“At approximately 10 a.m. local time, the amphibious ship USS Boxer was in international waters conducting a planned inbound transit of the Strait of Hormuz,” read the statement from Jonathan Hoffman, the chief Pentagon spokesperson. “A fixed wing unmanned aerial system (UAS) approached [USS] Boxer and closed within a threatening range. The ship took defensive action against the UAS to ensure the safety of the ship and its crew”

The Pentagon’s statement differs in language from the wording used by President Donald Trump, who claimed that “the drone was immediately destroyed” after receiving multiple calls to stand down. According to CNN, U.S. defense officials claim that the drone was disabled not by bullets or missiles (“kinetics,” in the jargon), but rather by jamming.

Specifically, it was almost certainly jamming via MRZR LMADIS, or a Polaris MRZR vehicle sporting a Light Marine Air Defense Integrated System. While a Corps official said as early as May that the deployment of the system to the Middle East was winding down, observers online noted that a MRZR LMADIS system was visible on the deck of the USS Boxer, an amphibious assault ship, for its transit through the Strait of Hormuz — a presence revealed in pictures uploaded earlier that day.

The incident is just 28 days after Iran shot down a Global Hawk over the Strait of Hormuz. While the drone jammed by the Boxer has yet to be identified, Iran has drone bases in the area (which is, again, a narrow body of water that borders Iran, and over one side of which Iran claims sovereignty). Non-state actors have also operated drones in the surrounding area, sometimes supplied by regional powers. The downing of the Global Hawk almost risked escalation, though the uncrewed nature of drones versus the certainty of human casualties in the event of a retaliation appears to have calmed tensions.

Instead, the recent exchange of attacks on drones hearkens back to an earlier era of tension in the Strait, invoking the attacks on commercial vessels during the Iran-Iraq war which ultimately involved US naval patrols and is recorded in history as the Tanker War. Jamming drones, in theory, poses less of a risk to other commercial or otherwise peopled aircraft, but the likely consequence of the exchange of drone destruction in the summer of 2019 is that commercial aircraft will steer clear of the area. After all, the greatest tragedy of the Tanker War involved a U.S. cruiser mistaking a civilian Iranian airliner for an attacking jet fighter, leading to the deaths of 290 people.

It is also worth noting that the immediate answer to a drone threat to ships appears to be simply parking a ground vehicle carrying a counterdrone system on the deck. The counterdrone field is full of possible solutions, but most have focused on either facilities or dismounted patrol protection, looking at handheld devices or larger installations at fixed points at bases. There is likely an underserved market in counterdrone solutions for ships, especially aimed at stopping mid- to low-end fixed-wing craft.

DARPA, the Department of Defense’s blue-sky agency, launched April 29 a program to use artificial intelligence to best match interventions for individuals. It is called “Teaching AI to Leverage Overlooked Residuals,” or TAILOR.

Specifically, DARPA is looking for submissions about how to use AI for “Human Performance Optimization,” or HPO. Crucially, DARPA is looking for alternatives to one-size-fits-all approaches, because universal recommendations based on group averages can work at cross-purposes to individual need.

“This approach frequently (mis)characterizes individual variance as statistical ‘noise,’ ‘residuals,’ or ‘error,’” reads the solicitation. “The resulting interventions (e.g., diet, physical training regimen, brain stimulation) are at best suboptimal and at worst deleterious for each person.”

Diet, physical training, and brain stimulation aren’t the flashiest parts of the military, but they’re fundamental to how everything else operates. DARPA specifically cites an interest from Special Operations Command’s “Close Combat Lethality Task Force” in getting this optimization right, tailoring interventions to individuals and teams into an advantage.

To get there, DARPA is asking proposers to come up with “third wave AI approaches,” to demonstrate contextual reasoning and transfer learning, meaning a solution is capable of adapting to changed circumstances, environments, and can use related knowledge for abstraction. DARPA also wants teams to move away from traditional AI that trains on large data sets, has opaque processes, and are hard to adapt to new contexts. Proposers will be expected to focus on at least on area of human performance, with DARPA aiming to have a full portfolio of research on physical cognitive, and social performance data.

It’s a lot of technical build-up, but the final outcome is relatively straightforward: after given the data and goal for a specific individual or team, the tool needs to evaluate how much that individual or team would benefit from a given intervention (such as a change in diet or exercise or brain stimulation.) The systems will also be evaluated on how well they can explain or accommodate counterfactuals. Phase II is much the same process, but with datasets from other teams, and with evaluation from subject matter experts in government.

