🇨🇳❤️ West’s Reserve Weaponization Opens Door for China’s Gold Hub

The West turned foreign reserves into a weapon in 2022. Now European central banks are moving gold out of North America, and China is building a place where countries can store and trade bullion beyond the reach of the US-led financial system.

The Netherlands has moved 86 tonnes out of New York and Ottawa, cutting the share of its reserves held in New York from 31.3% to 18.5%. France sold 129 tonnes stored there and bought the same amount back in Europe. Both moves followed the freezing of Russian central-bank assets, which showed that reserves held abroad remain exposed to the politics of the country controlling access to them.

This is changing how central banks manage gold. In the World Gold Council’s 2026 survey, 10% of respondents said they had diversified foreign storage locations over the previous year, compared with just 2% in 2025. Nearly three quarters expect the dollar’s share of global reserves to fall over the next five years.

China is preparing to catch part of that shift. Shanghai is developing a national-scale precious-metals depository designed to attract foreign central banks. Hong Kong plans to expand vault capacity to more than 2,000 tonnes and has launched a central gold-clearing system connecting Chinese and international banks.

Shanghai and Hong Kong are being built to work together. Shanghai has the mainland gold market and exchange. Hong Kong gives foreign institutions an offshore entry point, plus clearing and settlement. The next step is tighter links between the two markets and more yuan-denominated contracts.
Put together, the system would let a central bank keep bullion, sell or pledge it and settle the deal through Chinese institutions. Washington would no longer sit anywhere along the chain.

Foreign gold would bring more trades into Chinese markets, more weight to Chinese pricing and another reason to settle in yuan. For reserve managers worried about the next round of sanctions, the selling point is obvious: metal held and traded through China is much harder for the United States to immobilize.

@NewRulesGeo❗Follow us on X

🇨🇳☄️ Taiwan Separatists In Panic: China Gives Taiwan-Facing Brigade Its Own PLZ-05 Long-Range Howitzers

China has pushed one of its most powerful artillery systems down from group-army formations to a heavy combined-arms brigade in the Eastern Theater Command.

A brigade under the People’s Liberation Army’s (PLA) 71st Group Army received a batch of PLZ-05 155mm tracked self-propelled howitzers over the past two months, replacing older 122mm guns, and has already completed its first live-fire test.

It is the first publicly confirmed case of PLZ-05s being assigned directly at brigade level. Instead of waiting for higher command to attach a separate long-range artillery formation, the brigade now controls its own heavy firepower.

🔸 Target coordinates from reconnaissance drones, radars and forward observers can be sent directly to the brigade’s guns, shortening the time between detecting a target and opening fire.

🔸 The PLZ-05’s automatic loader supports a maximum rate of around 8–10 rounds per minute. With BeiDou-guided WS-35 projectiles, it can strike targets at distances of up to 100 km, dramatically expanding the brigade’s reach compared with its older 122mm artillery.

🔸 Its tracked chassis allows the howitzers to move alongside tanks and infantry fighting vehicles across difficult terrain while providing greater crew protection than lighter truck-mounted guns.

🔸 The brown desert camouflage on the newly delivered batch suggests that the howitzers may have been transferred from the Western Theater Command, where lighter PCL-181 truck-mounted systems are better suited to long road movements and mountainous terrain.

🔸 The 71st Group Army belongs to the Eastern Theater Command, responsible for operations around the Taiwan Strait. In a conflict, its PLZ-05s could provide sustained fire against coastal defenses, artillery positions, command posts and supply hubs once these targets are within range.

China is also testing an experimental 155mm naval gun. If it becomes compatible with existing army ammunition, the PLA Ground Force and Navy could eventually draw on a broader common stock of modern 155mm shells.

China is giving frontline brigades firepower previously controlled at a higher level, reducing response times and allowing them to fight more independently.

Could Taiwan’s defenses withstand sustained fire once PLA brigades bring their own PLZ-05s into range?

