# Cooking Thermometers & Timers: research notes Date: 2026-10-25. Scope: thermometers used to check food or cooking-surface temperature during home cooking — instant-read digital and dial thermometers, leave-in oven-safe probes, wireless/"smart" probe thermometers (wired-probe and truly-wireless), infrared surface thermometers, candy/deep-fry thermometers, and the standalone kitchen timers the taxonomy leaf bundles in with them. One register for the whole leaf; the `device_type` and `intended_use` splits do the dividing the label's "&" implies, rather than separate registers. Appliance thermometers (fridge/freezer/oven-air dial gauges) are mentioned only where they clarify a boundary; they are a different buying decision (checking an appliance, not food) and are not a facet here. No sibling leaf in `supplies > kitchen` overlaps materially: `cookware-stove-accessories` and `pressure-cookers` do not carry built-in probes in the add-on schema read for this run, so nothing is swallowed from a sibling. Quote labels: **verbatim** = the sentence in quotation marks was read directly in text I fetched myself (curl + tag-strip, or pdftotext on a downloaded PDF) and I saw it in that text; **fetched excerpt** = the fetch tool summarised or lightly reworded the page for me; **search excerpt** = from a search-engine summary only, page unseen. WebSearch reported its session budget exhausted before any query for this run returned results, so every source below is a direct fetch (curl, mostly through the skill's own `get.sh` helper), a WebFetch call, or an Internet Archive (Wayback Machine) snapshot where the live page blocked the fetch (403). No search-excerpt-only numbers appear in category.json. ## Seed Seed run output (recorded verbatim from the brief): - Example: kitchen-knives `steel_hardness` — "a numeric spec that only makes sense as one side of a trade-off, and a visual grammar borrowed from the hobbyist community that puts the trade-off on one axis." - Persona: "A print production manager, who thinks in what the spec sheet says versus what comes off the line." 1. **The thermometers equivalent of steel hardness is response time versus stated accuracy.** Both are printed numbers (seconds, ±°F) that only make sense as two ends of one trade: ThermoWorks' own explainer says thinner, faster sensors (thermocouples) are also the most shock-resistant, while FSIS's primer gives the slow end explicitly — 10 s for a thermistor, 15–20 s for an instant-read dial, 1–2 min for an oven-safe dial that averages over 2–2.5 inches of stem. **Kept** as the `response_time_s` split and the `stated_accuracy_f` annotation, with a proposed "speed vs. accuracy" two-axis map as the visual grammar (no single trade body publishes this chart for this category, so it is logged as a proposal, not an existing convention). 2. **The print-production manager's "spec sheet vs. what comes off the line" question maps onto thermometer calibration.** A press runs to a target curve on paper and then you measure the actual sheet; a thermometer has a printed accuracy spec and you have to check it against a known temperature (ice water at 32°F, boiling water at 212°F) to know whether the unit in your hand actually hits it. FSIS's primer gives the exact method **verbatim**; ThermoWorks' accuracy article makes the same point and adds that ASTM recommends testing in an ice bath rather than in food, because food is never at one stable temperature. **Kept**: calibration is a judgement (not a facet on its own, since most instant-reads either can or can't be calibrated depending on model, per FSIS), and `calibration_adjustable` is a boolean facet. **Dropped**: a numeric "drift rate" facet — no source gives a number for how fast a given sensor type drifts, only that physical shock and thermal extremes cause it (ThermoWorks). ## Sources and what each supports ### Regulators and standards bodies **FDA Food Code 2022 (PDF, downloaded and read with `pdftotext`)** https://www.fda.gov/media/164194/download - verbatim § 4-203.11: "(A) FOOD TEMPERATURE MEASURING DEVICES that are scaled only in Celsius or dually scaled in Celsius and Fahrenheit shall be accurate to ±1°C in the intended range of use. (B) FOOD TEMPERATURE MEASURING DEVICES that are scaled only in Fahrenheit shall be accurate to ±2°F in the intended range of use." - verbatim § 4-502.11 "Good Repair and Calibration": "if temperature or pressure measuring devices are not maintained in good repair, the accuracy of the readings is questionable." - verbatim Annex 5 (risk-based inspection guidance): "Infrared thermometers are inappropriate for measuring internal cooking temperatures." - verbatim Annex 7 inspection form note: "Food temperature measuring devices that are scaled only in Fahrenheit should be accurate to ±2°F in the intended range of use... scaled only in Celsius or dually scaled... should be accurate to ± 1°C." - Supports: the regulatory accuracy floor used in `stated_accuracy_f` and judgement 3; the calibration/good-repair requirement behind `calibration_adjustable`; the infrared exclusion in judgement 2. Confidence high — primary regulatory text, read directly. **USDA FSIS, Safe Minimum Internal Temperature Chart (Wayback Machine snapshot; live fsis.usda.gov returned 403)** https://web.archive.org/web/20250102085348/http://www.fsis.usda.gov/food-safety/safe-food-handling-and-preparation/food-safety-basics/safe-temperature-chart - verbatim table: "Beef, Pork, Veal & Lamb Steaks, chops, roasts — 145°F (62.8°C) and allow to rest for at least 3 minutes. Ground Meats — 160°F (71.1°C). Ground Poultry — 165°F. All Poultry — 165°F (73.9°C). Eggs — 160°F (71.1°C). Fish & Shellfish — 145°F (62.8°C). Leftovers — 165°F (73.9°C). Casseroles — 165°F (73.9°C)." - Supports: the safe-temperature reference used for the `temp_range_max` visual grammar's overlay band and for grounding `intended_use: everyday_meat_baking`. Confidence high. **USDA FSIS, Kitchen