Should the research prove fruitful, the military will be one step closer to specific responses for every individual, a personalized tool that’s general issue.

While it isn’t as flashy as dazzle paint or as iconic as juggalo makeup, camouflage netting has a long history of success in concealing people, buildings and vehicles from aerial surveillance. Swedish defense giant Saab has a camouflage netting just for individuals, designed to keep them hidden from the special cameras carried by drones.

Earlier this year, Marines tested the Barracuda Individual Warfighter Net, according to remarks Steve Simmerer of Saab Defense and Security gave at the Modern Day Marine event in late September. The tests took place at the East Coast School of Infantry.

The net, which weighs 6 pounds and packs up into a pouch, is designed to mitigate certain kinds of radar and thermal imaging, as well as the general visual obfuscation that comes with a mess of netting thrown over a person.

Like all netting, it works best when the person using it isn’t moving. That, combined with a design focus on applying the camouflage netting in the field when a threat is detected, make it somewhat more of a niche item, and one dependent on how well other forces can detect a drone to signal the change. With drones designed to fly silently and resemble wildlife, even troops equipped to hide could be caught unaware.

A new report published today by the Center for the Study of the Drone at Bard University found that in the president’s new budget request, the Department of Defense is asking for three times as many uncrewed vehicles for 2019 as it did in 2018.

That’s 3,447 new drones, to be exact, but don’t expect the skies to go dark with Reapers anytime soon. Despite the numbers requested, the drone budget itself is up by only 26 percent. The Pentagon isn’t just buying more drones, it’s acquiring a whole range of new systems, for everything from undersea surveillance to cheap scouting quadcopters to lasers designed to shoot down the cheap quadcopters of adversaries. The drone century is only just beginning.

“This is the biggest we’ve seen in years that we’ve tracked it,” said Dan Gettinger. Gettinger is co-director of the Center for the Study of the Drone, and its report on drone spending, published today, tracked everything unmanned in the Pentagon budget, as the center has done since 2013. This year, adding up programs across services, the Center found that the Pentagon is asking for $9.39 billion in drone spending. “We’ve seen budget hover in the 4-6 billion annual range since fiscal year 2013,” Gettinger said.“A budget for just over 9 billion for unmanned systems, that’s a pretty substantial increase.”

That $9.39 billion breaks down into $2.6 billion for the Air Force, $3.7 billion for the Navy and the Marines, $1.7 billion for the Army, and almost $1.3 billion across the rest of the Pentagon.

“This year the Navy has just skyrocketed past the Air Force,” said Gettinger. Part of that is a $982 million request for maritime systems, though last year Congress delivered over $200 million less on sea drones than the Navy asked for. “Congress has taken a dim view of unmanned maritime projects in the past and has cut funding for some of those.“

The rise of the underwater drone

Part of the shift in spending from Air Force to Navy is the nature of the drones purchased. When it comes to the iconic drones of the last administration, long-endurance machines like the MQ-9 Reaper and the RQ-4 Global Hawk, the Air Force is simply buying fewer than it used it, and what units it does buy are mostly replacements for aging or lost models. These legacy systems still dominate the budget: maintaining and operating existing machines still costs a pretty penny. In the margins around those big-ticket items, there’s a quiet shift taking place: a new emphasis on cheap, small drones.

Quadcopters show up in every service except the Air Force, as well as Special Operations Command and the generic overall Pentagon budget. There’s $140,000 to buy 100 quadcopters for CTEF, or the Counter-ISIS Train and Equip Fund. The Navy and Marine Corps is asking for 200 InstantEye quadcopters, likely part of a “quads for squads” program. SOCOM wants 50 quadcopters listed as just “quadcopter,” as well as over 500 NANO VTOL drones (which are likely to be quadcopters) and 160 Micro VTOL drones (also almost certainly quadcopters). And the Army’s asking for 1084 drones as part of the Soldier-Borne Sensor program, which wants to put palm-sized quadcopters at the disposal of infantry. Altogether, that’s over 2,000 quadcopters of various models. Excluding the ones requested by SOCOM, which are grouped together with multiple varieties of larger drones, the Pentagon is asking to spend about $30 million on small 1,400 drones, for about the same price as two MQ-9 Reapers.

“There’s not a lot of quadcopter action in prior budgets,” Gettinger said. “Last year was the first year we saw some substantial movement on that front, with some funding for quadcopters for CTEF, but it was nothing compared to what we see in FY 19.”