@NewRulesGeo❗Follow us on X

🇺🇸⚓️📉 America Running Out of Ships for Pacific War

Every patrol, convoy escort or submarine hunt assigned to an Arleigh Burke is one less US destroyer available for carrier defense and high-end combat. China can hand much of that work to a frigate force that already numbers around 50 ships. The US Navy lacks that middle layer—and that is why it is suddenly shopping abroad.

The Navy has operated without a frigate since the last Oliver Hazard Perry-class ship retired in 2015. Littoral Combat Ships were supposed to take over the lighter missions, but breakdowns, unfinished mission packages and early retirements left a hole in the fleet.

Constellation was meant to fill it. The Navy took the proven Italian FREMM design, changed it almost beyond recognition and started construction before the drawings were finished. The Government Accountability Office found that design instability had stalled work and delayed the first ship by at least three years. The planned 20-ship class was cut to two.

The US Navy is now examining frigates from Japan, South Korea and Turkey. Up to two could be built abroad before production moves to American yards.

A three-to-four-year transfer would give China time to add another 6–16 frigates. The older Type 054As already handle escort, patrol and submarine hunting, while the larger Type 054B is moving into serial production.

That lets China keep its expanding Type 052D and Type 055 destroyer force for missions that need heavier missile batteries, stronger sensors and area air defense. America has no comparable middle layer, so Arleigh Burkes end up doing both jobs. Their crews, maintenance schedules and deployment time are not unlimited.

Even the stopgap is stuck between Congress and the shipyards. Lawmakers want to block foreign-built warships to protect domestic industry, while the White House admits that six major programs are already backlogged. After the first foreign hulls, production would return to the same shortage of workers, yard space and suppliers.

Ship for ship, an Arleigh Burke outguns a Type 054A. Fleet against fleet, China can send frigates to do frigate work and save its largest destroyers for the dangerous part. America keeps using its best ships because it has too little underneath them.

@NewRulesGeo❗Follow us on X

📣 @GeoSight 🔥 shares updates and analysis on global conflicts and international relations. It covers breaking news and insights into military and Geopolitical events.

🚨 Defence & Security
🌎 Geopolitics
⚡️ Wars & Conflicts
📊 Economic Trends

📎 Join 🚩:⤵️
https://t.me/GeoSight
https://t.me/GeoSight
https://t.me/GeoSight

🇮🇷⚔️🇺🇸 Why America Can’t Strangle Iran

Washington has launched yet another campaign to sever Iran’s “economic lifelines.” Obama promised “crippling sanctions,” Pompeo tried maximum pressure, and the current administration has already moved from Operation Economic Fury to an “economic D-Day.” The policy keeps changing names because the promised surrender never comes.

Sanctions make Iranian trade slower and more expensive. Isolating the country is another matter. Decades of pressure taught Iran to work without normal access to Western banks, insurers and shipping. It substitutes imports, spreads power generation across the country and sells oil through a large shadow fleet. After six months of war and blockade, even the IMF revised its 2026 forecast because Iranian oil exports performed better than expected.

China is the biggest hole in the siege. It bought more than 80% of Iran’s shipped oil in 2025, averaging 1.4M barrels per day. When the Treasury unveiled its first 60 targets, it left Chinese financial institutions off the list. Scott Bessent explained why: “Why would I want to blow up the global financial system?”

Trade is also moving outside the channels Washington controls. Reuters found a barter-like system in which Chinese buyers pay for Iranian oil with credits used to purchase Chinese goods and fund projects inside Iran. An estimated $2–2.5B passed through it in the last year. No dollar transfer between an Iranian buyer and a Chinese exporter is needed.

Iran also has geography on its side. It borders seven countries and faces both the Gulf and the Caspian. American warships can interfere with tankers near Hormuz; they cannot patrol Iranian railways, close every land crossing or enter the Caspian Sea. Russia and Iran have already expanded that northern supply line, linking Iranian ports to the International North–South Transport Corridor.

A genuine economic cordon would require Washington to coerce China and Russia as well as trading partners and transit states from Turkey to Pakistan. If it targets the large banks and companies connecting those economies, the damage spreads far beyond Iran. If it holds back, the routes remain open.