Thermometers primer (Wayback Machine snapshot; live page 403)** https://web.archive.org/web/20240104031004/http://www.fsis.usda.gov/food-safety/safe-food-handling-and-preparation/food-safety-basics/kitchen-thermometers - verbatim: thermocouples "respond so rapidly" and are "not designed to remain in the food while it's cooking... Thermocouples can be calibrated for accuracy." - verbatim: thermistors take "roughly 10 seconds to register the temperature" and are "not designed to remain in the food while it's cooking... Not all thermistors can be calibrated." - verbatim: oven-cord thermometers (thermistor-type, cord to a base unit) "cannot be calibrated." - verbatim: "Oven-safe" bimetal thermometers sense "from its tip and up the stem for 2 to 2 1/2 inches. The resulting temperature is an average... if not left in the food while cooking, they can take as long as 1 to 2 minutes to register the correct temperature... it is not appropriate to measure the temperature of any food less than 3 inches thick." The "potato nail effect" (heat conducted along the stem reads hotter than the food beside it) is named and described. - verbatim: "Instant Read" bimetal thermometers measure "in about 15 to 20 seconds... Some models can be calibrated." - verbatim: candy/jelly/deep-fry thermometers "measure temperatures ranging from 100 to 400°F." - verbatim: "Most available food thermometers will give an accurate reading within 2 to 4°F." - verbatim calibration method: ice-water bath to 32°F or boiling water to 212°F, stem immersed 2 inches, wait 30 seconds minimum before adjusting the calibration nut; "Even if the food thermometer cannot be calibrated, it should still be checked for accuracy using either method." - Supports: the whole mechanism section's response-time figures for thermistor/bimetal/instant-read bimetal; the "not designed to remain in food" basis for `device_type: instant_read`'s branch; the calibration method in judgement 1; the classic candy-thermometer range used alongside TempPro's modern digital range; the consumer accuracy baseline in judgement 3. Confidence high — primary government primer, read directly. **Food Standards Australia New Zealand, Thermometers (consumer/business guidance page)** https://www.foodstandards.gov.au/business/food-safety/thermometers - verbatim: "Under Standard 3.2.2 — Food Safety Practices and General Requirements, food businesses... need to have an accurate and accessible thermometer... the thermometer is accurate to within 1°C." - verbatim: "Infrared 'gun' thermometers are useful for quick checks and for packaged food — but only measure the surface temperature." - verbatim: "Place the probe into the thickest part of the food and wait until the temperature stabilises before reading it... have it calibrated regularly." - Supports: a second, independent regulator's accuracy floor (±1°C, close to FDA's ±1°C/±2°F) and the infrared-is-surface-only claim in judgement 2 (paired with the FDA Annex 5 line, a different regulator making the same call independently — sources agree). Confidence high. ### Specialist reviewer **CNET, Best Meat Thermometers of 2026** https://www.cnet.com/home/kitchen-and-household/best-meat-thermometers/ - verbatim test method: "We use a medical-grade thermometer as our baseline for accuracy and test the meat probe variance to that temperature... By filling a tall glass with ice cubes and adding water... you can bring that water down to 32 degrees Fahrenheit... we insert both the medical-grade probe and the meat probe... and stir it until the medical-grade thermometer reads 32 degrees Fahrenheit." - verbatim: "Accuracy has really improved in the past few years, with most of the top brands claiming their sensors are within plus or minus 0.5 degrees Fahrenheit." - verbatim (ThermoPro Lightning unit tested): "delivers highly accurate thermometer readings with a plus-or-minus 0.5-degree Fahrenheit accuracy in 1 second." - Supports: a second, independent (non-brand) source for the ice-bath calibration test (paired with FSIS) and for the ±0.5°F premium-accuracy figure (paired with ThermoWorks) — both pairs are two-sourced. Confidence high. ### Brand explainers and spec sheets (ThermoWorks) **ThermoWorks blog, Accuracy: How to Know If Your Thermometer is Accurate** https://blog.thermoworks.com/accuracy-how-to-properly-compare-thermometers/ - verbatim: "Thermapen ONE has an industry-leading accuracy spec of ±0.5°F (±0.3°C) whereas ThermoPop 2... has an accuracy spec of ±1°F (±0.5°C) over much of its range." - verbatim: "the American Society for Testing and Materials (ASTM)... recommends testing thermometer accuracy and response time in a properly made ice bath." - verbatim: thermocouple sensors use "changes in the amplitude of voltage"; thermistor sensors use "changes in the speed of the flow of electrons." - verbatim: "Type K thermocouples in your Thermapen[] are known to maintain their accuracy more durably than thermistors." - Supports: the premium-accuracy figure (paired with CNET), the ice-bath test standard (paired with FSIS/CNET), the thermocouple-vs-thermistor mechanism and judgement 1. Confidence high for what the page states; it is a thermometer brand describing its own products favourably, so the specific ±0.5°F figure is given as ThermoWorks' own product spec, not a category-wide claim. **ThermoWorks blog, Thermometer Speed — How to Properly Compare Instruments** https://blog.thermoworks.com/thermometer-speed-how-to-properly-compare-instruments/ - verbatim: "Thermapen ONE takes a full reading in one second or less... Typical digital thermometers available on Amazon or at big box stores take 10 or 15 seconds for a full reading." - verbatim: "electronic thermometers are faster than mechanical thermometers... thermocouple sensors are typically faster than resistance technologies (like the thermistor or the RTD)... reduced tip probes are faster than standard-diameter probes." - verbatim: a "time constant" of 0.5 s implies "a full reading with that probe at 2.5 seconds (or five times the listed time constant)." - Supports: the response-time range and the thermocouple/thermistor speed ordering in mechanism.families; the "10-15s typical" figure (paired with FSIS's 10s thermistor / 15-20s instant-read dial figures). Confidence high for the brand's own claims about its product and about typical competitor speed. **ThermoWorks blog, How to Choose the Right Alarm Thermometer: a Guide** https://blog.thermoworks.com/how-to-choose-the-right-alarm-thermometer-a-guide/ - verbatim shared spec across their leave-in alarm line: "High Temperature Alarm Accuracy ±1.8°F... Temperature Range from –58 to 572°F... Water Resistance Rating to IP65." - verbatim: "BlueDOT has a line-of-sight range of 95 ft. and a transmission rate from 2-60 sec." - verbatim: "Smoke X2 features next generation RF technology and a line-of-sight signal range up to 6,562 feet away." - verbatim: SquareDOT's second channel "Average mode shows you a running average temperature calculation of the last 15 minutes." - Supports: `stated_accuracy_f` typical range at the budget/mid end (±1.8°F, paired with TempPro's own ±1.8°F spec below), `temp_range_min`/`temp_range_max` typical band, `wireless_range_ft` at the Bluetooth end (95 ft) and the long-range dedicated-RF end (6,562 ft — a second brand independent of TempPro, so the "RF can reach far beyond Bluetooth" claim is two-sourced across two vendors). Confidence high. **ThermoWorks blog, Why Professionals Don't Use Wireless Ambient Sensors** https://blog.thermoworks.com/why-professionals-dont-use-wireless-ambient-sensors/ - verbatim: "the ambient sensor on the end of it is reading 50°F lower than you expected... the heat sink effect and evaporative cooling." - verbatim: "RFX performed the best with an average of being within 5.25°F of the control temperature" in "an ambient temperature test The Barbecue Lab did with 9 different wireless probes," and "ThermoWorks is 4 times more accurate than the competition when using a wired ambient probe." - Used for judgement 4 (the heat-sink/evaporative-cooling mechanism only). The specific numbers (5.25°F, "4 times more accurate") are **not** used in category.json: they come from a third-party test (The Barbecue Lab) that I could not independently fetch (`bbqlab.txt` — Cloudflare "Just a moment..." challenge page, no content), and ThermoWorks is both citing that test and selling the wired alternative it favours. One brand's citation of an unseen third party does not meet the two-source bar for a number. Confidence medium for the general physical claim (heat sink + evaporative cooling), low for any specific figure (not used). **ThermoWorks blog, The Best Candy Thermometers: What You Need to Know** https://blog.thermoworks.com/candy-thermometers/ - verbatim: "The difference between the 'Firm Ball' stage and the 'Hard Ball' stage, or between 'Hard Ball' and 'Soft Crack' is literally one or two degrees." - verbatim: "Thermapen® ONE is accurate to within ±0.5°F (0.34°C) throughout the entire candy-making range... speeds of 2-3 seconds per reading." - verbatim: "ThermoPop®... ±2°F(1°C) accuracy is still tight enough for candy work." - verbatim (on classic glass/dial candy thermometers): "It is slow to come to temperature. It is inaccurate and falls out of calibration... The many sharp edges, nooks, and crannies on a traditional glass thermometer can easily cause seed crystal formation." - Supports: judgement 5 (candy needs tighter accuracy because stage windows are narrow) and the `intended_use: candy_frying_oil` branch rationale. Confidence medium — detailed and physically plausible, but a single brand source for the "1-2°F stage width" specific claim; not independently verified elsewhere in this run, so it stays a medium-confidence judgement rather than a numeric facet threshold. **ThermoWorks blog, What Kind of Thermometer Do I Need?** https://blog.thermoworks.com/kind-thermometer-need/ - verbatim: "An instant-read thermometer allows you to quickly check the temperature of a food in different places in rapid succession at a single point in time, while a cooking alarm thermometer allows you to track changes in temperature at a single fixed point in the food over time." - verbatim (comment thread, a customer and ThermoWorks staff): "the third general type is an infrared (IR) thermometer for accurately assessing the temperature of cooking surfaces... infrared measuring doesn't work properly on shiny surfaces like stainless steel... if there is reflection the light collected by the sensor is... 'diluted'." - Supports: the core instant-read vs. leave-in distinction that frames `device_type`, and the infrared reflective-surface caveat kept in the infrared mechanism family and `notes` (not a numeric claim, so no second source needed). Confidence high for the instant-read/alarm distinction (matches FSIS independently); medium for the shiny-surface caveat (single source, but physically standard emissivity behaviour, not a disputed claim). **ThermoWorks blog, Probe Safety: A Guide** https://blog.thermoworks.com/probe-safety-a-guide/ - verbatim: "probes rated to 640°F [can] burn out in a low-and-slow smoker that is running at 275°F... When wood burns, it does so at temperatures in excess of 1,100°F (593°C), hot enough to burn out either the sensor or the cable insulation." - verbatim: common mistakes — threading cables through a chimney (exhaust heat is "concentrated" and can exceed the cable's rating), draping cables across metal grates ("the metal in the grates can channel heat"), and letting a probe transition sit over a water pan's steam. - Supports: the `probe_max_temp_f` facet's rationale (a probe's temperature *rating* is for ambient diffuse heat, not direct flame or exhaust) and a judgement about probe placement. The specific "640°F" and "1,100°F" figures are illustrative numbers from one