59. Exclusive Interview with Antigravity: What It Took to Cross DJI’s Moat

Summary

• Antigravity’s head of drone R&D and product, 纽维, spoke to the media for the first time, arguing that A1 is not competing head-on within established categories such as traditional camera drones or FPV drones, but defining a new category: the “panoramic drone.” The product starts with specific functions to win—immersive flight and the freedom to “shoot first, edit later”—and targets users intimidated by learning to fly: “At the product-definition and commercial-logic level, we are fundamentally operating in a lane separate from the traditional market.”
• On Insta360’s post-launch stock drop and rumors that sales had “cooled,” 纽维’s response was: “We don’t feel much about it—let the bullets fly for a while.” He confirmed that the RMB30M in sales reported by JK over the first 48 hours excluded overseas channels that had not yet launched, and said the figure was “definitely in line with expectations.” Products priced in the high-thousands of yuan have a long decision cycle: customers need to experience them in stores before gradually coming around, and the team never expected to win over everyone immediately after launch.
• The market sees getting around DJI’s more than 100 patents as the hardest part of building a drone from scratch; 纽维’s view is that “this is a fact, but it is not quite as unsolvable as it looks.” “It doesn’t mean there is only one route from A to B”—for example, A1’s rear arms fold horizontally rather than rotating backward as in the patented Mavic design. DJI itself built its business by working around other companies’ patents, he noted; “later entrants may simply have a somewhat harder time.”
• His technical self-assessment was unusually candid: flight control is “on par” with the rival, several other areas are “slightly behind, but not by much,” and panoramic image quality objectively still trails that of traditional drones. The attack plan is to shore up the battery first, then tackle point-to-point video transmission—a narrow market where talent and suppliers are concentrated around the leading rival. 纽维 said the team had “spent an enormous amount of effort” solving it; latency of a few hundred milliseconds is enough to make users dizzy.
• The Antigravity team encountered non-compete restrictions at 10 to 20 suppliers, and one core sole-source supplier abruptly cut off supply after the tooling parts had returned, leaving “five seconds of absolute silence in the office.” The team was forced to replace the supplier and redo the work from scratch. 纽维 said the experience was made public because the team wants the market to be “more open and fair”; it now routinely prepares second-source options and contingency plans for at-risk materials.
• The endgame call—the “south-slope, north-slope” framework—is worth keeping in view: once drones achieve L4 autonomous driving, natural-language interaction and automatic camera movement, panoramic capture “may indeed no longer be necessary.” But airborne compute is unlikely to advance that quickly, and camera movement is fundamentally subjective creative work, so the 2 forms should coexist for a considerable period. Insta360 is “of course not just climbing the south slope”; it is also studying technology stacks for products it has not yet built and does not rule out any form of drone.
• The central conclusion on moats is that hardware has no moat, a single product innovation has no moat, and “a sustained organizational atmosphere and soil for innovation is the moat.” 纽维 acknowledged that there is “nothing particularly difficult to copy” about A1; the long-term edge is continuing to find unmet needs across the full shooting-recording-sharing chain. The RMB6,799 price sacrifices short-term profit as a good-faith offer—matching the experience at normal industry gross margins would put it at RMB8,000-9,000, but “price wars are a privilege of giants; we don’t have that privilege.”

Deep dive

1. A Gaming World Champion and 陶冶 of Bambu Lab: Why 纽维 Joined the Drone Industry

• 纽维 volunteered an “embarrassing” chapter from his past: he was consumed by gaming in college, won a world championship, and spent several extra years in school. That was how he met a gaming partner who stayed up all night to play with him—Dr. 陶冶, now the founder of Bambu Lab. After returning to China, 陶冶 spent time at DJI and in 2015 gave 纽维 the chance to attend a workshop summer camp in Shenzhen.
• The camp changed the direction of his life: “There were many smart, idealistic people who were highly grounded and willing to express what they really thought… This truth-seeking atmosphere was something I had never encountered in Chinese companies before.” In 2016, he gave up pursuing a PhD, returned to China and joined another drone company; the interview did not identify the company.

