Infrastructure plan for six networks includes orders for ground equipment for commercial space companies

The Tsinghua team has released VeriLoop Coder-E1, an open vertical coding model based on Qwen3.6-27B. This model is designed for repository-level code repair and agent software. The model demonstrated high results across various benchmarks: SWE-bench Verified 85.20, SWE-bench Pro 62.38, Terminal-Bench 2.0 76.40, and DeepSWE 33.63.

Among all open-source models with 32 billion parameters or less, VeriLoop Coder-E1 ranks first in three benchmarks and second in DeepSWE. When compared against all open-source models, it achieved second, first, first, and fifth place, respectively.

The main advantage of this development lies in fine-tuning using the PEFT method in a narrow domain, coordinated via Self-Harness. With the base layer of Qwen3.6-27B frozen, the project enhances adherence to tool contracts, links evidence to inferences, identifies uncertainties, explains validation failures, performs local repair, and controls rollback using only a small number of trainable parameters loaded through a detachable Surface Host adapter.

Self-Harness manages multi-round chains: the first round generates candidates based on problem descriptions, repository context, interface constraints, and acceptance criteria. Subsequent rounds recompile the generated code, check error reports, tool logs, interface conflicts, and counter-evidence, packaging them into structured Markdown work packages. This allows the model to fix locally refuted issues by relying on verified results rather than mechanical query repetition.

This cycle represents a closed, evidence-driven loop, not merely trial and error. Each generation round undergoes checking for counter-evidence, exploration, revision, and re-verification; only verified fixes proceed to the next stage. When verified fixes alter how future problems are detected, evidence is defined, and fixes are implemented, the cycle transitions into a 'Evidence-Driven Spiral' mode, following the 'refutation-exploration-repair-verification-internalization' chain.

The team redefines the concept of recursive self-improvement: it is not a system that constantly modifies itself, but methods adjusted by evidence that change the system's approach to error detection, assessment, and correction, remaining falsifiable by new evidence. Drawing on Karl Popper, unfalsifiability is considered a drawback. Passing verification does not imply recursion occurred; the recursive threshold is crossed only when a fix reveals transferable methodological flaws, enters future correction mechanisms, and changes the way unknown errors are detected and assessed. The boundary is intentionally set: the model's assets remain open, while the Self-Harness control stack remains closed.

The cluster of national data centers in Qingyan, Gansu province, has reached a computing power of 185,000P, with the share of intelligent computing at 99%. The center uses green energy at a price of 0.398 yuan per kWh and integrates over 700 enterprises in a chain from chips to applications.

Two years after the area was known as cornfields, the East Data and West Computing park in Qingyan, Gansu province, now features modern data centers on the Loess Plateau. The first phase of the 'Zhihui' project will provide 108,000P, construction of which is planned to be completed by October 30, with all available capacity pre-booked before construction began.

Qingyan possesses four natural advantages: location, climate, geology, and electricity, but pricing is the decisive factor. The direct supply model of green energy stabilizes tariffs at 0.398 yuan per kWh, which is 30–50% lower than in eastern China. An average data center consuming 10 million kWh monthly saves between 20 and 50 million yuan annually.

Gansu Mobile built the country's first assembled micro-modular industrial data center in just 100 days, reducing the timeline by 20 months compared to traditional schedules. Among the eight nodes of East Data and West Computing, Qingyan stands out because, unlike the others that only allocate quotas for green energy, the Qingyan government is involved in building and connecting the power supply with integrated storage systems.

The second level of development includes the network and ecosystem. Due to its central geographical location, Qingyan has created a circle with a latency of 1–3–7–14 milliseconds; every millisecond reduction increases training efficiency by several percentage points. The ecosystem integrates hub nodes, intelligent computing centers, and fiber optic networks involving over 700 enterprises, forming a chain from chips to applications: upstream are Enflame Technology and Loongson, in the middle are Kingsoft Cloud and Capital Online, and downstream are Moonshot AI and Zhipu AI.

Computing is also integrated into agriculture: a smart frost protection system saved flowering apple trees, and the Ruixue apple base, created based on a digital map, achieved 85% high-quality fruit, bringing households an income of 2,000 yuan, while the intelligent mining operation Huaneng Qingyang 5G-A made the energy extraction process safer.

The computing power of the Qingyan cluster exceeds 185,000P, with intelligent computing accounting for over 99%, covering more than 25% of the demand for national AI model tokens. Among the eight hub nodes, Qingyan ranks first in terms of growth rate, power increase, intelligent share, and consumption sufficiency. Talent attraction has transformed from a policy into an ecosystem approach: the arrival of China Mobile attracted Kingsoft Cloud and Capital Online, who connected with Xiaomi partners, implementing the 'attract one, bring the chain, build the cluster' model.

Qingyan's vision is to transform computing into a public service, accessible like water and electricity. The 'Zhihui' project positions itself as a token factory, stimulating industry modernization from selling server racks to selling tokens. Future computing resources will be distributed like an electrical grid: enterprises will provide computing power, and end-users will consume it on demand with real-time payment. Electrical signals passing through fiber optics across the Qinling Mountains reach terminals thousands of kilometers away, becoming intelligent dialogues, AI images, and autonomous driving solutions. Qingyan's computational wheat fields are growing, representing not just a set of 185,000P digits, but a western city redefining itself in the era of the digital economy.