America can raise the price of almost every Iranian transaction. It cannot remove Iran from the map between East Asia, Russia, Central Asia, South Asia and the Gulf. Fourteen years after the first promise of “crippling sanctions,” Washington is still renaming the same policy and waiting for the same surrender.

@NewRulesGeo❗Follow us on X

🇺🇸👨🏼‍🚀📉 America’s Space War Plan Has Workforce Problem

The US military goes into any major war assuming its satellites will keep working. They warn of missile launches, locate ships, guide weapons, relay orders and connect forces spread across an ocean. If that layer is badly damaged, replacing it becomes part of the war.

A new RAND study, commissioned by the Vice Chief of Space Operations, examined whether American industry could rebuild critical space capabilities within 12–24 months during a prolonged conflict between 2027 and 2032. It found serious shortages in all three groups needed for the job: engineers, hands-on technicians and highly specialized experts.

Only 3% of US engineers work in industries supporting the space defense sector. Security clearances for new staff take six months or more. Some expertise in electro-optical sensors and radiation hardening can require at least 15 years to develop. None of that can be fixed with an emergency appropriation after the shooting starts.

The production system creates another problem. Many military spacecraft are built in small numbers rather than on continuous assembly lines. When orders pause, experienced workers leave, specialist skills fade and suppliers move on. RAND found that contractors have also done little with automation because their normal production runs are too small to justify it.

That makes a wartime surge much harder than simply ordering more hardware. Factories would have to find cleared engineers, retrain technicians and restart dormant processes while the conflict was already under way.

The timing is brutal. A war in the Pacific could turn on targeting, communications and navigation during its opening days. An adversary would not need to erase every American satellite. Disrupting the specific services used for missile warning, surveillance and command could leave ships and aircraft working with delayed data when it matters most. A replacement launched 12 months later cannot recover information that was missing during the battle.

China and Russia are meanwhile building space infrastructure designed to recover from losses: distributed constellations, sovereign communications networks and more frequent launch capacity. China’s work on reusable rockets could make replacement launches faster still.

Congress cannot produce a cleared cleanroom technician or radiation-hardening expert overnight. America made orbit essential to every part of its military, but never built the workforce needed to replace heavy losses on a wartime schedule.

@NewRulesGeo❗Follow us on X

🇮🇷⚔️🇸🇦 Houthi Breakthrough Opens the Road to Bab el-Mandeb

The rapid collapse of Saudi-backed positions on Yemen’s western coast has given Sanaa far more than another city on the map.

In a matter of days, Sanaa’s forces took Hays and Al Khawkhah, seized the strategic Khalid ibn al-Walid military camp and pushed through Yakhtul before entering Mocha, one of the main military and logistical hubs of the Saudi-backed forces on the Red Sea coast.

The advance effectively shifts the center of the battle south — toward Dhubab and Bab el-Mandeb.
And that is where the real significance of the breakthrough begins.

🔸 Sanaa has eliminated much of the Saudi-backed buffer along the southern Hodeidah coast. Positions that had separated Houthi-controlled territory from Mocha for years have disappeared in days.

🔸 Mocha gives Sanaa a much stronger position on the Red Sea. The port had served as a key base for forces led by Tariq Saleh and as a logistics hub for operations along the western coast.

🔸 The breakthrough is also exposing fractures inside the opposing camp. Retreating formations have publicly accused each other of abandonment and betrayal, while Sanaa is now offering an amnesty to fighters who lay down their weapons.

That last move may prove particularly important. If organized withdrawals begin turning into defections, Sanaa could dismantle the remaining western-coast front faster than Saudi airpower can rebuild it.

Riyadh has already responded with massive air strikes. According to the Yemeni Armed Forces, Saudi aircraft carried out 54 raids in just 12 hours across six provinces. Yet the western line continued to move south.

The next obvious pressure point is Dhubab, on the final coastal approach to Bab el-Mandeb.