Besides putting small drones into the sky, the Pentagon is also looking to spend more on tools to shoot drones down. In this space, the Navy is also looking to fund multiple counter-drone laser projects, but somehow real-life lasers aren’t even the biggest ticket item. The Department of Defense is asking for $1.07 billion on counter-drone tools in this budget, up from $453.9 million last year Much of that increase is in the Army’s “Low-slow-small Integrated Defense System” (LIDS), which appear to be some form of guided missile.

“The military is taking a lot of different approaches to solving counter-drone issue,” Gettinger said. “Some of that involves outfitting existing defense systems with [counter-drone] capability and some of it involves thinking about how UAS threat can evolve in the future.”

Speaking of guided missiles, the Center report notes that the Army and Navy are asking for a combined total of 1,618 Switchblade loitering munitions. Related to both cruise missiles and drones, they don’t quite fit in either category. Gettinger noted that technically the Switchblades are included in the budget for missiles, making them closer to Hellfires than Reapers. Complicating it further, the Navy’s submarine-launched Blackwing drone is based on the Switchblade body, though it’s formally billed as an ISR platform, rather than a weapon.

“There’s new funding for Army combat ground vehicle program,” says Gettinger, “It’s not a huge amount of funding but it’s definitely part of the project that we’re going to keep an eye on in the future. They’re going to field test a system, remotely operated M119 armored personnel carrier.”

Specifically, that’s $40 million under the “Robotic Combat Vehicle Experimental Unit Prototypes project,” which wants to see robots accompany human-containing vehicles into battle.

It’s hard to say how much of this request will make it through Congress. Yet as a statement of where the Pentagon wants to go, it’s an invaluable guide. By piecing together the whole of the drone budget, the Center finds a pivotal moment, where drones are smaller and more numerous and better dispersed, and distributed too among boats and in submarines and driving on the ground and carried in backpacks. Drones are a big part of how every service sees war in the future playing out, and from cheap scouts in the field to elaborate lasers built to shoot down hostile drones, the notion that drone war is a separable part of warfighting can largely be laid to rest. Drones are an evergreen part of war, now.

Begun, the drone wars have.

Now, the military has plenty of drones, from Global Hawks on down, but nothing currently in the inventory could match both the capability and the disposability he wanted. So he was going to buy a camera drone kit for less than $1,000, and try to field it in the wargame.

I never made it out to see how the drone faired, but five years later, adding a drone into a Marine squad is no longer just the stuff of daring candidates looking for an edge over their peers in friendly wargames. Last month, Marines in the California desert practiced with quadcopters of their own.

From USNI News:

The InstantEye will hardly be the first quadcopter tested by the military, nor will it be the first small drone. Perhaps the most iconic drone carried by infantry is the RQ-11 Raven, a hand-tossed fixed-wing designed to break apart like LEGOs when it landed. The Ravens had a unit price of $35,000, and were sold with all the relevant equipment to operate a batch of three for $250,000. That’s bargain basement when compared to the price of other aircraft, but still pricey in the world of small drones.

To get scouting capabilities dirt cheap, the Marines and the Army have bought and tested commercial off-the-shelf quadcopters, like DJI Phantoms. These quadcopters typically run only a few hundred dollars, making them as cheap as a functional drone can get. The cost savings wasn’t worth the potential risk of the drones being built compromised, so citing cyber security concerns last summer, the U.S. Army ordered units to stop using drones from China-based manufacturer DJI.

The InstantEyes quadcopters, meanwhile, are built by InstantEye Robotics of Andover, Massachusetts. And while InstantEye’s announcent of a recent sale of 800 quadcopters to the Marine Corps didn’t include unit price for the drones, they were reported as costing around $1,000 in 2014. If the cost is similar, then it means that the Marines might finally have a squad-based aerial scout that matches or exceeds those of rivals in both capability and disposability.

And it means the Marines are now much closer to Commandant Neller’s vision of a drone in every squad.

Specifically, the Launch Forth platform is partnering with the United States Marine Corps to offer “design challenges” to meet logistical needs. From a press release:

The Modular Logistics Vehicle (MLV) project will result in the design of a new modular vehicle system that meets a range of logistical missions. Through concept and technical design, challenge participants will work to design a new vehicle that allows for modular adjustments to improve the Marine experience and their corresponding operations. Use cases could range from personal movement, mechanical repair, construction and administrative support.

