The Thermometer That Measures Heat Moving Through Food.
Most cooking thermometers answer one question: what is the temperature where the tip happens to be? Combustion's second-generation Predictive Thermometer turns the probe into a tiny instrument array. Eight sensors watch heat move from the cooking environment through the surface and toward the coldest internal point, while an onboard physics model estimates when the food will reach its target. It does not merely report temperature. It tries to understand the cook.
By Dr. Rob Urban · On heat transfer, cold spots, carryover cooking, food-safety lethality, wireless range, 900°F hardware, open protocols and why eight temperature sensors can reveal more than one perfectly accurate tip
First, the current-model check
Use the second-generation Predictive Thermometer. It is the current Combustion probe and materially improves heat tolerance, sealing, battery life and wireless capability over Gen 1.
The name stays the same, but the hardware does not. Combustion now sells the second-generation Predictive Thermometer as its current core tool. It still uses the eight-sensor architecture and physics-based prediction system that made the original interesting, but the probe is now nitrogen-flushed, hermetically sealed with a brazed ceramic bond and rated IP69K.
The upper half of the current probe is rated from minus 5°F to 900°F, while the lower electronics-bearing half is limited to 220°F and must remain inserted in food. Combustion lists continuous upper-sensor exposure to 750°F and a peak ambient-sensor limit of 900°F. The probe remains just 4.8 millimeters in diameter, with listed sensor accuracy of plus or minus 0.4°F.
The new generation also expands the radio and charging ecosystem. It works with first- and second-generation accessories, supports Combustion's MeatNet architecture and can gain internet-scale monitoring through current WiFi accessories. Combustion's current materials also cite dramatically longer cooking battery life than the original generation, depending on operating mode and accessory configuration.
The first-generation device established the idea. The second generation is the version worth preserving now because it makes that idea much harder to kill with heat, water or distance.
A single temperature can lie without being inaccurate
The problem is not sensor precision. The problem is where the sensor happens to be.
A conventional probe thermometer can be extraordinarily accurate at one point and still give a misleading picture of the food. A roast is not one temperature. Neither is a chicken breast, steak, loaf, turkey or side of salmon.
Heat moves from the cooking environment into the surface, then through the food by conduction. Geometry matters. Bone matters. Fat matters. evaporation at the surface matters. Thickness matters. The side facing a hot pan behaves differently from the side facing circulating oven air. Two locations only a centimeter apart can be at meaningfully different temperatures.
That creates a practical problem. The cook is usually trying to protect the coldest relevant internal region because that point controls doneness and, in many cases, safety. But the person inserting the probe has to find that region by feel.
Miss the thermal center and an accurate sensor faithfully reports the wrong place.
Combustion's central design decision is almost embarrassingly logical: if one sensor can miss the coldest spot, put several sensors along the probe and let software decide which internal reading is currently lowest.
That turns placement error from a catastrophic measurement mistake into something the instrument can often compensate for.
The probe is an instrument array
TrueCore searches for the coldest relevant internal point instead of assuming the tip found it.
The second-generation Predictive Thermometer contains eight temperature sensors distributed along its length. Four sensors below the minimum insertion line can serve as internal food sensors. The upper sensors monitor the region closer to the surface and cooking environment, with the outermost sensing position acting as an ambient reference.
Combustion's TrueCore system looks across the internal sensors and reports the lowest relevant internal reading rather than simply labeling the pointed tip “the core.” That matters because the thermal center can move during cooking.
Imagine a thick steak in a hot skillet. The bottom surface heats aggressively, the upper surface may be cooler, and the geometric center is not guaranteed to be the coldest thermal location at every moment. Flip the steak and the thermal field changes again. A probe inserted slightly off-center may still have another internal sensor closer to the true cold region.
This is a more interesting idea than making the tip sensor faster. It acknowledges that food is a spatial temperature field.
For cooks, the experience becomes simpler even though the measurement system becomes more complex. Insert the probe beyond the minimum line, set a target and let the device determine which internal sensor best represents the coldest point it can see.