2. Eight Years of Industry Observation: Mavic Pro Was a Revolution

• In 2016, drones offered roughly 20 minutes of flight time and were not portable. The flagship product was Phantom 4, which had just integrated machine-vision obstacle avoidance. Then, in the second half of the year, the roughly 500g Mavic Pro arrived: “It was a revolution. No one had imagined that a drone could be made so small.” His former employer’s revenue consequently jumped from several billion yuan to around RMB20B.
• After years of accumulation, every submodule had made substantial progress. Today’s drone market spans a product matrix from around RMB3,000 to above RMB10,000, with near-RMB20,000 models at the prosumer level.

3. Why He Left His Former Employer: Self-Doubt and JK’s Candor Over a Late-Night Dinner

• 纽维’s question to himself was whether strong product sales were really attributable to his own product definition, solution selection and quality management—or to the platform’s ready-made technology, industrial capabilities and brand. “After going through 2 generations of products, I couldn’t get an answer. I wanted to give myself a bigger challenge.”
• He had a late-night dinner with JK at the end of 2023, and alignment on productization, future technology and the competitive landscape was one reason he joined. What moved him most was JK’s willingness to lay out Insta360’s existing problems, the challenges ahead and “some things he was afraid of”: “A founder who is willing to face his own difficulties, who has a sense of urgency and the resolve to make adjustments—that was what really moved me.”

4. Why the Door to Drones Is So Hard to Pry Open: Too Big for Big Companies, Too Hard for Small Ones

• 纽维’s explanation of the market consensus is that consumer drones require exceptionally high R&D, mass-production and quality-control capabilities, while the market is only worth a few tens of billions of yuan. The economics are unattractive for any company: “It may be a circle that big companies don’t care about and small companies can’t operate in.” Asked whether the door was easy to pry open, he answered directly: “Of course it isn’t.”
• His former employer went deep in every submodule, building a formidable moat on supply-chain capabilities and talent density: “In the narrow field of consumer drones, the most critical and strongest suppliers and talent are actually concentrated there.”

5. Why Insta360 Has a Shot: Reusing Flying-Camera Strengths to Define a New Category

• The entry thesis has 3 layers. First, Insta360 now has enough capital: “This is no longer an era when a team of a few dozen people can build a consumer drone well.” Second, a consumer drone is fundamentally a flying camera, making Insta360’s imaging technology readily extensible. Third, the leaders of the core functions have years of industry experience, and the team is “grounded and able to separate truth from noise.”
• Commercially, the company entered through a position distinct from the traditional market: addressing the high cost of learning to fly and the difficulty of camera movement, while elevating immersive flight—a capability traditional drones do not offer—to the center of the product. “We want to open up a new category, the panoramic drone, so we think we have a chance.”
• The team’s average age is no more than 30, while its functional leaders are around 35. JK is more of a supervisor: “A lot of the time, we’re pretty disobedient.” The team will argue with JK and reject his requests, partly because JK himself has never built a drone.

6. Stock Down, Sales “Cooling”? “Let the Bullets Fly for a While”

• On Insta360’s stock drop after the launch event, 纽维 said the team “honestly didn’t feel much.” JK did not pass the pressure down either. Internally, the team cut a cake, did an awkward dance and handed out red envelopes to celebrate: “We’d suffered for so long, working day and night. The aircraft was finally released and shipped.”
• He directly addressed rumors of “terrible sales.” The RMB30M reported over the first 48 hours excluded overseas channels that had not yet been rolled out and was “definitely a sales level in line with expectations.” His view of the purchase funnel is that a product costing several thousand yuan requires customers to try it in a store before gradually accepting it. For users unable to make sense of a flood of positive and negative reviews, he recommends trying the product offline.
• The project may have spent 2 or 3 years in its early phase. After 纽维 joined, he completely rebuilt the system and spent roughly 2 more years on it. The team had also built multiple generations of demos earlier that never became actual products. On JK’s description of the drone as a “10-year decision” and the project’s having started around 5 years ago, 纽维 said: “That’s roughly the story as I understand it.”