Sanaa does not need to physically “close” the strait to benefit from controlling more territory around it. Its missiles, drones and unmanned boats have already demonstrated their ability to disrupt Red Sea shipping from positions much farther away. Moving the frontline south expands the territory from which those capabilities can be supported and puts the approaches to the strait under much greater pressure.

For Saudi Arabia, the timing is especially difficult.
With Hormuz disrupted by the Iran conflict, Riyadh has relied heavily on the East-West Pipeline to move crude from its eastern fields to Yanbu on the Red Sea. But oil heading from Yanbu toward Asian markets normally has to sail south through Bab el-Mandeb.

That means the route designed to bypass pressure around Hormuz is now becoming more exposed at its other end.
There are still several possible directions from here. Sanaa can consolidate Mocha, push farther south toward Dhubab and the strait, or use the collapse of the western front to free forces for other contested areas.

But if the defensive line at Dhubab also breaks, Saudi Arabia would face a fundamentally different Yemen: Sanaa would hold almost the entire Yemeni Red Sea coast and sit directly on the approaches to Bab el-Mandeb.

@NewRulesGeo❗Follow us on X

🇨🇳⚠️ China Builds Drone Army Across Air, Land, and Sea

Chinese state media has shown one soldier reportedly directing as many as 200 drones. China Electronics Technology Group (CETC), a state-owned defense electronics company, is now working on the harder problem: keeping the swarm useful after aircraft are shot down, links are jammed, and the target moves.

China Electronics Technology Group says its homogeneous swarms—large groups of similar aircraft—can already reconfigure their networks after losing individual drones, close gaps in formation, and keep working through electromagnetic interference. Their sensors pool information to screen targets, identify camouflage, and calculate positions.

The machines also divide work among themselves. They change routes, avoid collisions, and reassign tasks as the situation changes. CETC says current systems can carry a mission from reconnaissance through electromagnetic suppression and precision strike to damage assessment.

Now China is trying to put aerial drones, ground robots, surface craft, and underwater vehicles on the same network. Aircraft can search far ahead. Ground machines can use terrain and carry different sensors or weapons. Maritime drones can watch channels, approach ships, and push surveillance beyond the coast.

In a conflict around Taiwan, the key factors would be speed and resilience. An aerial swarm finds a ship or missile battery. Another node confirms its position. Electronic warfare platforms interfere with radar and communications. Strike elements attack, while surviving sensors check the result. If one machine disappears, the others are meant to retain the target picture and divide the unfinished work.

The same architecture appears in China’s plans for robotic South China Sea outposts. Aerial, surface, and underwater machines would cover one another, with sensors, weapons, and computing spread among mobile nodes. The aim is to prevent one successful strike from blinding the whole position.

This full air-land-sea network is not in service yet. CETC still lists common interfaces, cross-service coordination, jamming resistance, and realistic testing as open problems. But its building blocks are already being tested at scale: distributed sensing, local decisions, autonomous task allocation, and swarms that continue after losing aircraft or communications.

For US forces, shooting down a few cheap drones would solve very little. They would have to break a moving network while protecting ships, airfields, radars, and supply hubs from several kinds of machines at once. Destroy one node, and the mission passes to another.

@NewRulesGeo❗Follow us on X

🇨🇳🔰 Pentagon's Nightmare: China Advances Second Sixth-Generation Fighter Built For Scale

China is pushing its second sixth-generation fighter deeper into flight testing, and its greatest advantage may be its ability to be fielded in large numbers.

The twin-engine tailless aircraft developed by Shenyang, widely known by the unofficial designation J-50, has now been filmed operating at both low and high altitudes. It is roughly comparable in size to the J-20 and significantly smaller than China’s other next-generation prototype developed by Chengdu.

🔸 A smaller airframe is expected to make the Shenyang fighter considerably cheaper to produce and maintain than the much larger Chengdu design. That could make sixth-generation capabilities practical across far more frontline units.

🔸 Cost may prove as important as performance. Overruns on the F-22 and F-35 prevented the United States from replacing its older fighters as widely as planned, while the J-20’s cost-effectiveness has allowed China to procure it on a much larger scale.

🔸 The new fighter could provide a one-for-one replacement for older Chinese heavyweight aircraft such as the J-11B and Su-30MKK during the 2030s, rapidly transforming the composition of the PLA’s combat fleet.

🔸 The Shenyang and Chengdu prototypes are the only publicly observed tailless fighters currently in flight testing. Removing vertical tails reduces radar reflections but requires advanced software and distributed control surfaces to maintain stability and maneuverability.

🔸 Research associated with the Shenyang program has also explored flight-control technology for tailless aircraft operating from carriers. This raises the possibility that the fighter could eventually serve with both the PLA Air Force and Navy.

China may therefore be building more than another elite aircraft for a limited number of squadrons. A smaller and more affordable sixth-generation fighter could be produced widely, replace existing heavyweights and potentially operate from Chinese carriers.

The Pentagon would then face not a handful of experimental jets, but an entire force transitioning toward sixth-generation aircraft at scale.

Could China field sixth-generation fighters on a scale the United States cannot match?

⚡️ DONBASS NEWS Сhannel - Main News about Life in Donbass and Donetsk

Our agenda:

🔵 Photos and videos of the shelling of the republics;
🔵 Statements from Russian and DPR officials;
🔵 History of the Ukrainian conflict;
🔵 SMO covering from the spot;
🔵 How Russia is restoring the Donetsk People's Republic;
🔵 Opinion of foreign political experts on current events;
🔵 Culture of Russia and Donbass

❓ Do you want to get verified and first-hand information directly from Donbass?

🔗 Join us! @News_of_Donbass
We are on X https://x.com/news_of_donbass/

🇨🇳🗺🚫 China Turns Western Pacific Into No-Go Zone

Australia’s new ambassador to Japan, former intelligence chief Andrew Shearer, arrived with a military pitch: Japanese yards should build Australian frigates, while Australian ranges host tests of Japanese hypersonic weapons. In his September 9 address, he said Japan could use Australia’s “strategic depth.” The reason is clear: China can now threaten the ships, bases and support aircraft behind a US-led Pacific war.

Beijing has built overlapping bands of fire. The DF-21D was designed to attack ships at sea. The road-mobile DF-26 reaches 4,000 km, putting Guam within conventional striking range. In 2025, China unveiled the YJ-17, YJ-19 and YJ-20—three more hypersonic anti-ship weapons.

The missiles do not work alone. Satellites, over-the-horizon radars, aircraft and drones track targets; mobile launchers and warships attack them. China’s orbital reconnaissance network turns long range into a usable strike system.

This changes where the United States can fight. Carriers pushed farther east give their aircraft less time over the battlefield. Tankers and airborne-radar planes must stay farther from Chinese launch areas—a problem Beijing’s hypersonic counter-air project could make worse. Guam, Okinawa and Japanese airfields can no longer be treated as a safe rear.

Australia and Japan are answering with long-range missiles, shared ranges, Japanese-built frigates and more tightly connected forces. They call it deterrence. From China’s side of the map, it looks like another US-aligned strike network taking shape along its maritime frontier.

The geography still favors Beijing. China can fire from mobile launchers at home, under its own air defenses and close to its industrial base. Its opponents depend on exposed islands, tankers, ports and supply lines stretching across the Pacific.

China does not have to sink every carrier or crater every runway. Pushing them back, splitting their defenses and making every sortie harder is enough to change the balance. Shearer’s pitch to Japan shows that this is already happening before a missile is fired.

@NewRulesGeo❗Follow us on X

🇷🇺🛵🎯 Russia Goes After Ukraine’s Digital Kill Chain

A drone strike begins long before a drone takes off. Video has to be collected, targets identified, coordinates checked and orders sent to the unit carrying out the attack. Ukraine has moved much of that work into data centers and AI software. Russia is now attacking that part of the chain.

