Original Title: "Exclusive Report by LatePost | Tang Jie's Internal Memo States Implementation of Touch High Plan, 'Not Summiting is Failure'"
Source: LatePost Team

LatePost has learned exclusively that today, the founder of Intellisys, Tang Jie, released an internal memo outlining Intellisys's understanding of the upcoming competition in AGI. In the memo, Tang Jie stated that Intellisys will continue to pursue the so-called "Counterintuitive" route, initiating the "Touch High Plan," which focuses on AGI research rather than short-term commercialization.

On the road to AGI, there are several must-cross peaks, which are also the most turbulent areas of today's technological wave. The four peaks listed by Tang Jie are:

- Long Horizon Task

- Autonomous Agent System

- Fully Self Training

- Ultimate Security Governance

Among them, Ultimate Security Governance is particularly emphasized. Intellisys plans to invest billions of resources to tackle mechanical explainability, which means clarifying the neural logic behind model decisions and driving black-box systems towards transparent and explainable systems.

On January 8th of this year, Intellisys debuted on the Hong Kong Stock Exchange with an issue price of 116.2 Hong Kong dollars per share for its large-scale model. In the following six months, Intellisys's stock price soared to a peak of 2980 Hong Kong dollars, more than 24 times the issue price, and its market value once exceeded 1.3 trillion Hong Kong dollars.

On July 8th, shares held by 11 cornerstone investors were unlocked, with over 25 million shares of Intellisys, valued at over 40 billion Hong Kong dollars, entering the market. The market had expected selling pressure, but the stock price of Intellisys rose instead of falling. The next day, Intellisys announced a new share placement at a price of 1588 Hong Kong dollars per share, with a discount of approximately 13%, raising approximately 31.4 billion Hong Kong dollars. This was the highest single placement record for a Hong Kong AI company this year.

According to Intellisys, the funds raised from the placement will mainly be used for base model research and development, computing power infrastructure construction, commercial expansion, and global ecosystem layout. Tang Jie stated in the open letter that the assessment of the leap to the upper limit of intelligence is also "the cognition we most want to convey to everyone."

In a series of comprehensive evaluations, the Genosis GLM-5.2 model has been recognized to have reached the forefront of overseas model capabilities and, due to its open-source nature, is popular within the technical community.

The following is the full content of an internal memo:

《The Giant Wave is Here》

— To every Genosis member and partner who cares about the future of artificial intelligence

Who We Are: "Essence, Counterintuition, Focus"

Genosis has never been a company chasing trends. It grew out of a lab, carrying with it the methodology of this lab for twenty years. This methodology can be summarized in three words: Essence, Counterintuition, Focus — think deep enough to dare to choose against the grain; choose against the grain enough, and you must persevere long enough.

Looking back, almost every key decision we made seemed "counterintuitive." In 2006, we sat on the cold bench with an academic search system on a desktop computer because we had realized that behind it was the "mining of the evolution of disciplines," a question worthy of a ten-year answer; from 2021 to 2022, when "making machines think like humans" was seen by the vast majority as a lunatic moonshot, we reallocated resources, bet billions of parameters, and created the GLM-130B — that was a year and a half before ChatGPT shook the world; and on January 8, 2026, the day Genosis went public on the Hong Kong Stock Exchange, we saw it as a brand new starting point, firmly committing to a full return to basic model research and vigorously impacting the next generation of models.

While others ring the bell, we reset to zero. This is not a pose, it is a belief — since the destination is AGI, short-term gains or industry trends are merely scenery along the way to the endgame.

What has supported us to this day is an extreme focus and a pure idealism. It took us ten years to go from an academic search system on a desktop computer to tens of millions of users; we have been on the path of large models for nearly ten years, and we will firmly continue to delve deep and keep going. Today's Genosis is a group of people willing to question the essence, dare to be counterintuitive, and focus on getting things done — this is the source of Genosis's core competitiveness.

How We See This Era: The Boundary of Intelligence is Being Redefined

If we have learned one thing in the past twenty years, it is this: the real business opportunity does not lie in product and model fine-tuning, but in the leap of the boundary of intelligence. This is our fundamental judgment on the current AI transformation, and it is the cognition we most want to convey to everyone.

This transformation is fundamentally not an innovation in product or business model, but the technological revolution itself has elevated the "boundary of intelligence." Whoever can first raise this boundary by an inch can redefine the capability boundary of various industries. The new generation of AI enterprises focusing on first principles are contending for this inch of breakthrough.