brand's own blog post and are not repeated by a second source in this run, so neither number is placed in category.json; only the general, physically well-established point (open flame and flue gas run far hotter than a smoker's set temperature) is kept, in `probe_max_temp_f.why_experts_care` and notes. Confidence high for the general point, low (single-source) for the two illustrative figures. ### Brand spec sheets (MEATER) **MEATER, MEATER Plus product page and FAQ** https://meater.com/products/meater-plus - verbatim: "The internal sensor supports a maximum temperature of 212°F... The ambient sensor can handle a maximum temperature of 527°F." - verbatim: "Alpha Collection probes have a maximum ambient temperature of 527°F, while Pro Collection probes go up to 1,000°F... Alpha Collection probes also use a two-sensor design rather than the six-sensor system found in Pro Collection probes." - verbatim: "MEATER Block — The premium Alpha Collection option, featuring four probes, built-in WiFi for whole-home range." - verbatim: "MEATER Plus and MEATER SE have a Bluetooth range of up to 165 ft/50 m, depending on obstructions." - verbatim: "Rechargeable: Over 24 hours of continuous cooking... Probe contains Li-ion Battery." - Supports: `probe_max_temp_f` range (527°F Alpha / 1,000°F Pro — a second brand independent of the TempPro/CHEF iQ listing figure below, so "premium ambient probes reach about 1,000°F" is two-sourced), `probe_count` (up to 4, MEATER Block), `wireless_range_ft` (165 ft Bluetooth), `wireless_connectivity` (Bluetooth vs. built-in WiFi), `truly_wireless` (no cable, rechargeable battery in the probe itself), and the `multi_sensor_array` mechanism family (six-sensor Pro design). Confidence high — read directly from the brand's own FAQ. ### Brand spec sheet (Combustion Inc.) **Combustion Inc., Predictive Thermometer product/FAQ page** https://combustion.inc/products/predictive-thermometer - verbatim: "Your Predictive Thermometer sends readings by Bluetooth. A WiFi Booster, WiFi Display or Giant Grill Gauge receives those readings and sends them on to your phone... Beyond Bluetooth range: WiFi... you can check food temperatures and time until ready wherever your phone has internet access." - verbatim: "the second-generation thermometer transmits at +8 dBm, with an unobstructed Bluetooth range rated at 330 feet (100 meters)." - verbatim: "Each WiFi accessory supports up to four Predictive Thermometers... Linked Combustion tools form MeatNet, relaying readings between them." - verbatim: "With its 900°F max temperature rating and brazed hermetic seal, the Predictive Thermometer doesn't just survive high heat." - "8 Sensors" (product page heading). - Supports: a third independent brand for the `multi_sensor_array` mechanism family (8 sensors, vs. MEATER's 6); a second independent brand (after MEATER) for "WiFi gives effectively unlimited range once connected to the internet, Bluetooth is short-range"; `probe_max_temp_f` at 900°F (close to, and corroborating, MEATER Pro's 1,000°F and CHEF iQ's 1,000°F listing claim below). Confidence high for the brand's own stated specs. ### Brand spec sheets (TempPro — formerly ThermoPro's direct-to-consumer line; pages note "Your order may arrive in packaging with our previous brand name") **TempPro, TempSync PLUS TP829B Wireless Meat Thermometer product page** https://temppro.com/a/shop/products/tp829b-wireless-meat-thermometer - verbatim: "Wireless Range: 1000ft. Temperature Measurement Range: 14°F to 572°F (-10°C to 300°C). Measurement Accuracy: ±1.8°F (±1.0°C). Sensor Type: NTC." - verbatim: "No apps, no complex steps... Equipped with advanced RF technology." - Supports: `wireless_range_ft` at the dedicated-RF end (1,000 ft — paired with ThermoWorks' Smoke X2 6,562 ft as the second, independent-brand source that RF substantially outranges Bluetooth); `temp_range_max` at 572°F; `stated_accuracy_f` at the budget/mid end (±1.8°F, matching ThermoWorks' own alarm-line spec exactly); `sensor_type: thermistor_ntc`. Confidence high for the brand's own spec table. **TempPro, TempTrack CORE TP16/TP16S Leave-In Meat Thermometer product page** https://temppro.com/a/shop/products/tp16-leave-in-meat-thermometer - verbatim: "a 40" stainless steel mesh cable and a 6.5" food-grade stainless steel probe that can withstand temperatures up to 716°F (380°C)." - verbatim: "Temperature Measurement Range: 32 to 572°F (0 to 300°C). Measurement Accuracy: ±1.8°F (±1.0°C). Sensor Type: NTC. Refresh rate: 2 seconds." - Supports: `probe_max_temp_f` at the leave-in end (716°F, a different brand/model from MEATER's 527–1,000°F figures, widening the sourced band); `temp_range_min`/`temp_range_max`/`stated_accuracy_f` corroboration (same ±1.8°F figure as ThermoWorks' alarm line, independently). Confidence high. **TempPro, TempDip ELITE TP511 Digital Candy & Deep-Fry Thermometer product page** https://temppro.com/a/shop/products/tp511-candy-thermometer - verbatim: "Temperature Measurement Range: -58 to 572°F (-50 to 300°C). Measurement Accuracy: ±1.8°F (±1.0°C)... Response Time: 5 seconds." — but the page's own marketing copy elsewhere says "accurate to ±0.9°F" and a comparison table on the same page gives "±0.9°F(±0.5°C)" for this exact model. **This is an internal contradiction on the brand's own page** (±1.8°F in the spec table vs. ±0.9°F in the prose and comparison table); logged under "Corrections" below and not relied on for a single-sourced number — the facet's typical range instead uses the figures that are repeated identically across TP829B, TP16 and TP511's spec tables (±1.8°F) plus ThermoWorks' matching ±1.8°F alarm spec. - verbatim: "IPX5-rated waterproof design," "Calibration Function for Lifetime Accuracy," "NSF Certified for Safety and Quality." - Supports: `temp_range_min` (-58°F, repeated on TP829B too — same brand's whole NTC line, not cross-brand, logged as such); `waterproof` IP-code example (IPX5); `calibration_adjustable` listing language ("Calibration Function"). The NSF claim is noted in vocabulary only, not made a facet (see "Against the existing add-on schema"). Confidence medium — brand's own page, with an internal number conflict noted above. **TempPro blog, Truly Wireless vs. Wireless Meat Thermometer: Which to Choose** https://temppro.com/blogs/news/truly-wireless-vs-wireless-meat-thermometer-which-to-choose - verbatim: "A standard wireless meat thermometer still uses a physical wire, but only part of the way... The wireless communication happens between the base unit and your phone. The probe itself remains physically tethered by a wire." - verbatim: "A truly wireless meat thermometer removes that final cable entirely... Every component of the temperature sensing and transmission system is contained within the probe." - verbatim (TP920 wired-probe example): "a robust 650-foot Bluetooth range"; (TP980 truly-wireless example): "unlimited monitoring range through your home Wi-Fi network... Supporting up to 4 probes... extreme accuracy (±0.5°F)... The probe itself must be recharged, typically after 24-50 hours of use." - Supports: `truly_wireless` as a real, named distinction in the trade (not a generic wired/wireless split — the brand itself markets "TempSync" vs. "TempSpike" lines on exactly this axis); `wireless_range_ft` at a second Bluetooth data point (650 ft, a boosted implementation, alongside MEATER's 165 ft and ThermoWorks' 95 ft — the figure varies a great deal by implementation, which the facet's `typical` text says explicitly); trade-off 1 (battery/recharge cost of going cable-free). Confidence high for the brand's own distinction and specs, medium for "±0.5°F" as a single-model claim. **TempPro blog, How Important Is a 500' Range to a BBQ Griller?** https://temppro.com/a/shop/blogs/news/how-important-is-a-500-range-to-a-bbq-griller - verbatim: "This innovative device uses Bluetooth 5.2 technology to establish a strong and stable wireless connection over a 500' range." - Supports: a third Bluetooth range data point (500 ft, "the longest in the market" per the brand) — read in context with the TP920 (650 ft) and MEATER (165 ft) figures, confirming that rated Bluetooth range varies by a factor of 4-5x across real products rather than being one fixed number. Confidence high for the brand's own claim about its own product; medium for "longest in the market" (marketing superlative, not verified against every competitor). **TempPro blog, How to Choose Cooking Timers: 6 Features to Look For** https://temppro.com/blogs/news/how-to-choose-cooking-timers - verbatim: "look for a loud and adjustable alarm, a large easy-to-read display, simple controls, multiple timers, flexible placement, and reliable battery life." - verbatim (TM02 spec example): "Pick from three alarm volumes (70–105dB) or switch to silent mode... Each timer can be set for up to 23 hours 59 minutes." - Supports: `timer_channel_count` (dual/multi-timer as a named feature) and `timer_loudness_db` (70-105 dB). Confidence medium — a brand buying-guide article naming its own product's numbers; no second brand's timer dB spec was found in this run (see "Could not reach"), so `timer_loudness_db` carries only one sourced number and is marked `backfill_cost: high` and `researched` rather than treated as a settled category-wide range. ### Retail listings (what data is already present — read for presence only, per `references/sources.md`) **Amazon search results, "meat thermometer" (titles read from the fetched results page; product-detail tables did not render)** https://www.amazon.com/s?k=meat+thermometer - Read directly: ~40 titles including "Meat Thermometer Digital, 1-2 Sec Instant Read"; "CHEF iQ Sense... WiFi Unlimited Range | 1000°F Heat Safe | 5 Sensors"; "ThermoMaven 1 Sec Professional Digital Meat Thermometer, ±0.5℉ Accuracy"; "MEATER Plus Smart Meat Thermometer, Honey, Bluetooth, 165ft"; "Smart Wireless Meat Thermometer with Display Dock, 820ft Long Range Bluetooth 5.3... IP67 Waterproof Dual Sensor... Ceramic Handle"; "Escali AH1 Stainless Steel Oven Safe Meat Thermometer... NSF Certified"; "OXO Good Grips Thermocouple Digital Food Thermometer"; "TempPro TempSwift 2 Probes Meat Thermometer Digital with 2 Display Channels." - Supports: `response_time_s` is reliably in titles (direct numbers: "1-2 Sec", "0.5 Sec", "1 Sec"), reclassifying it to `source: listing` with high confidence. `wireless_range_ft`, `probe_count`, `wireless_connectivity`, and `probe_max_temp_f` are each in at least one title verbatim ("165ft", "820ft", "2 Probes", "Bluetooth", "WiFi Unlimited Range", "1000°F Heat Safe"), so each is marked `listing` for premium/wireless models specifically, with the hint noting it is inconsistent across the catalog. `stated_accuracy_f` and `sensor_type` ("Thermocouple") appear in only a handful of titles, so both stay `researched`. "NSF Certified" appears in a title but is not made a facet (AGENTS.md rule 5: certifications are cut on sight). Confidence high for presence; this is retail-listing evidence used only for the `source`/`listing_field_hint` fields, per `references/sources.md`. **Amazon search results, "oven thermometer"** https://www.amazon.com/s?k=oven+thermometer - Read directly: titles including "Stainless Steel Oven Thermometer 50-300°C/100-600°F"; "Meat Thermometer for Oven, Analog Leave-in... 40-220°F for Beef Poultry Pork Lamb Seafood"; "Digital Oven Thermometer -58°F to 572°F/-50°C-300°C... High Low Temperature Alarm"; "Gozney Infrared Thermometer, Built-in Laser Thermometer with LED Screen." - Supports: `temp_range_min`/`temp_range_max` are reliably in oven/leave-in-dial titles (explicit ranges), so these two facets are marked `listing` for that device type specifically, with the hint noting instant-read titles rarely state the full range. Confirms an `infrared` device_type exists as a distinct retail line (Gozney). Confidence high for presence. **Amazon search results, "kitchen timer"** https://www.amazon.com/s?k=kitchen+timer - Read directly: titles including "TempPro TM02 24H Kitchen Timer with Clock Mode & Dual Digital Timers"; "FMP Digital 4 Channel Commercial Kitchen Countdown Timer"; "Digital Kitchen Timer - 4-Channel Cooking Timer with Alarm, Magnetic Back, Loud & Silent Modes"; "60 Minute Wind Up Timer... 