The competition is in two stages, with a now-closed concept design phase and a technical design phase open until March 25th. There’s prize money involved: $10,000 to be divvied up among seven design challenge winners, and more than $40,000 for technical prize winners. In addition, the Marine Corps will select a winning design April 23rd and then move on to proof-of-concept and prototyping stages.

If all this sounds very Silicon Valley, that’s because it is.

Launch Forth is San Francisco-based and bills itself as “a SaaS (software as a service) platform for product design powered by a robust community of designers, engineers and solvers.” In essence, it’s a way to crowdsource designs, rather than commissioning them. When people submit their entry in the contest, they do so with the understanding (at least clicked through on the submissions page) that Local Motors gets rights to the winning designs. (Silicon Valley, incidentally, *loves* Barrow’s quote about logistics).

Contest-based rights transfer aside, what is the Marine Corps hoping to get from the Forth? From the contest brief (emphasis theirs):

Similar to the first Design Challenge, the objective of this Challenge is to give Marines operating on base improved functionality and experience while conducting day to day logistical operations. This new vehicle will incorporate modular adjustments to improve the Marine experience and their corresponding operations. New modules could be designed and customized on-base as new needs and use cases become discovered. (NOTE: None of these use cases are in combat. Please do NOT include weapons on the design.)

The vehicle can be no longer than 11 feet 8 inches, no taller than 6 feet, no wider than 5 feet 4 inches. In addition, at least part of the vehicle must be 3D printed. It can weigh no more than 2,000 pounds empty, and 4,000 pounds fully loaded. And it must have a minimum 10 inch clearance. Beyond all that, the design must contain four distinct modules to accommodate tasks like Personnel Transport (room for six occupants and four bags) or Administration (a solar panel power supply, as well as everything needed to run a computer, a printer, and a label maker, as well as other requirements), as well as specified construction and mechanical support modules needs. All the modularity feeds directly into another major requirement: it should be possible for two marines to change out modules within 15 minutes.

If the concept and technical design stages are successful, and the Marine Corps finds the skeleton of a design that can ultimately perform multiple missions in the field, then the exercise will have been a steal: two months of concept design time, $50,000 in total prize money, and a new wealth of designed concepts to tweak and manage in the future.

But it’s important to remember that, despite this being a Local Motors project and a vehicle design competition, the partnership isn’t really about building a vehicle. This is really about the Marine Corp contracting with a software company, to then coordinate and reward crowdsourced designs, to get new concepts ready for possible fielding and delivery by someone else down the road.

All of this matches a slew of recent Pentagon efforts, from sending a single Marine to adapt an open-source drone design to Marine needs, or the entire reason for the Defense Innovation Unit Experimental. If the Pentagon is to adapt to the pace of technological change, and to field new machines with anything like the rapidity needed, it’s likely that someone needs to figure out a new approach, a new way to go from “here is a problem we have” to “here is the technology that meets that need.” To do that, the Pentagon is increasingly turning to the expertise of amateurs, with technology companies in the Bay Area happy to play the middlemen.

Today, the blog of the Army’s Mad Scientist Laboratory (part of U.S. Army Training and Doctrine Command) is showcasing an entry on how AI will shape the future of the Army itself, down to the electro-mechanical organs of individual soldiers. In his short story “Sine Pari,” senior concept developer Howard R. Simkin of the U.S. Army Special Operations Command imagines a typical day recruiting for Special Operations in the middle of the 21st century.

From “Sine Pari”:

“My uncle was in Special Operations during ‘the Big One.’ Next to my dad, he is the coolest person I ever met, so …” He searched for words, “So I decided to come and check it out.”

“Okay.” I began. “This isn’t your uncle’s Special Operations. Since the Big One, we’ve made quite a few” – I caught myself before saying changes – “upgrades.” I paused, “Roberto, before we take the enhanced reality tour, I’d like to know what augments you have had – if any.”

“Sure.” Roberto paused for a moment,“ Let’s see…I’ve got Daystrom Model 40B ER corneal implants, a Neuralink BCI jack, and a Samsung cognitive enhancement implant. That’s about it.”

The essence of “Sine Pari,” is one of hackable bodies and hackable systems. The young recruit has already augmented his body, and has childhood experience hacking the robots and household AIs he grew up with. We don’t know the kinds of wars prospective recruit Roberto Preciado will fight, or even the nature of The Big One fought before this time. Instead, we get a portrait of a highly advanced scanning system, helpful AIs in the place of aides, and a social comfort with biomechanical body enhancements.