The evolution of the intelligence upper bound has a clear path. Artificial Intelligence is undergoing a transition from perceptual intelligence to cognitive intelligence—machines are no longer just "seeing" and "hearing" but are starting to "understand" and "reason." The next step is pointing towards AGI.

We have a simple yet strict definition of AGI: AGI is not the intelligence of a single genius but the sum of human intelligence. It should have the ability to create "relativity"-level original knowledge, which is the sole criterion we use to measure if we have truly reached the peak. On the road to this endpoint, there are several mountains that must be crossed, which are also where today's technological waves are most turbulent:

First Peak: Long Horizon Task Capability

The most exciting breakthrough today is teaching models to complete an extremely long task—not immediate Q&A, but planning and execution spanning weeks, months, and even years. For example, a model can tirelessly search for vulnerabilities in software, essentially learning the mindset of a top security expert and then amplifying it through the machine's endurance.

Second Peak: Fully Autonomous Intelligent Agent System

Above the long-horizon task, intelligent agent collectives that can autonomously drive, collaborate, and operate 24/7 will become a new form of productivity. We have mentioned the "One-Person Company OPC" before, but the pace of technology is faster than expected—we are moving towards the "Fully Automated Company NPC." The three challenges of memory, continual learning, and self-judgment that were once thought to require a paradigm shift to solve have gradually been eliminated under the dual drive of technology and application: long-context and retrieval-augmented generation (RAG) are approaching the rudiments of memory; the increase in model iteration frequency itself is approaching continual learning; cutting-edge models have already shown the buds of self-judgment.

Third Peak: Self-Evolving

This is the most difficult and enticing peak. AI training AI has taken shape—models writing their own code, cleaning and synthesizing data by themselves, and self-training. This may require some computing power but saves the most valuable human effort and time. In the era of large models, speed is crucial, and rapid iteration will directly widen the intergenerational cognitive gap. When leading overseas companies begin building clusters of a million to even two million chips, the true purpose is likely to let the model train itself.

What Will Happen After Crossing These Three Mountains?

AI will start learning what "I" is, what self-awareness is; further on, it will touch upon human emotions; even further, consciousness itself. From perception to cognition, from cognition to generalization, from generalization to superintelligence (ASI) — this path has already been laid out, the giant wave has arrived, and it is irreversible.

This is not just our opinion. Google DeepMind made a stark statement in their report "From AGI to ASI": even if the capabilities of a single model remain at human levels forever, as long as computing power continues to grow, superintelligence may be forcibly "squeezed" out. They speculate that if globally runnable AGI instances grow tenfold every year, there will be one billion of them within five years. These intelligent entities, sharing the same underlying brain, with a hundredfold increase in thinking efficiency and zero-cost replication of experience at the group level, are equivalent to ASI. In other words, the transition from AGI to ASI requires breakthroughs at the algorithmic level as well as the aggregation of massive computing resources.

This irreversible trend will penetrate the entire technology stack from top to bottom: when AGI arrives, today's applications may need to be rebuilt as AI-native, or may no longer be needed at all; operating systems may be rewritten, and when you turn on your computer in the future, you may see an "LLM OS" where all functions are generated on demand; further down the line is the challenge to the von Neumann architecture that has been running for eighty years. Finance, law, e-commerce, the internet... no industry will be left untouched. Many friends have come to me saying they want to transform their businesses and keep up with the pace of AI, but those who truly perceive that "this irreversible transformation has already begun" are still very few.

Our Direction of Full Effort: "Touch High"

After recognizing the trend, all that's left is to choose. And the choice of the Intelligence Spectrum has always been "counterintuitive" — at a time when the industry is accelerating commercialization, we have decided to aim higher.

We have named this strategy the "Touch High Plan." At the historical moment when artificial intelligence is leaping from perception and cognition to fully general intelligence, the Intelligence Spectrum will, in a "touch high" posture, challenge the physical and algorithmic limits of current technology. Over the next two years, we plan to strategically invest — not to pursue short-term application monetization, but to aim directly at the next peak of AGI.

