry residual (ch1 = 17.66 ± 0.38/0.28, 3.4σ). The three arguments given are reasonable, but the residual is not negligible. If real, it would change the color to ~3.09 mag, the photometric distance to ~5.6 pc, and T_eff to ~349 K. While §4.2 discusses this alternative, it is absent from the abstract and conclusion. I recommend that the abstract/conclusion state that the primary limits are based on the adopted 5σ non-detection and explicitly note the alternate sub-threshold interpretation, or that Table 1 include the forced-photometry values as a separate row.","section":"§3.2, §4.2"},{"comment":"The proper motion fit has a poor reduced chi-squared for μα (χ²_ν = 13.9), and the paper acknowledges that the formal uncertainties may be underestimated due to WISE blending. This matters because the predicted positions anchor the JHK non-detection limits (§3.4, §3.5) and the v_tan < 44.9 km/s limit (§4.1). The authors should add a systematic error floor to the WISE astrometric points (e.g., based on the fit scatter) and demonstrate that the NIR non-detections and v_tan limit are robust to the resulting enlarged positional uncertainty. Without this, the reader cannot assess how conservative the derived limits are.","section":"§3.3, Table 1"}],"minor_comments":[{"comment":"Use 'December 2012' instead of 'late-2012' or '2012 December' for consistency.","section":"§2.3"},{"comment":"The photometry header 'JM KO', 'HM KO', 'KM KO' appears to be a formatting artifact; it should read 'J (MKO)', 'H (MKO)', 'K (MKO)'.","section":"Table 1"},{"comment":"The reduced proper motion H_ch2 ≈ 20.1 mag is mentioned without a definition or equation; define it or cite the standard formula so the value is reproducible.","section":"§4.1"},{"comment":"The fitted T_eff(ch1−ch2) polynomial is shown as a dotted line, but the coefficients and the number of calibration points are not given. Please provide the polynomial coefficients and the calibration sample size, ideally in a footnote or in the text.","section":"Figure 3"},{"comment":"The term 'L VF' should be 'LVF' (linear variable filter) consistently throughout the section.","section":"§3.6"},{"comment":"Several references are listed only as arXiv e-prints (e.g., Meisner et al. 2022, Caselden et al. 2026, Leggett 2026). If any have been accepted or published, please update to the journal reference. Also, the Meisner et al. 2022 entry lacks a DOI.","section":"References"},{"comment":"The notation 'ch1' and 'ch2' is used throughout; ensure the first use explicitly defines the Spitzer/IRAC channel wavelengths (this appears in footnote 14, but might be worth stating in the main text as well, especially since the abstract uses [3.6] and [4.5]).","section":"§3.1"}],"recommendation":"major_revision","confidential_remarks":"The discovery of SMDET-1 is believable and well-supported by the Spitzer and WISE data. The main concern is that the two headline physical quantities (T_eff < 391 K and d < 7.4 pc) are photometric limits that depend on the calibration of the red edge of the ch1−ch2 color relations and on the treatment of the 3.4σ ch1 forced-photometry residual. The authors already discuss these caveats, but they are not reflected in the abstract. I think a major revision is appropriate, asking the authors to either quantify the calibration systematics or present the limits with explicit assumptions. The paper is otherwise well within the scope of the journal and would be of interest to the brown dwarf community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: SMDET-1 is real and new. The discovery layer is solid. WISE time-resolved coadds show a 1.3\"/yr moving source, and Spitzer/Deep GLIMPSE resolves it at 4.5 μm at a bracketing epoch, which rules out the usual single-epoch artifacts. The paper is honest that the object is a candidate, and it goes out of its way to quantify every limit it uses. That transparency is the main reason I trust the discovery.\n\nWhat's genuinely new is the object itself. It was sitting in unTimely catalog detections and Deep GLIMPSE, but nobody flagged it as moving because it's in a crowded plane field and blended with brighter contaminants. The SMDET deep-learning search is the actual tool, and this paper is a demonstration of it finding something previous methods missed. The methodology is described elsewhere, but the application here is legitimate.