100% Mechanical." - Supports: `timer_channel_count` is reliably in titles ("Dual", "4 Channel"), marked `listing`. Mechanical-vs-digital timer type is confirmed as a real retail distinction but is not made a facet — it tracks `device_type: timer_only` plus `sensor_type`'s absence, and the mechanism-section distinction (spring vs. quartz/electronic) is enough; a mechanical/digital facet would fail the generic test (every battery-powered gadget has this split). Magnetic-back mounting is visible in product photos and is cut per AGENTS.md rule 1 (visually-derived fields stay out of the schema). Confidence high for presence. ### Mechanism and vocabulary (encyclopedic) **Wikipedia, IP code** https://en.wikipedia.org/wiki/IP_Code - verbatim: "a cellular phone rated at IP67 is 'dust resistant' and can be 'immersed in 1 meter of freshwater for up to 30 minutes.'" IPX5 = "Splashing of water... from any direction shall have no harmful effect." - Supports: the IP-rating vocabulary entry and the proposed IP-code refinement of the existing `waterproof` boolean (see "Against the existing add-on schema"). Confidence high — standards-derived encyclopedia entry, cites IEC 60529. **Wikipedia, Candy thermometer** https://en.wikipedia.org/wiki/Candy_thermometer - verbatim: "A candy thermometer... can read higher temperatures, usually 400°F/200°C or more... Types of candy thermometers include liquid thermometers, coil spring or 'dial' thermometers... and digital thermometers." - Supports: a second, independent source (alongside FSIS) for "candy thermometers read higher than general cooking thermometers, around 400°F+." Confidence high for the general claim. ## Data points used for derived metrics No `derived_metrics` are proposed for this category. Unlike formulations (cost per serving) or machines (watts per dollar), nothing in this category's facets combines into a number an expert computes silently while shopping — accuracy, speed, range and temperature ceiling are each read on their own, not divided or multiplied against another facet or against price. Forcing a derived metric in here would fail the generic test (inventing a "dollars per degree of accuracy" number nobody actually uses). This is a deliberate omission, not an oversight. ## Against the existing add-on schema `zamana-expertise/data/schema-parts/supplies/kitchen/add-ons/thermometer.jsonc` today has only: `temp_range_min`/`temp_range_max` (value + f/c unit), `read_speed_seconds`, `waterproof` (boolean), and `sizing.probe_length`. Read alongside `kitchen-standard.jsonc` (materials-by-role, `max_heat_temp`, `dishwasher_safe`) and `supplies-standard.jsonc` (`high_signal_markdown`, generic `sizing`). AGENTS.md field-kind vocabulary used below: **vertical anchor**, **category quality currency**, **hard filter**, **sizing**, or **noise/cut**. | Facet in this register | In add-on today | Proposed AGENTS.md kind | Note | |---|---|---|---| | `response_time_s` | yes, `read_speed_seconds` | category quality currency | Already there and well-named; this register confirms it is reliably in titles (listing, not just researched) — worth promoting to a currency the pipeline commits to backfilling, not just storing when present. | | `stated_accuracy_f` | no | category quality currency | The single number ThermoWorks, CNET and TempPro all treat as the headline spec; absent from the add-on entirely. Mostly on brand pages, not titles — researched, low backfill cost for the top N. | | `temp_range_min`/`temp_range_max` | yes | sizing-adjacent hard filter | Already there; keep both. Confirmed reliably on oven/leave-in-dial listings, less so on instant-read titles — the hint should say so rather than assume uniform coverage. | | `waterproof` | yes (boolean) | hard filter | Keep the boolean for backward compatibility, but propose upgrading to an IP-code enum (`ip20`, `ipx5`, `ip65`, `ip67`, `none`) since real listings already state the code (IP67, IPX5 seen in titles/spec pages) and the code carries more than a yes/no (splash vs. full immersion matters for a probe that gets dunked in brine). | | `sizing.probe_length` | yes | sizing | Keep as-is; no change proposed. | | `sensor_type` | no | category quality currency (secondary) | Rarely on listings, reliably on spec pages as "Sensor Type: NTC" or similar. Explains *why* speed/accuracy differ, so worth the research cost for the top N, not a blanket filter. | | `wireless_connectivity` | no | hard filter | Real buyers filter on "works with my phone" vs. "just a display" — Bluetooth/RF/WiFi/none. Reliably in titles for wireless models. | | `wireless_range_ft` | no | category quality currency | Reliably in titles for wireless models specifically ("165ft", "820ft", "1000FT"); varies 4-5x across implementations even within one technology, so the typical-range text must say so rather than imply one number per technology. | | `truly_wireless` | no | hard filter | Not usually spelled out in titles; inferred from the product line name (TempSpike vs. TempSync; MEATER and Combustion are always cable-free). Worth the research cost — it is the #1 axis TempPro's own buying guide organizes around. | | `probe_count` | no | hard filter | Reliably in titles ("4 Probes", "Dual Probe", "2 Display Channels"). Cheap, high-value, currently invisible to the schema. | | `probe_max_temp_f` | no | category quality currency | A genuinely different number from `temp_range_max` (reading range vs. survival rating) that no add-on field currently captures; occasionally in titles for premium wireless probes ("1000°F Heat Safe"), otherwise