This investment will focus on four core engines:

First, Long-term Tasks. Transitioning AI from "instantaneous answers" to "grand projects," developing a new generation memory architecture that allows models to go through the entire project life cycle "learning while doing and remembering," and possessing the top-level ability to autonomously decompose grand goals (such as "designing a new type of anti-cancer drug molecule") into thousands of executable sub-tasks.

Secondly, Autonomous Intelligent Agent System. Transitioning from "Intelligent Assistant" to "Digital Employee," building an intelligent agent society with thousands of different professional "personalities" and "skills," allowing them to autonomously debate, collaborate, review code, schedule resources, and achieve "self-driving" level of digital productivity.

Thirdly, Fully Self Training. As high-quality human data is about to run out, turning computing power into the fuel of evolution—constructing a high-quality synthetic data factory, achieving knowledge creation from nothing through AI-to-AI competition (Self-Play), and empowering systems with the ability to refactor their own code within a secure sandbox, enabling evolution to surpass the physical limitations of human engineers.

Fourthly, Ultimate Security Governance. This is one of the key points I want to emphasize among the four engines.

The more powerful the capability, the more robust the security constraints must be. Since its inception, Brainwave has established the principle: AI must serve the well-being of humanity and national strategy. The company has abandoned external security patches and insists on embedding human ethics, social norms, and national laws and regulations as fundamental axioms in the model's value function; it plans to invest billions of resources to tackle "mechanical explainability," clarify the neural logic behind model decisions, drive black-box systems to transparent and interpretable systems; while actively engaging in international AI governance to prevent AI technology abuse.

This sense of urgency is not groundless. When the most advanced overseas models have delayed full public release due to risk considerations, and their company leaders publicly warn of AI's profound impact reshaping the global power structure, we should be even more aware: the realization of superintelligence and superalignment research must advance simultaneously. This is also the proposition we repeatedly examine when facing disruptive technologies—history has repeatedly shown that when a technology reaches a power level capable of changing the course of civilization, security is no longer incidental but the fundamental premise for the technology's survival and permitted application.

Open Ecosystem: The Underlying Logic of Inclusive Intelligence and Secure Governance

We have always believed that artificial intelligence, as a strategic technology leading the future, cannot achieve long-term development without an open and collaborative industrial ecosystem. The value of cutting-edge intelligence lies not only in the technological breakthrough itself but also in whether it can empower various industries and benefit every developer. We firmly believe that true security is not built on technological isolation and barriers but stems from wide-ranging co-construction, sharing, and oversight under the sun.

Based on a profound understanding of technology inclusion, Brainwave has provided its strategic response. Recently, we have released the most powerful open-source model to date, GLM-5.2, supporting truly usable million-scale contexts, maintaining a leading position in long-range tasks, open to all users, and officially open-sourced under the most permissive MIT license—anyone can download, deploy, and commercialize without constraints. This is the company's firm stance expressed in product form.

We choose to believe in another path: cutting-edge intelligence should not only belong to a few, nor should it be subject to arbitrary rules. It should be open, accessible, buildable, and serve every developer.

This is not contradictory to "reaching new heights"; instead, it is two sides of the same coin: with one hand reaching upwards to challenge the limits of intelligence, and the other hand paving the way down to make the most cutting-edge capabilities as open and inclusive as possible. The heights we reach belong to all humankind, and the paths we pave belong to everyone.

Epilogue: Why Now, Why Us

Some may ask: Why, after the listing of IntelliSpectrum, continue to invest core resources and "reach new heights" in the most uncertain direction? Because we believe in a simple truth: true summit conquerors will turn the mountain into a road.

The essence we want to clarify, once condensed through the "Enlightenment Big Model" project into the consensus of hundreds of scientists, has become the cornerstone for a generation of entrepreneurs to take off through IntelliSpectrum's industry investment and the entire ecosystem. Today, we want to build this road higher and wider—to a height that can protect ourselves, safeguard national security, a height that can give humanity the opportunity to explore more unknowns, explore the mysteries of the universe; as well as wide enough to allow every developer, every team, to come up.

In the era of AGI, these things that were once out of reach have for the first time become achievable. This is the greatest fortune of our generation, as well as the heaviest responsibility.

The giant wave has arrived, the trend is irreversible. IntelliSpectrum wants to be the one who faces the waves and reaches new heights.

Not reaching the summit is failure.

This time, what we aim to achieve is the height that belongs to all humankind.

Founder of IntelliSpectrum, Tang Jie

July 11, 2026

Original Article Link

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia