A recent document titled "Transcript of Liang Wenfeng's Four-Hour Investor Meeting" has circulated widely in the tech and finance sectors. According to multiple sources, during a private discussion, Liang Wenfeng outlined a "restrained" strategy: avoiding C-end traffic, not pursuing 3D or world models, only earning reasonable profits, open-sourcing the strongest models, and arguing that AI could eventually account for 10% of global GDP and thus cannot be monopolized. As the hype grew, DeepSeek publicly denied the report, calling it "a distortion of concepts and expressions." Media outlets sought verification from DeepSeek but received no response by the deadline. Liang Wenfeng himself did not confirm the document's authenticity, while the organizers claimed the text "preserved the original meaning as much as possible, with only minor edits." This widely circulated yet unverified material has split into three layers: the leaked text, media interpretations, and the denial from those involved.
The truth will be judged by time, but the cash flow structure reflected in the leaked material warrants careful observation by the insurance industry.
Setting aside the disputes, a scholarly analysis of the leaked logic reveals that on one side, Liang Wenfeng's High-Flyer Quant leverages self-developed supercomputing and millisecond-level trading systems to achieve a technical edge over ordinary investors in the A-share market. Retail investors execute T+1 trades and manual orders in seconds, while quant funds use intraday short-selling via underlying positions and margin, with microsecond-level order placement and cancellation through colocation. During market fluctuations, homogeneous factors resonate, leading to synchronized order cancellations and sell-offs. In this structure, the game between quant funds and retail investors is not a contest but a one-sided extraction, like a pump continuously siphoning liquidity from retail investors and even some institutional players into their own accounts. Data shows that in 2025, returns on High-Flyer's products widely exceeded 50%, while retail investors generally suffered losses of over 30%, earning it the label "retail investor harvester" in public opinion.
On the other side, DeepSeek uses the cash flow generated from the domestic market to fund its large model development, open-source initiatives, and API pricing, which it has driven down to a level where "equipment costs can be recovered in ten months." This benefits global developers and domestic small and medium-sized enterprises.
This creates a unique closed loop: cash flow extracted from the domestic secondary market through speed and technical asymmetry is used as a global technological public good. For an individual, this logic may be self-consistent—the quant side is unlikely to lose, as it has a dominant advantage over retail investors, preventing the pump from reversing. Using this as a foundation, if AI succeeds, the returns are global; if AI valuations decline, the money extracted from retail investors is merely considered a necessary cost for exploring the frontier of human technology. After all, there is no rigid redemption contract between the quant account and the AI account.
In contrast, the insurance industry, which has recently performed well in the equity market, does not have this kind of flexibility. From a fundamental perspective, no business should attempt this operational logic.
Insurance premiums are the price ordinary people pay for hedging against uncertainty, representing liabilities with duration and redemption commitments. A life insurance policy may be claimed 30 years later, and a pension policy may pay out 40 years later. This dictates that insurance must adhere to two boundaries: on the underwriting side, profits come from actuarial science, suitability, and cost coverage; on the investment side, profits come from duration matching, safety margins, and cross-cycle allocation. Value can be transferred between these two boundaries, but the liability side cannot forever rely on the investment side to fill gaps. While investment is indeed used to cover liability costs—a fundamental aspect of insurance operations—short-term investment gains must not be mistaken for normalcy, nor should they be used to ignore the foundational value of underwriting.
It cannot be overlooked that the insurance industry has long held a misconception. When equities perform well and tech stocks rise, insurers see substantial paper gains, leading to slack in underwriting. Sales teams can then raise dividend illustrations, and expense departments can continue spending on scale, thinking, "The asset side has performed well this year, so the loss from interest rate spreads can be deferred for now." In reality, this uses short-term equity market gains to mask the long-term crude development of underwriting, substituting the asset side's beta for the liability side's alpha. This is structurally similar to "using domestic market extraction from retail investors to fund global tech exploration," but with different foundations. The former extracts funds from retail investors through speed and technical asymmetry, while the latter uses the long-duration funds entrusted by policyholders for cross-cycle risk smoothing.
The fundamental difference between insurance and quant funds may lie here. Quant funds have a dominant advantage over retail investors, acting like a pump that doesn't reverse, with no rigid payment obligations to individual investors. If insurers use asset gains from a bull market to cover underwriting losses, when the bubble bursts, equity values decline, long-term interest rates fall further, and high-guaranteed-rate policies enter their payout peak, the high-return policies sold with misleading sales tactics must be paid out in cash. At that point, there is no other quant account to fall back on. The only way forward is to repeat the old path: issue new products that deviate further from reality, using more aggressive illustrations, looser underwriting, and a shorter-sighted focus on scale, collecting new premiums to pay off old obligations. This closes the interest rate spread spiral, and the lessons from Japan in the 1990s and some domestic insurers in the 2000s are not far behind.
Therefore, for the insurance industry, the investment side must be more cautious, and the underwriting side must be more refined.
Investment caution does not mean avoiding equities or tech stocks, but rather ensuring that equity allocation adheres to duration characteristics. Tech stocks are part of a portfolio, not a narrative lifeline. Profits from a bull market should first strengthen solvency and set aside provisions for interest rate spread losses, rather than being used as capital to fuel new highs in premium scale.
Refined underwriting means truly implementing suitability management, not packaging high-variable dividends as guaranteed returns, not betting liability costs on asset gains, and squeezing out expense ratios, relying on risk pricing rather than sales tactics to make money. In short, the dignity of insurance lies in being able to fulfill promises to customers even with unremarkable investment performance, through careful underwriting. The true maturity and stability of the insurance industry are rooted in the long-term stability of the liability side, not the short-term volatility of the asset side.
The same cash flow is described in online discourse as both "the grace of Saint Liang" and "the coldness of a harvester," which perfectly captures the two sides of Liang Wenfeng. He uses the extraction from ordinary investors in the domestic market to fund global technology exploration, because there is no rigid redemption contract between the two extraction points. The insurance industry is entirely different. The liability and asset sides are linked through insurance policies, forming a long-term fiduciary responsibility. Any attempt to "use the stock market to support underwriting" essentially bets the policyholders' future on current valuation sentiment.
True business goodness is not about using cash flow from Market A to fund dreams in Market B, but about doing the best within the boundaries of Market A. The insurance industry does not need to envy another mountain's fire; its own mountain is meant to serve as a shield. What the stock market earns belongs to the stock market; the stability of underwriting is the very foundation of the insurance industry's existence. This is the most solid essence of "insurance being insurance."