Then it does something thermometers normally refuse to do
It predicts the future.
Combustion's Prediction Engine does not merely extrapolate a straight line from the last few temperature readings. The company describes it as a physics-based model of heat flow through the food, continuously updated by what the sensor array observes.
The distinction matters because cooking curves are not linear. The rate at which the center warms depends on the temperature difference between the surface and the interior, the food's geometry and effective thermal properties, evaporative cooling at the surface, changes in heat input and the way the cook manipulates the food.
A steak moved from low oven heat to a ripping-hot pan changes its thermal boundary conditions instantly. A smoker entering the evaporative stall can spend a long time with little apparent core-temperature progress. Opening an oven door changes the environment. Flipping food changes which side receives the strongest heat flux.
The current thermometer watches those changes and revises its remaining-time estimate as the cook evolves. Combustion says a prediction usually appears after roughly the first third of the cook, when enough thermal history exists to model the trajectory meaningfully.
That is an important philosophical choice. The device does not pretend certainty before it has enough evidence.
A useful prediction engine is not the one that guesses earliest. It is the one that knows when the data has become good enough to make a useful estimate.
Measure a temperature field
Multiple sensors show how the food and its surroundings are changing, rather than reducing the cook to one point.
Infer heat flow
The onboard software uses the evolving thermal pattern to construct a physics-based estimate of the cook.
Update remaining time
Change the pan, oven, smoker or orientation and the countdown can correct itself rather than staying married to an obsolete estimate.
Surface temperature explains why the core is doing what it is doing
Ambient oven temperature is not the same thing as the air touching the food.
Recipes routinely describe an oven as “350°F” or a smoker as “250°F,” but the food does not experience a perfectly uniform number. The boundary layer immediately around the food can be cooler because energy is being absorbed, water is evaporating and airflow is imperfect.
Combustion uses the sensor array to identify a surface-region temperature in addition to the core and external ambient readings. That is useful because the surface is the gateway through which heat enters the food.
If the surface sits below the user's target internal temperature, the core cannot eventually rise above it by conduction alone. This is why Combustion's own guidance notes that very low-and-slow cooks may not receive a useful completion prediction until the surface temperature climbs above the target core temperature.
The surface reading can also make the barbecue stall easier to understand. Evaporation from wet meat consumes energy and cools the surface. Pit temperature can remain steady while the meat's surface and core behave differently from what a simple oven-number mental model predicts.
This is one of the thermometer's greatest educational features. It makes heat transfer visible.
A person can watch the gap between ambient, surface and core and begin to see why raising the oven temperature changes cooking rate, why a sear creates a new thermal gradient and why “the grill is 400°F” tells only a small part of the story.
Food safety is time plus temperature
SafeCook replaces the familiar one-number shortcut with lethality tracking.
Most home cooks learn food-safety temperatures as fixed thresholds: poultry to one number, ground meat to another. Those numbers are useful because they are simple, but microbial lethality is fundamentally a time-and-temperature relationship.
At sufficiently high temperatures, pathogenic bacteria die progressively over time. Higher temperatures achieve a given reduction faster. Lower temperatures can achieve the same target reduction if the food remains there long enough under appropriate validated conditions.
Combustion's SafeCook feature uses the thermometer's continuous temperature history to track this accumulated safety effect rather than looking only for one instantaneous endpoint. The company describes the feature as performing the same basic time-temperature lethality mathematics used in commercial food safety, then presenting the result directly to the cook.
This fits the product unusually well because continuous measurement is exactly what a leave-in probe can do better than an occasional spot check. The thermometer knows not only that the core reached a particular temperature, but how long it spent across a whole sequence of temperatures.
The caveat is important. Food-safety calculations depend on the food category, initial contamination assumptions, target pathogen and validated model. A consumer feature should not be treated as permission to improvise preservation, canning or professional HACCP processes outside its intended scope.
Still, the concept is powerful. The traditional endpoint number is an intentionally conservative shortcut. A sensor that knows the entire thermal history can represent the underlying science more faithfully.