What is missing from this portrait, yet implied by this world, is the massive amount of cybersecurity back-end needed to secure not just the devices a person uses, but the devices people put into their bodies. To take this to one logical end, we need only remember how, two weeks ago, a popular jogging app made headlines because of the paths it revealed around military bases. Jogging apps that log the GPS position of a user, and then upload it later, create maps of some value, and are a good way to think about the vulnerabilities inherent in aggregate data.

In “Sine Pari,” we get a recruit with three augmentations: electronic eyes, a brain input link, and a brain enhancement. Notably, these are also all assumed to be from major brands, ones where it’s entirely possible the security flaws can be studied and exploited for profit. Now, it’s possible that after recruiting, our new Special Operator swaps out commercial parts for models built to military specifications, which would be a tidy way to close that potential plot hole. It’s also possible, following the logic of this world, that when Roberto Preciado joins the force, his cognitive enhancement contains malicious code, his eyes record everything he sees, and his brain-link transmits that information to an interested third party, who suddenly has eyes on the inside a Special Operations unit.

When we think of the potential the future brings, we have to think of the risk as well. Electronic eyeballs are one compromise eye from picture-perfect spy cameras. Brain uplinks and cognitive enhancements may greatly reduce training times and allow the rapid learning of complex skills, but they could also be a pathway to compromise or hijack the brain of the person in question. As the tools of war change, the adversarial nature of war, the danger of technologies exploited against their users, remain.

Fortunately, both for the fictional protagonist here and for TRADOC’s Soldier 2050 concept more broadly, this is not a stand-alone project. On March 8 and 9, TRADOC is co-sponsoring a Bio Convergence and Soldier 2050 Conference in Menlo Park, California.

With IEDs set on the side of the road, or attacks springing suddenly from concealed attackers in ambush, being first is a risky proposition. Which means it’s an ideal job for a robot, immune from such fleshy concerns as panic or mortality.

Last month, U.S. Army Tank Automotive Research, Development and Engineering Center (TARDEC) and the U.S. Army Armaments Research, Development and Engineering Center (ARDEC) collaborated to see if a robot Humvee could do the same course as human crews.

That course was Scout Gunnery Table VI, which the Army uses to certify crews for combat vehicles. The course tests the vehicle’s ability to move and shoot and communicate, engage stationary and moving targets. Those targets range from infantry cut-outs to armored vehicle silhouettes. For testing a robotic Humvee, the Army will send it alongside a human-driven command-and-control vehicle, as a sort of wingman team. The robot Humvee will operate without people inside, using three different autonomous systems:

First is the TARDEC-developed Robotic Technology Kernel (RTK), the autonomy system for planning and controlling the vehicle’s mobility. RTK contains driving cameras for remote operation, LIDAR (light detection and ranging) sensors for object classification, stereo cameras for terrain classification, computers for computation, radios for communication, and all the essential hardware, cables and mounts. The system can be manually driven through teleoperation or autonomously driven through waypoint navigation.

In addition to the autonomy system, the robotic vehicle has a Picatinny Lightweight Remote Weapon System that can mount either a minigun or a machine gun. That means that this is an armed robot, though one with a human-in-the-loop, as we’ll get to in just a moment. The robot Humvee, with its autonomy system and remote weapon system, also uses the Autonomous Remote Engagement System (ARES), which, according to the Army can, uses a vision-based system to both track targets on its own and follow targets designated by a human user.

So, this is a Humvee that can be remotely controlled or navigate autonomously to waypoints, that mounts a weapon, and that can watch and track targets. Crucially, it takes a human to press the trigger. In this case, that human is one of the five people in the wingman command-and-control vehicle. There’s a commander, a driver and then, in the back, there’s a gunner and a remote driver (piloting the robot) and a remote weapons operator (controlling the robot’s gun). The hatch gunner can also use a designator to mark targets for the robot.

The autonomy we see here is a sort of limited, directed autonomy: the robot can go where it’s sent, and point a gun at a target it’s told to follow, but this is not a machine operating entirely independently, given a mission it then executes on its own. Instead, it is a robot tag-along, a vehicle crewed remotely.

As robotics move more and more into the battlefield and become a standard part of war, it will look a lot like this: an aging military machine, retrofitted into a robot, taking point on patrol and firing back while a human, inside a follow-on vehicle, tells the robot what to shoot.