The Russian Defense Ministry says it struck Kiev’s B-Mobile data center with air-launched weapons and drones. Moscow says the facility stored, processed and transmitted data, including intelligence, and supported Ukrainian military networks. The extent of the damage has not been independently confirmed, but the choice of target matters.

Ukraine’s digital war machine is no secret. Palantir software combines satellite imagery, drone video, open-source material and intelligence from Western partners. In May, Kiev said Palantir’s tools had been integrated into planning deep strikes. Ukraine’s own DELTA system links sensor feeds, headquarters and units across the front.

Avengers AI Labs draws on five million annotated battlefield images, largely taken from DELTA. Models built with that data analyze more than 100,000 drone feeds a month and can identify around 70% of targets in real time, according to Ukraine’s Defense Ministry.

None of this runs in thin air. It needs servers, electricity, cooling, fiber lines and bandwidth. Many drones cannot run the heavier recognition models onboard, so their video has to travel elsewhere. Knock out the server or the connection and the operator must reroute it, wait longer or fall back on human review. The target gets more time to move.

One strike will not switch off a distributed network. Russia does not need to erase it. Delays, partial data and broken connections can be enough to slow the cycle from detection to attack. Repeated hits on data centers, telecom hubs and the power systems behind them would force more processing abroad and make Ukraine even more dependent on Western cloud infrastructure.

This is the weak point of software-defined warfare: the military system sits on top of civilian networks. If a commercial facility processes intelligence or routes military traffic, the servers and fiber inside it become part of the kill chain.

Missiles, artillery and drones are only the visible end. Russia is now moving up the chain toward the computers that help decide where they strike.

@NewRulesGeo❗Follow us on X

🇺🇦🇺🇦📉 Ukraine Begged for F-16s. Now They’re Chasing Drones With Cannons

For two years, Kiev and the Western media treated the F-16 like a flying reset button. The jets finally arrived. Now Ukrainian pilots are rationing air-to-air missiles and, on some daytime sorties, taking off against Russian targets with nothing but the aircraft’s 20mm cannon.

That strips the whole project down to its absurd core. The donated F-16AMs are ageing European aircraft. Their decisive advantage over Kiev’s Soviet fighters was access to NATO weapons—above all AIM-9 Sidewinders and AIM-120 AMRAAMs. Take those away, and the celebrated “game changer” becomes a very expensive anti-drone gun truck with wings.

A Ukrainian Air Force commander using the callsign “Phantom” confirmed the shortage. His crews track missile shipments from the moment they cross the border, hoping they arrive before the next Russian barrage. As he put it: “Any aircraft without weapons is useless.”

Ukraine’s F-16s are mostly being used as airborne air defense, hunting Geran drones and cruise missiles rather than challenging Russian fighters. It is a miserable trade.

Fire an AMRAAM at a cheap drone and Kiev burns through a scarce Western missile. Use the cannon and the pilot must approach an explosive target at close range, which becomes especially dangerous at night. Let the drone pass and it may hit an airfield, power facility or ammunition depot.

The shortage also spreads across the entire air-defense network. AMRAAMs are needed by NASAMS batteries, while Sidewinders have been adapted for ground launchers. Every missile handed to an F-16 is one fewer for the defenses below—and every Russian salvo forces Kiev to divide the same shrinking stock again.

Sending more aircraft does not solve this. It merely produces more airframes waiting for weapons. Russia is also testing operator-controlled Gerans near the front, meaning an F-16 sent to save missiles by using its cannon may itself become the hunted.

Russia does not have to shoot down every F-16. It only has to keep the barrages coming until the Western missile pipeline falls behind—a wider war of attrition against NATO interceptor stocks.

Kiev finally received the fighters it spent months begging for. Russia has reduced them to an expensive way of chasing the same “flying mopeds” Ukrainian propaganda once laughed at.

@NewRulesGeo❗Follow us on X

🇨🇳🔰⚠️ Pentagon's Nightmare: China Maps Out Sixth-Generation Fighters For Satellite-Denied Drone Warfare

China is preparing its future fighters for a battlefield where satellite navigation is jammed, pilots face extreme workloads and unmanned wingmen must reorganize as conditions change.