Gen 2 is where the hardware catches up to the ambition
A predictive model is useless if the probe dies in a grill, leaks in sous vide or loses radio contact through a metal cooker.
The second-generation probe's most important improvements are brutally physical. Combustion says it is nitrogen-flushed and hermetically sealed with a brazed ceramic bond, earning an IP69K rating. That is the highest ingress-protection level in the standard system, intended for resistance to dust and high-pressure, high-temperature water exposure.
The upper sensing region can tolerate peak temperatures to 900°F, making the current probe far better suited to grilling, broiling and high-heat finishing than the original hardware. The lower half still has a much lower thermal limit because batteries and electronics remain less heat tolerant than a metal probe tip. That is why the minimum insertion line is a real engineering boundary, not decorative paint.
Combustion's current homepage lists the probe at 4.8 millimeters in diameter and plus or minus 0.4°F sensor accuracy. That is impressively thin for a wireless leave-in device containing eight sensors, a processor, radio and battery.
Battery management is also now contextual. Combustion publishes different runtime expectations based on whether the probe is in instant-read, cooking or lower-power modes, with current second-generation marketing citing extremely long cook-time capability under optimized operation.
The result is a probe that feels more like instrumentation than a kitchen gadget. Thermal zones, seals, radio behavior, charge states and ingress protection all have to coexist inside something thinner than many ordinary meat probes.
The radio system is called MeatNet, which is ridiculous enough to be memorable
The serious idea underneath the name is a mesh-like repeater architecture.
Wireless cooking thermometers have an awkward physics problem. Ovens, smokers and grills are large metal enclosures, and metal is very good at attenuating radio signals. The probe is also tiny, which limits antenna size and battery capacity.
Combustion addresses this with a family of repeating accessories. Displays, Boosters and newer WiFi devices can receive probe data and extend it onward. Multiple compatible tools can participate in what the company calls MeatNet.
The architecture is valuable because the probe does not need to do every communication job itself. It can transmit locally to a nearby repeater, while a larger accessory with more power and a better antenna handles the next hop. Current second-generation WiFi accessories can then bridge the data to the internet for effectively unlimited monitoring range.
This modularity also preserves first-generation hardware value. Combustion says first- and second-generation accessories remain compatible, and current setup guidance notes that both probe generations use the same software workflow.
That is an unusually good decision in connected consumer hardware. New radios and WiFi features expand the system without requiring every existing Display or Booster to become electronic waste.
Combustion made the probe unusually open
A kitchen thermometer with an open protocol is a small rebellion against disposable smart hardware.
Combustion publicly emphasizes that its thermometer platform is open source and modular. That matters because connected kitchen products often become dependent on proprietary apps, undocumented cloud services and accessory ecosystems that can disappear long before the physical hardware fails.
The Predictive Thermometer's important calculations happen inside the probe rather than requiring the cloud for every prediction. The app provides graphs, configuration, firmware updates and advanced features, but the core architecture is not simply a dumb sensor streaming raw data to a remote server.
That local intelligence is important for resilience. A thermometer should remain a thermometer even when internet connectivity is poor. A prediction should not need a round trip to a data center while dinner is in the oven.
The open approach is also culturally aligned with the kind of customer Combustion attracts. This is a product for people who want to see all eight sensor traces, understand the API, build integrations and argue about heat transfer for fun.
That may be a small market compared with mass-market kitchen gadgets. It is also the kind of market that creates unusually durable technical communities around tools.
Carryover cooking exposes the limit of “done temperature”
Food does not stop changing when it leaves the heat source.
Pull a roast from an oven and the outside layers remain hotter than the center. Heat continues flowing inward even though the oven door has closed behind it. The core temperature can therefore continue rising after cooking stops.
This carryover effect depends on size, shape, surface temperature and cooking method. A thin steak cooked gently behaves differently from a large roast with an extremely hot exterior. That makes a fixed “pull it five degrees early” rule inherently approximate.
Combustion's multi-sensor architecture is unusually well positioned to reason about carryover because it can see the thermal gradient that drives it. If the surface is far hotter than the core, more energy remains available to move inward after removal. If the gradient is small, carryover should be smaller.
The company's current guidance is still conservative: set the predictive target a little below final doneness and watch the core continue toward the desired number after removal. Combustion has also said it continues developing more complete carryover prediction as more data becomes available.
I like that incompleteness. It reveals what makes the product interesting scientifically. Prediction is not a magic feature added once. It is a model of a physical system that can improve as the company learns more about how cooks, foods and heat sources behave.
The instant-read function is almost a side effect
A device built for thermal history can also answer the old question quickly.
Take the current probe out of its charger and it begins in instant-read mode. Combustion says it can deliver an accurate tip reading in under five seconds, with setup documentation describing typical stabilization in roughly three to four seconds.
That makes the device useful for burgers, fish fillets and quick checks where leaving a probe in place is unnecessary. But the instant-read mode is intentionally the highest-power operating state because the thermometer transmits more frequently.
This is a funny inversion of the traditional category. Most digital thermometers are built primarily for instantaneous spot measurements and only some can remain in food. Combustion is architected around continuous sensing and prediction, with instant read included because the hardware can already do it.
The distinction matters because continuous data creates entirely different capabilities. A single measurement tells you state. A time series tells you process.
Once the thermometer has watched the whole cook, it can estimate slope, identify plateaus, calculate accumulated food-safety lethality and show how interventions changed the thermal path.
That is why the product feels closer to a data logger than a fancy kitchen thermometer.
The founder explains the product
Chris Young came to cooking through science, engineering and Modernist Cuisine.
Combustion Inc. is led by Chris Young, a chef and technical food obsessive whose career has repeatedly sat at the boundary between cooking and applied science. He is widely associated with the Modernist Cuisine project and with the idea that rigorous measurement can make cooking more understandable rather than less creative.
That background is visible in the Predictive Thermometer. The product does not merely chase a more accurate sensor specification. It asks a more scientific question: what would you measure if you wanted to understand the entire heat-transfer process rather than one point?
The first-generation thermometer introduced the eight-sensor probe, TrueCore detection and onboard prediction architecture. The current second generation retains that software philosophy while substantially upgrading physical survivability, radio capability and high-temperature tolerance.
Combustion has also expanded outward from the probe into a broader thermal-control ecosystem. WiFi Displays and Boosters extend communication, the Giant Grill Gauge measures the grill environment, and the Combustion Engine adds wireless draft control for charcoal cookers.
That trajectory is revealing. The thermometer is becoming the sensing heart of a larger closed-loop cooking system.
The company is effectively moving from “measure the food” toward “measure the system, predict the food and eventually help control the heat source.” That is classic control-system thinking translated into barbecue and dinner.
Chris Young develops a public reputation around applying scientific and engineering thinking to professional and experimental cooking.
Combustion launches an eight-sensor wireless thermometer built around TrueCore and an onboard physics-based prediction engine.
Continuous temperature history becomes the basis for integrated time-and-temperature food-safety calculations.
The probe receives major hardware upgrades including IP69K sealing, 900°F peak upper-zone tolerance and improved radio and battery performance.
WiFi accessories, Giant Grill Gauge and Combustion Engine extend the platform from measurement toward networked cooking control.
There are still constraints
Eight sensors reduce placement sensitivity. They do not eliminate physics, food variability or the need to use the probe correctly.
The minimum insertion line matters because the electronics-bearing lower section cannot tolerate the same temperatures as the upper probe. Very thin foods may simply be too small to insert the device safely. A 4.8-millimeter probe is thin for wireless hardware, but it is still larger than some professional needle probes.
The prediction also depends on the cook becoming sufficiently observable. Combustion's own guidance says an estimate usually appears only after meaningful progress has occurred, and low-and-slow conditions can delay prediction when the surface remains below the target core temperature.
Wireless performance is still subject to metal, distance and geometry. That is why Boosters and Displays exist. A signal leaving a probe buried in meat inside a closed metal smoker is solving a much harder communication problem than a phone sitting on a desk.