brand-page only. | | `calibration_adjustable` | no | hard filter (boolean) | Rarely a title bullet, sometimes a brand-page bullet ("Calibration Function"); FSIS explicitly varies this by sensor type and model, so it is not safely inferable from `sensor_type` alone. | | `timer_channel_count` | no (no timer facets at all today) | hard filter | Reliably in timer titles ("Dual", "4-Channel"). The add-on schema has no timer-specific fields; this and the next are the minimum needed to not lose the "& Timers" half of the leaf. | | `timer_loudness_db` | no | category quality currency | The single feature buying guides call most important for a timer and the thing no title states; single-brand sourced in this run (see "Could not reach"), so backfill cost is marked high rather than low. | Add-on fields this register considers noise for this category: none of the four existing fields are noise — all four are kept and three are confirmed by retail-listing evidence to belong where they are. The one *candidate* field this research actively rejects is a dedicated **NSF/ANSI 51 or other certification field**: it appears on real listings (`"NSF Certified"` seen twice in titles) but AGENTS.md rule 5 cuts certifications on sight as Jev/markdown material, and the FDA Food Code's own accuracy requirement (§4-203.11) is the governing standard that actually drives a buying decision here, not a voluntary materials certification. NSF stays in the vocabulary section only. ## Branches Traced to the register sentence or a stated chain of sourced facts, per leg: - `device_type = instant_read` → then_ask `response_time_s`, `stated_accuracy_f`, `sensor_type`; makes irrelevant `probe_count`, `wireless_connectivity`, `wireless_range_ft`, `truly_wireless`, `probe_max_temp_f`, `timer_channel_count`, `timer_loudness_db`. Chain: FSIS says instant-read sensors (thermocouple, thermistor, instant-read bimetal) are each "not designed to remain in the food while it's cooking," so there is no probe to leave in, no wireless link to speak of, and no cable-heat rating to worry about; speed and accuracy are the two numbers ThermoWorks' speed and accuracy articles both treat as *the* comparison. SUPPORTED. - `device_type = leave_in_probe` → then_ask `temp_range_max`, `probe_max_temp_f`, `calibration_adjustable`; makes irrelevant `wireless_connectivity`, `wireless_range_ft`, `truly_wireless`, `timer_channel_count`, `timer_loudness_db`. Chain: FSIS describes leave-in bimetal and oven-cord thermometers as designed to stay in the oven with a base unit, not a wireless link; TempPro's TP16 page shows the two numbers that actually differ across leave-in models are the reading ceiling and the cable's heat rating, and FSIS notes calibratability varies by model for this type specifically. SUPPORTED. - `device_type = wireless_probe` → then_ask `wireless_connectivity`, `wireless_range_ft`, `truly_wireless`, `probe_count`, `probe_max_temp_f`. Chain: this is exactly the axis MEATER's FAQ and TempPro's "truly wireless vs. wireless" guide organize their entire product lines around; `probe_max_temp_f` matters most here because wireless probes are the ones marketed for open-flame/searing use (MEATER Pro, CHEF iQ). SUPPORTED. - `device_type = infrared` → then_ask `temp_range_max`; makes irrelevant `probe_count`, `wireless_connectivity`, `wireless_range_ft`, `truly_wireless`, `calibration_adjustable`, `probe_max_temp_f`, `timer_channel_count`, `timer_loudness_db`, `temp_range_min`. Chain: FSANZ and the FDA Food Code (two independent regulators) both say infrared only reads a surface, so there is no probe at all — no count, no cable, no wireless link, no "calibration" in the probe sense. The reading ceiling still matters (a griddle or pizza-oven surface check needs real headroom). SUPPORTED. - `device_type = timer_only` → then_ask `timer_channel_count`, `timer_loudness_db`; makes irrelevant `stated_accuracy_f`, `temp_range_min`, `temp_range_max`, `waterproof`, `sensor_type`, `wireless_connectivity`, `wireless_range_ft`, `truly_wireless`, `probe_count`, `probe_max_temp_f`, `calibration_adjustable`, `response_time_s`. Chain: a standalone kitchen timer has no temperature sensor of any kind (confirmed by the Amazon "kitchen timer" title sweep — none of the sampled titles mention a probe or temperature spec); every temperature-related facet is simply inapplicable, not merely unimportant. SUPPORTED. - No branch for `device_type` + any other combination: not applicable, all five values are covered above. - `intended_use = candy_frying_oil` → then_ask `temp_range_max`, `stated_accuracy_f`; makes irrelevant `probe_count`, `wireless_connectivity`, `wireless_range_ft`, `truly_wireless`, `timer_channel_count`. Chain: ThermoWorks' candy article says stage boundaries are "literally one or two degrees" apart (accuracy) and FSIS/TempPro both put candy/frying thermometers' ceiling well above everyday cooking (400-572°F); candy and frying are single-pot, single-reading tasks, so multi-probe wireless tracking (built for multiple proteins over hours) does not apply. SUPPORTED. - `intended_use = low_slow_bbq_smoking` → then_ask `wireless_connectivity`, `wireless_range_ft`, `probe_count`, `probe_max_temp_f`. Chain: this is exactly what TempPro's wireless-range guide and MEATER's multi-probe Block model are built for, and ThermoWorks' probe-safety guide explains why the heat rating matters specifically in a smoker's hot spots and chimney. No facet is made irrelevant here (accuracy and speed still matter for a smoker, just less acutely than for candy), so this branch has `then_ask` only. SUPPORTED. - No branch for `intended_use = everyday_meat_baking`: the rationale itself says this is the default path — FSIS's safe-temperature chart and FDA's ±2°F floor already describe the baseline every thermometer in the catalog should clear, so answering "everyday" narrows nothing further; it just means no extra questions follow. Nothing to trace beyond the rationale. SUPPORTED (trivial). - `response_time_s` (numeric split, no `values` list): no branches, per house convention for numeric splits without an enumerated value set (the `when` field must name a value of the split facet, and a bare numeric facet has none). ## Trimmed detail Facts cut from capped fields that carried a number or a source, kept here per field. - **summary**: "FDA's commercial floor is ±2°F; most consumer units land at ±1 to ±2°F, premium thermocouples reach ±0.5°F" (kept in `stated_accuracy_f.why_experts_care` and judgement 3). - **mechanism.plain**: "the 'potato nail effect', where heat conducted along a bimetal stem makes the area next to the probe read hotter than the food beside it" (FSIS verbatim; kept here and in `sensor_type`'s notes, not spelled out in the capped field). "A time constant is the time to reach 63% of a reading; five time constants give a full reading" (ThermoWorks; kept in the speed source entry above, not in the 150-word mechanism text). - **`device_type.why_experts_care`**: "FSIS: none of the instant-read sensor types are designed to stay in food while it cooks" (kept in Branches instead). - **`intended_use.why_experts_care`**: "classic candy thermometers read 100-400°F per FSIS; modern digital units from TempPro read to 572°F" (kept in `temp_range_max` and the Sources section). - **`wireless_range_ft.why_experts_care`**: the four separate brand figures (95, 165, 500, 650, 820, 1000, 6,562 ft) are kept in the Sources section rather than listed in the capped field, which instead gives the three-tier shape (Bluetooth / RF / WiFi) in words. - **`probe_max_temp_f.why_experts_care`**: ThermoWorks' specific "640°F probe, 1,100°F wood fire" illustrative numbers are not repeated in the facet (single-sourced); the field instead states the general point that the rating is for ambient heat, not direct flame. - **judgement 4's note**: "The Barbecue Lab tested 9 wireless probes; ThermoWorks says RFX averaged within 5.25°F and was 4x more accurate than competitors" is recorded in the Sources section as the specific (unverified) claim, while the judgement itself keeps only the general, physically-grounded mechanism (heat sink + evaporative cooling). ## Could not reach AmazingRibs' thermometer ratings page (both the live site and every Wayback Machine snapshot attempted returned "not archived"/blocked) — the specialist BBQ-community voice this category would most want is missing from this run; CNET's own test methodology (above) substitutes as the specialist-review source. The Barbecue Lab's 9-probe ambient test (Cloudflare challenge page, no content) — see judgement 4. America's Test Kitchen and Consumer Reports thermometer pages (both 404'd on the specific URLs tried). Serious Eats' instant-read thermometer guide (404). The Verge, Engadget, Smoked BBQ Source and a Tom's-Guide-style outlet's thermometer pages (404 or homepage-only redirect). University extension primers from Penn State, Nebraska, Georgia and Ohio State (each 404'd on the specific thermometer-guide URL tried; Ohio State's fetch landed on an unrelated Bacillus cereus page). ThermoWorks' own `thermapen-one` product page rendered as navigation only when fetched (no spec table in the static HTML); its specs are instead sourced from the brand's accuracy and candy-thermometer blog posts, which quote the same figures in prose. No second brand's timer-loudness spec was found, so `timer_loudness_db` remains single-sourced to TempPro. ## Corrections This is the first draft of this register; no prior blind-check pass has run yet. One internal contradiction was caught and resolved during drafting rather than after a blind check: TempPro's TP511 candy-thermometer page states ±1.8°F in its spec table but ±0.9°F in its marketing prose and comparison table for the same model. Resolved by not relying on either single figure from that page; `stated_accuracy_f`'s typical range instead uses the ±1.8°F figure that is identical and repeated across three TempPro models (TP829B, TP16, TP511's own spec table) and ThermoWorks' independently-stated ±1.8°F alarm-line spec — a cross-brand match, not a single disputed number. ## Corrections (2026-10-26) Applied after blind check.md (6 required fixes). - `timer_loudness_db` was single-sourced in violation of "no range without two sources." Fetched ThermoWorks' TimeStick Trio spec page directly (curl, grepped "Alarm Volume: 70 dB at 11.8 inches (300 mm)") and added it as evidence `tw_timestick_trio`; the 70dB floor is now corroborated by a second, independent brand, so the range stays as {70,105}. - Added `probe_cable_length_in` as a new numeric facet (TP16's 40" mesh cable plus a second source, ThermoWorks' DOT page, fetched directly: "Cable Length 47 inches (1.2m)"), referenced from the `leave_in_probe` branch's `then_ask`. - `device_type`'s cold-start question ("Instant-read, leave-in, wireless, or just a timer?") omitted `infrared`; changed to "Instant-read, leave-in, wireless, infrared, or a timer?" in both the split and the facet's own `cold_start_question`. - Added a `vocabulary` entry for "Bimetal dial" (previously used in `sensor_type` values, `mechanism.families`, and trade-off 4 but never defined). - Resolved the `candy_frying_oil` vs. `wireless_probe` conflict on `wireless_connectivity`/`wireless_range_ft` by removing those two from `candy_frying_oil`'s `makes_irrelevant` list, since `device_type=wireless_probe` genuinely asks for them and a wireless candy thermometer is not excluded by the schema. - Rejected the "add a backlit-display facet" half of fix 6: a named differentiator across TempPro's own product lines but a generic digital-display feature (any backlit electronic display, not specific to thermometers), so it fails the register's generic test. Logging the decision here per the check's own "or log as cut" alternative, rather than adding a `## Noise` section for one low-priority item.