7. A1 Dissected: A Panoramic Bubble on 249g, Stabilization Without a Gimbal

• The core innovation is putting an up-and-down dual-fisheye panoramic camera on a 249g aircraft, together with a remote controller, flight stick and goggles. 纽维’s “bubble” analogy: the aircraft records a spherical video bubble in the sky, while the person wearing the goggles appears to stand inside it and sees whichever direction they turn their head toward.
• The 2 differentiated experiences are genuine immersion and simple flight and capture. The cinematic view of an FPV drone “requires a highly skilled, highly experienced pilot”; let go of the controls and “you may crash many aircraft before you can shoot a smooth clip.” A1 can fly the desired route, after which users frame, crop and rotate the spherical footage frame by frame to recreate an FPV effect.
• Why no gimbal? Changes in aircraft attitude create large vibrations, traditionally offset by a physical gimbal. But “however you rotate a spherical object, it is still a sphere; only the portion you crop out rotates,” so the function can be handled electronically instead of with a physical gimbal.

8. Circumventing DJI’s 100-Plus Patents: There Is More Than One Route from A to B

• On the investor consensus that the hardest part was getting around DJI’s more than 100 patents, 纽维 said: “That is a fact, but it isn’t quite as unsolvable as it looks… It doesn’t mean there is only one route from A to B, and the supposedly best route often works only under certain conditions.”
• The folding structure is one example. A1’s rear arms fold horizontally, while the Mavic’s rear arms rotate backward, a design covered by patent filings that was “a very brilliant design at the time.” He added the historical perspective that DJI itself had taken many different routes to work around other companies’ patents: “The circumstances are broadly similar for every company. Later entrants may simply have a harder time.”

9. Category Life Cycles: Why a Third Phase Has Not Arrived in More Than a Decade

• 纽维’s framework has several stages: a technological breakthrough creates an expensive prototype; one company brings cost, reliability and usability down to an acceptable baseline, defining the category; the market expands and attracts more entrants; then the marginal impact of technology declines and functions consolidate into a single product—smartphones swallowing point-and-shoot cameras is the classic example. Insta360 wants to play the role of “creating a small category” within consumer drones.
• Why has DJI spent more than 10 years in the second stage without producing more specialized competitors? Market growth has slowed and the total market is only worth a few tens of billions of yuan. “There is no such thing as a small matter when you are sending something into the sky”—an old saying in the model community, where safety and reliability come first. That alone disqualifies many companies. The leading rival has also worked with suppliers to make every submodule “extremely sophisticated,” leaving entrants with high barriers and uncertain returns: “Even if the gross margin looks very high, this is not a market anyone can enter casually.”

10. The Safety Trade-Off Behind Not Allowing Inverted Flight

• JK keeps a GoPro Karma on his desk; after launch, the drone was involved in crash incidents and recalled. 纽维 experienced that episode firsthand at his former employer and “understands deeply the importance of system safety in drones.” The approach includes designing out lessons previous players paid to learn, testing in high winds, at high altitude, in extreme cold, rain and fog, and “labeling cautiously”: testing may show operation down to -20°C or lower, but the official rating is limited to -10°C.
• The restriction on inverted flight is deliberate, not a technical shortcoming. There is no rear-facing perception camera; the 2 forward-facing perception cameras are used for obstacle avoidance, which he likens to AEB in autonomous driving, and flying backward creates a crash risk. Traditional drones fly backward to capture a pull-away move from near to far, but “for a panoramic camera, the content in front and behind is the same; flying backward is simply a question of which part of the bubble you crop.” 纽维 called it a trade-off. As the host summarized, the product cannot simultaneously offer unrestricted forward and backward flight, safety and more efficient framing.