And the richness of the data can become a distraction. Eight sensor traces, surface temperature, ambient temperature, prediction, safety calculations and wireless accessories are delightful to an enthusiast. Someone who only wants to know whether a burger is 140°F may reasonably prefer a simpler instant-read thermometer.
That is not a flaw so much as product honesty. Combustion built an instrument for people who want to know more about the cook than the final number.
The deeper engineering lesson
More sensors are useful when they measure the structure of a problem, not merely add data.
Consumer electronics often use sensor count as decoration. Add another camera, another microphone, another metric and the specification table looks richer.
The Predictive Thermometer is a better example of multi-sensor design because the arrangement has a physical reason. Temperature exists along a spatial gradient from hot exterior to cooler interior. Multiple sensors along one axis sample that gradient directly.
That geometry enables features a single sensor cannot reproduce through precision alone. The device can identify the coldest internal point among several candidates. It can estimate a surface region. It can see temperature differences along the probe. It can feed a richer state into a prediction model.
This resembles instrumentation in precision manufacturing. The value of measurement often comes not from making one gauge more precise but from measuring the right variables in the right locations so the system's behavior can be inferred.
Combustion's eight sensors are interesting because each helps turn food from an opaque object into a partially observable thermal system.
Once that happens, prediction becomes possible.
Why it goes in the digital time capsule
Because it turns cooking temperature from a number into a model.
The Things Worth Wanting collection is interested in products that make invisible processes legible. Combustion's Predictive Thermometer does that unusually well.
A traditional thermometer tells you one temperature. This probe shows a gradient.
It finds the coldest internal location it can see rather than trusting the insertion point. It estimates surface temperature rather than pretending oven air is what the food experiences. It records thermal history instead of forgetting every measurement the moment it happens. It uses that history to predict completion and calculate food-safety progress.
The second generation makes the underlying idea more credible because the hardware is now better suited to the brutal places cooks actually want to put it. High heat, grills, smokers, vacuum bags and long cooks are not edge cases for a serious thermometer. They are the job.
I also like the direction of the ecosystem. The probe performs its important calculations locally, older accessories remain useful, newer devices can add WiFi, and the company exposes enough of the system that technically curious owners can understand what it is doing.
That feels different from the usual smart-kitchen pattern of placing a mediocre sensor behind an app.
Combustion built the sensor array first, then used software to extract more meaning from the physics.
That order matters.
The Predictive Thermometer belongs in the time capsule because it captures a moment when kitchen instruments stopped merely digitizing old gauges and began using computation to ask better questions about the process itself.
Not “what is the temperature?”
“Where is the coldest point, how is heat moving, when will it be ready, and has the entire thermal history made it safe?”
That is a much more interesting instrument.
A few practical questions
About the current Combustion Predictive Thermometer
Which version of the Combustion Predictive Thermometer is current?
The current product is the second-generation Predictive Thermometer. It retains the eight-sensor and prediction architecture while adding major hardware improvements including IP69K sealing, higher heat tolerance and improved wireless capability.
How many sensors does it have?
The current probe has eight temperature sensors distributed along its length, allowing it to observe internal, surface-region and ambient temperatures.
How hot can the second-generation probe get?
Combustion lists the upper half to 900°F peak, with continuous upper-sensor exposure to 750°F. The lower electronics-bearing half is limited to 220°F and must remain properly inserted in food.
How accurate is it?
Combustion currently lists temperature-sensor accuracy at plus or minus 0.4°F, or 0.2°C.
What does TrueCore do?
TrueCore identifies the lowest relevant temperature among the internal sensors rather than assuming the pointed tip is positioned at the food's coldest spot.
How does the prediction work?
The onboard Prediction Engine uses measured heat-flow behavior to estimate time remaining to a selected target temperature, revising the countdown as cooking conditions change.
Does it need WiFi?
No. Core probe operation and prediction do not require internet access. Bluetooth and MeatNet-compatible accessories handle local communication, while current WiFi accessories can extend monitoring over the internet.
What warranty does it have?
Combustion currently lists a two-year warranty for the second-generation Predictive Thermometer.