Two vastly different public reputations of Liang Wenfeng: he is deified as a saint in the AI world, while people are settling accounts with him in the A-share market arena.

Liang Wenfeng's nearly four-hour investor exchange meeting has gone viral across social media.

Strictly speaking, this material cannot yet be regarded as fully reliable official information. The circulated transcript notes that the recording took place on May 20, was transcribed via speech recognition and organized by AI without distinguishing between speakers, and some figures and proper nouns may contain errors; public reposters have also generally emphasized that the manuscript has not been confirmed by Liang Wenfeng himself.

But this does not seem to stop everyone from "listening to the sacred teachings of Liang".

Because the Liang Wenfeng described in the content perfectly fits people's imagination of a Chinese technology hero: he does not pursue an IPO, does not aim for profit maximization, and has no intention of becoming the next Tencent or ByteDance; he adheres to open source principles, emphasizes goodwill and restraint; he sets no KPIs, does not encourage overtime work, and believes that products, users, and revenue are all just by-products on the path to AGI.

Even when discussing API pricing, he does not talk about how to extract the highest possible profit, but rather that the equipment only needs to pay for itself in about ten months. His core logic is that the future benefits of AI will be large enough, so there is no need to pick up every tiny sesame seed encountered along the way.

This reads more like a manifesto of technological idealism.

As a result, Liang Wenfeng has once again been hailed as "Saint Liang".

The most popular narrative around DeepSeek is already a near-perfect national technology story: in a context where high-end chips are restricted and overseas large models hold a first-mover advantage, a group of Chinese engineers who do not come from the traditional Silicon Valley system have used less computing power and lower costs to build a model that is qualified to sit at the global table.

DeepSeek has not truly "lifted" the chip blockade, but it has at least convinced many people that a disadvantage in computing power does not equal a dead end in technology. Problems that others solve with more GPUs can also be overcome by Chinese teams through algorithms, architecture optimization, and engineering efficiency.

This sentiment has accumulated around Liang Wenfeng and eventually turned him into "Saint Liang".

But Liang Wenfeng has another side.

In the context of A-share investors, High-Flyer has long been one of the representatives of domestic quantitative investment. High-Flyer applied machine learning to trading in its early days, launched its first AI model in 2016, and gradually realized the AI transformation of investment strategies. Around 2021, its assets under management once exceeded 100 billion yuan, but it also experienced significant drawdowns and issued public apologies.

When many ordinary investors mention quantitative trading, their impressions are all about "harvesting retail investors", overcrowding in small-cap stocks, programmatic trading, and market fluctuations they cannot understand.

Therefore, the same Liang Wenfeng is the "Saint Liang" leading the breakthrough of domestic technology in the AI industry; but back in the capital market, he is easily seen as a synonym for the power of quantitative trading.

This is actually the most interesting part of Liang Wenfeng's public reputation.

"Saint Liang" and "Quantitative Liang" are not two different people, but the same set of capabilities applied in two different interest scenarios.

What Liang Wenfeng is truly good at has probably never changed: believing in models, believing in computing, believing in efficiency; concentrating resources to solve a small number of key problems, using engineering optimization to make up for insufficient resources; not over-relying on human experience, nor blindly believing in traditional organizational management.

These capabilities were first used in High-Flyer Quantitative, and later migrated to DeepSeek.

The problem is that quantitative trading is usually understood as a stock game. The so-called excess return is essentially a relative return: if you earn a little more than the market, it means someone else earns a little less than the market. The more advanced the technology, the easier it is for ordinary participants to feel that they are standing on the opposite side of the algorithms.

But large models, at least at this stage, are more like an incremental market.

As model capabilities improve, inference costs decrease, and code efficiency increases, both developers and ordinary users can benefit. DeepSeek drives down prices and open-sources its models, so society feels the technological dividend rather than the loss of counterparty traders.

The same knife is seen in the quantitative market as a tool cutting into other people's cake, but in the AI industry, it becomes a tool to reduce computing power costs and break through overseas technological barriers.

The public's moral evaluation of it will naturally be completely different.

Therefore, DeepSeek is not a betrayal of High-Flyer Quantitative. On the contrary, DeepSeek grew right out of High-Flyer.

Quantitative trading has trained the team's sensitivity to data, models, computing power, and costs, and also accumulated early capital and infrastructure. Without High-Flyer, it would be hard to imagine a DeepSeek with such a unique temperament. The most amazing part of DeepSeek is that the extreme efficiency methodology formed in the trading market has been moved to a field that is more likely to gain social recognition.

From this perspective, "Saint Liang" and "Quantitative Liang" explain each other.

It can even be said that it is precisely because Liang Wenfeng first obtained resources in the highly realistic, return-focused quantitative industry that he later qualified to talk about goodwill, open source, and restraint in the AI industry. Idealism does not float in a vacuum; beneath it are the capital, computing power, and engineering capabilities accumulated from quantitative trading.

Of course, the most noteworthy parts of the meeting minutes are not those golden quotes that are perfect for screenshots and reposting.

It is certainly touching when a founder says he does not pursue profits, will not build a super app, sets no KPIs, and does not encourage overtime. But a company cannot ultimately be judged solely by the founder's self-description. As DeepSeek grows larger, begins to raise financing, issues stock options, and faces issues of team stability and commercialization, the real test is whether the so-called "restraint" can be institutionalized, rather than remaining just Liang Wenfeng's personal value.

"Saint Liang" is a public projection, and the synonym for quantitative trading is also a public projection.

The former embodies people's expectations for the breakthrough of domestic technology: it is best to have a quiet, pure, money-averse genius who breaks the blockade for China.

The latter carries ordinary investors' unease about the complex market: it is best to find a specific person to take responsibility for those incomprehensible trades, drawdowns, and fluctuations.

The real Liang Wenfeng is probably more complicated than both of these images, but also more ordinary.

He is neither a saint who descended from the sky, nor just a quantitative capital hiding behind machines harvesting the market.

It is just that the same set of capabilities has produced completely different social effects in different arenas.

So, is he Saint Liang, or just Wenfeng?

Don't rush to canonize him, and don't rush to disenchant him either.

What really matters is not what he said this time, but whether DeepSeek can continue to implement "restraint" in its products, open source initiatives, pricing, and business decisions in the coming years.

Saints are born from quotations.

Companies ultimately prove themselves through actions.

This article is from the WeChat official account "Wu Duidui" (ID: esnql520), written by Wu Duidui, and published with authorization from 36Kr.