After reading a few quotes from George Soros, I began to grasp the essence of investing.

The person telling this story says: 'I don’t need to do anything—just hold long-term, and eventually dividends will recoup my cost.' It might actually work—but the probability is extremely low; historically, truly successful cases are exceedingly rare. III. How to Do It 1. What to Buy You can buy anything—as long as it generates profit. Don’t confine yourself unnecessarily. ・ For example, buying only stocks with low P/E ratios; ・ Buying only consumer-sector stocks; ・ Buying only internet stocks; ・ Investing exclusively in ChiNext; ・ Buying only small-cap stocks; ・ Buying only blue-chip stocks; ・ Buying only high-growth stocks; —all of these are self-imposed constraints... The crux lies in correctly predicting both the price direction of a given asset *and* the timing. If you correctly predict a stock’s upward move, you can buy and profit; if you correctly predict its downward move, you can short-sell and profit. The concept of a 'circle of competence' is a myth. Specific industries, company types, or markets (A-shares, B-shares, H-shares) should not serve as artificial boundaries for defining such a circle. Wherever you perceive potential for a turnaround—that’s where you should invest. Soros once said: 'Markets are foolish—and you needn’t be exceptionally intelligent either. You don’t need to understand everything, but you must understand *something* better than others.' 2. When to Buy This is the critical question—and different schools of thought apply distinct criteria. Benjamin Graham: 'I boldly distill the secret of successful investing into a four-word motto: margin of safety.' His method involves first estimating a company’s intrinsic value, then waiting until the market price offers a 40–60% discount before considering purchase. Graham’s approach is excellent—but flawed: it overemphasizes valuation while underemphasizing timing. Warren Buffett: 'It’s far better to buy a wonderful company at a fair price than a fair company at a wonderful price.' This statement has led countless investors to losses—because 'fair price' has no objective standard, 'wonderful company' is subjective, and intrinsic value isn’t an immutable benchmark; there’s no floor—only ever-lower floors. This vague maxim causes many to become careless at the point of purchase. Regarding the 'when to buy' question in secondary markets, Graham’s perspective is worth learning; Buffett’s, however, should be firmly discarded. On this very question, Soros, Livermore, and O’Neil have already delivered the most brilliant answers—their methods resolve *both* what to buy *and* when to buy, surpassing Graham’s in sophistication. Soros: 'The financial world is turbulent, chaotic, and inherently unpredictable. Only by discerning underlying realities can one prevail. Attempting to treat every market movement as part of a mathematical formula is futile. Mathematics cannot control financial markets—psychological factors do. More precisely, only by grasping the crowd’s instincts can one control the market: that is, understanding *when* and *how* the crowd will converge around a particular stock, currency, or commodity is essential for investor success.' This foundational statement by Soros underscores the paramount importance of timing—*when*, and *in what manner*, the crowd gathers around a given stock. Yet he stops there: after laying down this principle, he offers no concrete, actionable indicators. Jesse Livermore—the brightest star of the U.S. stock market—had long since answered this question brilliantly, providing a key buy signal: 'Always patiently wait for the market to reach a critical point before acting—and you’ll consistently profit.'

Peter Thiel liquidates all NVIDIA shares; Intel rumored to acquire SambaNova: A final verdict—from 'brutal aesthetics' to 'cost supremacy'.

Introduction — Recently, two seemingly unrelated yet subtly interconnected pieces of news erupted simultaneously across the tech and investment communities: Capital Retreat: Peter Thiel—the Silicon Valley venture capital legend—liquidated all NVIDIA shares held by his firm Thiel Macro in Q3. Similarly, SoftBank and Bridgewater have also begun withdrawing. Industry Restructuring: Rumors suggest Intel is evaluating an acquisition of AI chip unicorn SambaNova Systems. [Image] Viewed together, these two developments reveal an industry-wide truth: The first half of the AI compute era—defined by 'brute-force scaling of GPU memory, disregarding cost in model training'—is peaking; the second half—the era of 'architectural efficiency and maximum ROI (return on investment)'—has already begun. Today, as an architect, I’d like to guide you past the surface-level headlines and delve into the underlying technological logic driving this shift. This article comprises two parts; references are shared via IT Daddy—Knowledge Base: 1. The Ceiling of Brutal Aesthetics 2. Memory Wall: A Technical Breakdown I. The Ceiling of Brutal Aesthetics The 'Bigger Is Better' era of brutal aesthetics may be hitting its ceiling. Today’s AI arms race sounds like a contest of financial muscle alone. GPT-4 and Gemini Ultra each involve trillions of parameters and cost tens of millions of dollars to train. Without a cluster of ten thousand GPUs, you seemingly lack even a seat at the table. Although NVIDIA’s Q3 earnings remain impressive, Jensen Huang himself candidly admitted that the company’s financial reports have awkwardly entered a double-bind: If the results are strong, people suspect an AI bubble; If the results are weak, people suspect AI has peaked. This reflects a deeper industry assessment. Meanwhile, the rumors surrounding SambaNova hint at an alternative AI future—one that prioritizes efficiency and modularity, and no longer relies entirely on brute-force GPU-memory stacking. Today, from an architect’s perspective, we’ll unpack five counterintuitive truths behind this technology—and examine whether this logic could truly disrupt NVIDIA’s dominance. 1. Truth #1: 'Three cobblers, with their wits combined, surpass Zhuge Liang, the master strategist' (CoE Architecture > Monolithic Large Models) The trend is shifting away from training one all-knowing, monolithic giant model toward 'Composition of Experts' (CoE).

Piet Viljoen: Sasol’s Spectacular Comeback Turns a Battered Stock into a Market Star

3 days ago

Sasol Ltd reports a significant stock rally driven by new leadership restoring investor confidence. This turnaround repositions the chemical and energy giant as a key growth opportunity for the market.

BizNews

2026-09-25

Sasol CEO Simon Baloyi Receives Stock Award, Sells Majority Shortly After

2026-09-25

Sasol CFO Walt Paul Bruns Acquires Shares, Then Sells 7,890 on Vesting

2026-09-25

Sasol CFO Sells $100,203 in Stock

2026-10-01

Sasol CFO Sells 1,798 Shares of Stock

2026-10-01

Sasol CFO Acquires and Sells Shares Following Stock Award Vesting