When it comes to Incirlik, the Strava jogging data just adds to existing knowledge. Satellite views show B-52s stationed outside special bunkers, a quick search of “nuclear weapons Turkey” pulls up a ton of stories about the storage at the base in Incirlik (I’m partial to this one), and the knowledge that people on base like to run laps around available paths probably isn’t new information to anyone in the area. As analysts and uniformed personnel debated on Twitter, what Strava’s heat map adds isn’t much compared to what a half-interested observer probably already knows. But that’s the heatmap itself.

The bigger, specific question is what else Strava knows that isn’t on the heat map. And more broadly, the bigger danger is what happens when every technologies vital for everyday life record that information and share it widely. Strava accounts are linked to Facebook, Google, or email, and depending on the sign-up method, by simply making the account a user gives Strava the same data about their connections already siloed away in a social network. A click or two later, and the user can choose how to share their location with the app. Buried under setting is a “privacy” section. By default, anyone can view a Strava user’s profile, people logged into the app can follow a user, and can download user data. An Enhanced Privacy setting masks some activity, but users have to individually toggle each of several categories of activity to hide it from the app. If a user wants to keep a certain location hidden on the app, like their home or office, they have to go through the desktop client to set a hidden zone with a radius of three-quarters of a mile around a set location.

For someone who wants to just sign on, run, and share their run with their buddies, that’s a lot of work to make sure that, say, the daily jog around a patriot missile installation is secure. And someone who wants to make sure that the steps they take while on on duty count for their fitness tracker might not take the extra time to play with the sensor and hide their location data from other users. And even if they do, the information is still fed into the app and the company itself, where it is then collected and up to the company how that data gets used. The heat map doesn’t identify any specific users. It instead gives viewers something else: patterns of behavior around places.

A quick note on this: while the high traffic areas show brightest (this is the nature of a heat map, after all), it’s the thin lines around sensitive locations, detailing paths walked by maybe one person doing security, that are probably most interesting to anyone plotting something nefarious. That’s true universally: the paths of any group of armed people running security with Strava on can be found here. While it might be of special interest to those tracking the activities of the Pentagon, it’s relevant to anyone tracking any military. And a caveat: this is just the data of people who use the Strava app, so while it documents some activity, the absence of anything on the Strava map only means no Strava users recorded data there, not that nothing happened there.

Strava is hardly the first app to capture ephemeral behavior and turn it into a public, geolocated document. When the messaging and video tool Snapchat introduced a world map option, allowing any user to view any users or posts set to public, commanders had to remind those serving to change their privacy settings, to keep locations private from acquaintances. When Pokemon took a map built for the augmented reality game Ingress and made it a massive phenomenon, quiet but mildly notable sites (like, say, a Commanding General’s house) became the target of casual interlopers.

We are living in a future where the default is to be connected, and the ways people are connected increasingly ask users to offer information, seemingly insignificant, in exchange for daily utility. To some extent, this is vital: what good is a jogging app if it can’t tell where the user went jogging? What is unclear, though, is what else the app knows, and how securely that data is stored, and if it’s possible for someone with access to the full dataset to de-anonymize it.

And then, there is the simple fact of the data itself: while the location of nuclear storage in Incirlik may be public, and the bunkers may be visible from space, those loops are still restricted areas, and cell phones, especially the kind that can record the path someone took on a jog, are restricted in those areas. If there is a risk here, it is the modern liabilities information shared online combined same danger that has plagued every endeavor from the dawn of time: human error.

But in the long run, it’s still a good idea to understand how such a group can harness the resources it commands, and turn them into the fuel for a grotesque war machine. To that end, on Halloween the World Bank published a study of ISIS’s oil production, based on what can be inferred from remote sensors.

Before we get to the findings, it’s interesting to the methods for tracking that production. From the report:

Notably, the VIIRS sensors captured gas flares from active wells. The researchers than compared production data from prior-ISIS records. Helping the process along, in 2015, the capture of an ISIS official responsible for logistics yielded a trove of papers and internal accounting, against which researchers could check their production estimates. (ISIS, it turns out, very much tried to standardize production within its borders, complete with quality control inspectors for everything from explosives to consumer goods.)

The study confirms what most informed observers of the war against ISIS could expect: production peaked in the six months after the fall of Mosul, and then declined a lot in 2015, and declined further still in 2016. The numbers themselves, too, are small: at its brief July 2014 peak, ISIS produced only 80,000 barrels per day, and throughout the entire run of the Islamic State, it produced only about as much oil as Abu Dhabi makes in 10 days.