\n\nThe soft spots are all in the characterization layer. The ch1 non-detection is converted to a limit using a locally fitted 5σ depth of 17.38, which is 0.42 mag shallower than the survey's quoted depth. The forced photometry leaves a 3.4σ residual that would shift the numbers — but even adopting the residual gives ch1−ch2 ≈ 3.09, which is still a Y-phototype color. So the ch1-depth choice doesn't threaten the core classification. The μ_α fit is poor (χ²_ν = 13.9), and the paper admits the uncertainties are underestimated. There is no parallax, no spectrum, as the paper says. The main real caveat is the one the paper states in §4.1: metal-poor objects can be large outliers in the T_eff–color plane. If SMDET-1 is metal-poor or dusty, the same color wouldn't guarantee T_eff < 391 K or d < 7.4 pc. High proper motion alone only bounds the distance to tens of parsecs. So the headline numbers are conditional, and the paper's own caveats are appropriate.\n\nBottom line: this deserves a serious referee. The discovery is strong, the characterization is appropriately modest, and the paper is a useful contribution to the nearby brown dwarf census. A referee should push for a parallax or at least a clearer statement of the systematic uncertainty in the photometric distance, but the central claim — a fast-moving, very red object with a Y-dwarf phototype — holds up. Send it to review.","headline":"A genuinely new fast-moving cold brown dwarf candidate with a solid two-instrument discovery; the headline distance and temperature are photometric upper limits that survive the main modeling choice but not the metal-poor-outlier caveat.","tokens_in":17989,"tokens_out":1712,"would_cite":true,"duration_ms":31866,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"SMDET-1 is a fast-moving infrared source that appears to be a Y dwarf within 7.4 pc of the Sun.","keywords":["brown dwarfs","Y dwarfs","proper motions","WISE","Spitzer","deep learning","solar neighborhood","photometric distance"],"falsifier":"A detection of SMDET-1 at 3.6 μm with a point-source flux exceeding the adopted limit, or a J-band detection at J ≤ 21.16, would overturn the color lower limit that drives the Y dwarf classification. Conversely, if forced photometry at the predicted position yields a clean 5σ ch1 detection, the current distance and temperature limits would be replaced by values near 5.6 pc and ~349 K.","tokens_in":16703,"feed_emoji":"🪐","tokens_out":3927,"duration_ms":38819,"temperature":0.7,"pith_summary":"This paper reports the discovery of SMDET-1, a faint, fast-moving infrared source that appears to be one of the coldest and closest known brown dwarfs: a Y dwarf with an effective temperature below 391 K and a photometric distance under 7.4 pc. The object was found by a deep-learning search over time-resolved WISE images, then confirmed and measured with Spitzer imaging that detects it only at 4.5 μm. The Y dwarf classification rests on a color lower limit: the object is redder in Spitzer's ch1–ch2 color than any T dwarf, with ch1–ch2 > 2.81 mag. If the paper is right, SMDET-1 enlarges the incomplete census of the very coldest objects in the solar neighborhood and demonstrates that pixel-level machine learning can uncover moving objects in crowded fields that traditional catalog searches missed.","feed_headline":"A likely Y dwarf is hiding 7.4 parsecs away","feed_subtitle":"A neural network sifting archival sky survey images found one of the coldest, closest neighbors to the Sun.","key_machinery":"The argument turns on the conversion of a non-detection into a constraint: the Deep GLIMPSE catalog detects SMDET-1 only in the 4.5 μm (ch2) channel, and the authors translate the absence of a 3.6 μm (ch1) detection into a 5σ magnitude limit by fitting a second-order polynomial to the ch1 signal-to-noise versus magnitude trend of ~1,600 nearby catalog sources. That ch1 limit, combined with the measured ch2 flux, produces the color lower limit that anchors the temperature, distance, and phototype estimates. The discovery itself rests on the SMDET neural network, a recurrent convolutional architecture that scans sequences of time-resolved WISE coadds for faint, fast-moving sources.","core_discovery":"The paper's central claim is that SMDET-1 is a genuine, high–proper-motion (≈1.3 arcsec/yr) brown dwarf, detected at 4.5 μm by both WISE and Spitzer and absent from near-infrared JHK imaging, whose Spitzer color limit places it in the Y dwarf regime. Because Spitzer's 3.6 μm channel fails to detect it, the authors derive a 5σ limit of ch1 > 17.38 mag from a local fit to sources around the object, giving ch1 – ch2 > 2.81 mag. Applying published color–luminosity and color–temperature relations, this yields T_eff < 391 K, a photometric distance < 7.4 pc, and a phototype later than Y0.8. The object is therefore presented as a nearby Y dwarf candidate and as evidence that the coldest population o","pith_inferences":["The paper's property estimates inherit a systematic risk: the adopted ch1 depth is 0.42 mag shallower than the survey's quoted sensitivity, and a 3.4σ residual flux sits at the predicted position. A deeper 3.6 μm observation would either confirm a real detection (shrinking the distance to ~5.6 pc) or push the limiting color even redder.","SMDET-1's nondetection in early SPHEREx data is presented as consistent with expectations, but a dedicated stacking analysis as more SPHEREx passes accumulate could provide an independent 4–5 μm detection without waiting for a new telescope.","This discovery suggests the low-mass cutoff of star formation may be even closer than the current census implies, and that similarly cold objects are likely hidden in the same archival data where traditional point-source catalogs fail due to crowding and blending."],"forward_implications":["SMDET-1, if confirmed, becomes one of the closest known brown dwarfs, with a photometric distance under 7.4 pc, joining the small set of Y dwarfs in the solar neighborhood.","The object's extreme ch1–ch2 color and J-band non-detection make it a strong target for JWST or large-telescope spectroscopy to measure its temperature and composition.","The successful recovery of a moving source in a crowded Galactic-plane field suggests that similar pixel-level deep-learning searches can find other overlooked nearby objects in archival WISE and Spitzer data.","A trigonometric parallax from high-resolution follow-up would convert the photometric distance into a precise measurement and test the Y dwarf interpretation.","If the sub-threshold ch1 flux is real rather than contamination, the object would be even closer (≈5.6 pc) and cooler (≈349 K), strengthening its status."],"fun_headline_variants":["AI finds a cold neighbor: Y dwarf just 7.4 parsecs away","Deep learning spots hidden Y dwarf candidate near Sun","Y dwarf found 7.4 pc away using AI on old survey images","AI uncovers a cold Y dwarf candidate just 7.4 pc away","A deep-learning search reveals a nearby Y dwarf"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The core premise is that SMDET-1 has no real 3.6 μm emission, so the measured 3.4σ residual at its predicted position is treated as noise or imperfect subtraction of a bright neighbor rather than a faint detection; if that residual is real, the derived color, distance, and temperature all shift, though the Y dwarf classification would likely survive.","fun_headline_variants_meta":{"raw":{"variants":["AI finds a cold neighbor: Y dwarf just 7.4 parsecs away","Deep learning spots hidden Y dwarf candidate near Sun","Y dwarf found 7.4 pc away using AI on old survey images","AI uncovers a cold Y dwarf candidate just 7.4 pc away","A deep-learning search reveals a nearby Y dwarf"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001008,"raw_usage":{"total_tokens":4154,"prompt_tokens":859,"completion_tokens":3295,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":3205}},"tokens_in":603,"tokens_out":3295,"duration_ms":58396,"temperature":1.0,"reasoning_tokens":3205,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T01:31:23.897181+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A detection of SMDET-1 at 3.6 μm with a point-source flux exceeding the adopted limit, or a J-band detection at J ≤ 21.16, would overturn the color lower limit that drives the Y dwarf classification. Conversely, if forced photometry at the predicted position yields a clean 5σ ch1 detection, the current distance and temperature limits would be replaced by values near 5.6 pc and ~349 K.","supporting_citations":[],"review_version":1}