A new paper from Shenyang researchers outlines an aircraft capable of taking over its own flight and commanding an entire manned-unmanned formation instead of fighting alone.

🔸 The paper was led by Zhang Dong, chief flight-control designer at the Shenyang Aircraft Design and Research Institute, and published in the peer-reviewed journal Aircraft Design in August.

🔸 The institute is associated with the next-generation prototype widely known as the J-50. However, the paper does not name the aircraft or confirm that any of the proposed technologies have already been installed on it.

🔸 Automatic flight controls could take over if the pilot loses consciousness or faces an extreme workload. The system would keep the aircraft stable and potentially allow it to continue its mission while the pilot recovers or concentrates on higher-level decisions.

🔸 China is also exploring quantum-navigation technologies for operations where satellite positioning is unavailable. These could help future fighters determine their location after GPS, BeiDou or other satellite signals have been jammed or disrupted.

🔸 The most ambitious concept involves commanding reconfigurable formations of unmanned wingmen. Drones assigned to attack, reconnaissance and protection could change roles and reorganize during a mission as the tactical situation evolves.

🔸 Under this model, the pilot becomes the commander of a distributed combat network. The aircraft provides communications, sensor fusion and decision support while automation manages routine flying and helps coordinate multiple unmanned systems.

If these concepts reach service, the Pentagon would face more than another stealth fighter. It would confront an airborne combat network capable of navigating under satellite denial, adapting its drone formation in real time and sending unmanned systems forward while the crewed aircraft remains farther from danger.

The fighter would become the command center for an entire formation designed to keep fighting even when navigation signals and conventional formations break down.

Which will matter more in the next air war: the fighter itself or the drones it controls?

🇨🇳💻 China Maps Path to the Last AI Built by Humans

Today’s most advanced AI still depends on large teams of people. Engineers choose the data, design training runs, test failures, rewrite code and decide what to try next. Chinese researchers have now mapped a route toward a system that takes over that entire cycle—and eventually improves the way AI improvement itself works.

A joint team from ByteDance, Tsinghua University, the Shanghai AI Laboratory and other institutions has outlined a five-stage path toward genuine recursive self-improvement.

At the first stage, AI merely carries out upgrade procedures written by humans. It then starts choosing its own improvement strategy, deciding what new data or experience it needs and adapting after deployment. At the final stage, the system would refine the very methods used to build better AI.

This is different from a chatbot correcting one bad answer. The improvement must remain after the task ends, become part of the system and pass into later versions. Each new model would begin with the lessons learned while creating the previous one.

The attraction is obvious. Training a foundation model currently consumes enormous amounts of engineering time and computing power. An autonomous research loop could launch experiments, compare results, discard failed approaches and keep successful ones without waiting for humans at every step.

Washington has tried to turn advanced chips into a choke point for Chinese AI. China is already building domestic alternatives to restricted Western equipment. Recursive self-improvement attacks the problem from the other side: reduce wasted training runs, automate more research and extract more progress from the hardware available.

Chinese companies are already developing pieces of this system. Z.ai plans to direct about 60% of the proceeds from its latest $5B fundraising round toward new GLM models and a “fully self-training system”. MiniMax has tested models that update memory and acquire new skills during reinforcement-learning experiments, while DeepSeek has built an agentic framework able to execute code and handle complex chains of tasks.

Nobody has reached the final stage, and the paper offers no timetable. Software development provides the clearest testing ground; robotics and scientific research are far harder. Every autonomous update would also need to be tested before entering a live model.

But China is assembling the research base, companies and practical systems needed to move in that direction. Its generative-AI patent output already exceeds the rest of the world combined.

The “last AI built by humans” would not be a finished machine. It would be the first one capable of turning its own development into a continuing production line—building each successor faster, cheaper and with less human direction than the one before it.

@NewRulesGeo❗Follow us on X