11. Tech Road Map and Unusually Candid Self-Assessment: Video Transmission Is the Hardest Nut to Crack

• The 4 supposedly major challenges—flight control, perception and obstacle avoidance, video transmission and battery—are, in 纽维’s view, only a partial list. These are narrow fields with very few qualified people and suppliers. The order of attack was to fix the battery first, then identify point-to-point video transmission as a major problem—“we spent an enormous amount of effort” solving it—followed by flight control and multisensor-fusion algorithms.
• Video transmission is difficult because its consumer use cases are narrow, mainly drones, and few strong companies are willing to develop it as a standalone capability. The leading rival raised the bar with custom chips, a proprietary protocol stack, algorithms and communications algorithms, lifting user expectations for resolution, frame rate, interference resistance, stability and latency. “If it is a few hundred milliseconds, people get dizzy and the controls feel sluggish.” That has become a major moat.
• Asked for an objective assessment, 纽维 offered an answer worth recording: “The current level of our flight control can stand shoulder to shoulder with the other side. As for the other areas, we may be slightly behind, but not by much.” When the host asked whether that might discourage buyers, he replied: “We come from a technical background. Objectivity matters a lot, and we don’t want to mislead users.” The product definition is different, however, and the resulting experience is “one of a kind.”

12. Supply Cutoff and Non-Compete Restrictions: Why Go Public

• Just after this year’s Spring Festival, with the tooling parts back, a core sole-source supplier said it could no longer supply materials because of a non-compete restriction. The original exclusivity agreement had applied only to one product; “when we were told supply would be cut off, that agreement should have been expanded.” “We were stunned in that moment. The whole office went silent for 5 seconds.” 纽维 compared it to making an entire table of food only for a pet to jump up and knock everything onto the floor. The solution was to start over, add manpower and replace the supplier.
• The unfair treatment mentioned in JK’s letter to all employees was real. Non-compete restrictions affected 10 to 20 suppliers, and the team now routinely prepares second-source options and contingency plans for at-risk materials, keeping inventory at relatively safe levels. Asked why the experience was made public, 纽维 said the team had been through many difficulties and wanted the market to be “more open and fair,” which would support market growth, help the supply chain build stronger capabilities and allow suppliers to innovate further upstream.

13. The External Display Debate, Virtual Cockpit and AI Editing: Won Over by the Counterintuitive

• Secondary-market analysts argued that the external display on the flight goggles was feature bloat and could hurt the cost structure. 纽维 initially agreed and strongly opposed it: “It takes up space, adds cost and requires effort to tune. I thought it was a waste.” JK responded: “We need to do some human-centered, counterintuitive things.” 纽维 was not convinced at first, but as more testers said they liked the display, he realized “there was another layer to it.” After a future OTA update, the display will support customizable text and memes.
• He initially considered the virtual cockpit—turning the aircraft into a mount ridden by a flying creature—equally impractical, but changed his mind after seeing a demo: “Everyone may have seen How to Train Your Dragon and wanted to become the protagonist, riding a mount through the sky.” The game-like mechanics are still being explored and development has begun, but the team will first focus on polishing the flight experience.
• AI editing is not yet capable of fully automated input-to-output production. Once users select keyframes from the footage and arrange the beat points, it can generate a short video of acceptable quality. 纽维 “wouldn’t dare say it is done very well,” but believes it should be best-in-class for the 360-degree panoramic-camera and drone use case. Insta360’s roughly 200-person team is tracking developments in VLM and VLA; AI is already used for stitching and chromatic-aberration correction, while automatic editing and end-to-end perception and obstacle-avoidance models remain under exploration. On the criticism that customers have “paid RMB7,000 for a toy,” he acknowledged that most skepticism concerns image quality, while the App and post-production workflow will be optimized. Ordinary users should have no problem generating a decent clip, though the number of clips usable by a studio “may not be particularly large.”

14. South Slope, North Slope and the Endgame: Opportunity or False Need? How Was RMB6,799 Set?

• The host’s endgame question was whether panoramic capture on drones might ultimately prove to be a false need. 纽维’s framework: the panoramic drone’s value is immersive flight and simple capture, while its weakness is image quality relative to traditional drones. The key question is whether simplifying flight is easier for traditional drones to solve, or whether closing the image-quality gap is harder. His view is that autonomous driving and automatic camera movement in the air depend on perception systems and compute, and “the progress of compute is predictably not going to be that fast.” Camera movement is also subjective creative work: “If you hand everything to AI, you lose a lot of the fun.” Under the absolute conditions of L4 autonomous driving, natural-language interaction, automatic route planning and freedom to shoot, “perhaps panoramic capture really would no longer be necessary.”
• The north-slope question is whether panoramic capture can close the image-quality gap, following the camera market’s pattern: once smartphone image quality became good enough, point-and-shoot cameras and near-professional all-in-one cameras disappeared. But Insta360 is “of course not just climbing the south slope.” “Even technology stacks for which we do not yet have a product, we are studying and discussing in depth. We do not reject any form of drone.”
• The pricing logic is straightforward: “Price wars are a privilege of giants; we don’t have that privilege.” RMB6,799 is deliberately positioned below RMB7,000, with short-term profit sacrificed as a good-faith offer. If the immersive experience were priced at normal industry gross margins, “it would probably come in at RMB8,000 or RMB9,000,” while no product with a comparable experience is currently available in the short-term market.

15. “The Innovation Soil Is the Moat”: JK’s Tragic Resolve, Blowups and Cash Tossing

• Asked what about A1 would be difficult to copy, 纽维 answered: “From a technical perspective, there is nothing particularly difficult to copy.” The long-term edge is the willingness to keep finding unmet needs across the chain from shooting to recording to sharing and to keep experimenting. The host summarized, and he agreed: “Hardware today has no moat, and a single product innovation has no moat. A sustained organizational atmosphere and soil for innovation is the moat.” “If you keep innovating, others can only remain followers.” Insta360’s roughly 200-person team is tracking VLM and VLA developments; AI is already used for stitching and chromatic-aberration correction, while automatic editing and end-to-end perception and obstacle-avoidance models remain under exploration.
• On JK’s social-media post saying that Insta360 might not lead the drone industry in the future, might even disappear from it, but could still leave its name in the history of technology, 纽维’s interpretation was “accepting everything that could happen, including facing the worst outcome—there is no other meaning to it.” He added: “We don’t think we will disappear, because we are very clear about the path ahead.” A1, a from-zero-to-one product, was the hardest product he had ever worked on. One colleague who lived through the dark period of the early demos put it this way: “It was really too hard. Getting this done has made our confidence explode.”
• One decisive moment came when 纽维 joined and test-flew an old demo: it turned doubt into conviction because “shoot first, edit later” dramatically improved the odds of getting a usable shot from a drone. JK appears to have lost his temper only once. 纽维 asked him to take a validation sample to a core supplier and negotiate a lower price, but the prototype failed to fly because of GPS-signal issues and unfinished hardware. “He felt it was quite embarrassing and really laid into us twice.” Asked afterward whether he apologized, 纽维 said: “No—I don’t think so.” When the host pointed out that it was a product problem, he agreed. At an August brand-promotion party, JK spontaneously threw cash, sending the event onto the trending lists; the team found it “a little awkward, but also kind of fun.” “We don’t dare throw cash this time.” JK later said on social media that he would send red envelopes instead. Insta360 CEO Leo closed with a line from a poem: “People do not recognize the soaring tree; only when it reaches the clouds do they speak of its height.”