The Japanese yen rose broadly in Asian trading on Monday at the start of the week against a basket of global currencies, extending its gains for a second consecutive day against the US dollar and hitting a two-month high. The move was supported by continued selling pressure on the US currency, as well as growing speculation about possible intervention by US and Japanese monetary authorities in the foreign exchange market.

The Federal Reserve Bank of New York conducted a review of the dollar/yen exchange rate with market participants, a step widely viewed as a strong signal of potential intervention, amid ongoing and intensified coordination between US and Japanese authorities to address sharp market volatility.

Price overview

• Japanese yen exchange rate today: The dollar fell 1.25% against the yen to 153.81, its lowest level since last November, from Friday’s close at 155.74. The dollar recorded an intraday high at 155.34.

• The yen ended Friday’s session up 1.65% against the dollar, marking the first loss for the US currency in three days and the yen’s largest daily gain since last August, driven by rising speculation over foreign exchange intervention.

• The yen gained 1.5% against the dollar last week, recording its first weekly gain in a month, supported by an acceleration in the unwinding of yen carry trades.

US dollar

The US dollar index fell by more than 0.5% on Monday, extending its losses for a third consecutive session and hitting a four-month low at 96.95 points, reflecting continued weakness in the US currency against a basket of major and secondary currencies.

The decline comes alongside an acceleration in dollar selling, amid rising concerns over possible intervention by monetary authorities in both the United States and Japan to curb volatility and stabilize price movements.

This is in addition to increasing political and economic risks in the United States, accompanied by declining confidence in dollar-denominated assets and a widening state of uncertainty across global markets.

Joint US–Japan intervention

Sources told Reuters that the Federal Reserve Bank of New York had reviewed dollar/yen exchange rate levels with market participants, a move seen as a strong signal of potential intervention, amid continued and intensive coordination between US and Japanese authorities to counter sharp market fluctuations.

Senior Japanese officials, including the finance minister and top diplomats, confirmed on Monday that they are in “close coordination” with the United States on foreign exchange issues, based on a joint statement issued in September 2025.

Japanese Prime Minister Sanae Takaichi warned that the government would “take necessary action” against any abnormal or speculative movements in the market.

Japanese interest rates

• In line with most market expectations, the Bank of Japan left its benchmark interest rate unchanged on Friday at 0.75%, the highest level since 1995.

• The decision to hold rates was approved by an 8–1 vote, with one board member calling for a 25-basis-point hike to 1.0%. The bank opted to pause in order to assess the impact of the rate increase implemented in December 2025.

• The Bank of Japan raised its economic growth and inflation forecasts for the fiscal year ending March 2026, signaling readiness to continue tightening monetary policy and gradually raise borrowing costs.

• Bank of Japan Governor Kazuo Ueda said the central bank would continue to raise interest rates if economic conditions and prices evolve in line with expectations, stressing the importance of inflation trends in policy decisions.

• Market pricing for a 25-basis-point rate hike at the Bank of Japan’s March meeting remains below 20%.

• Expectations for a 25-basis-point hike at the April meeting have risen to above 50%.

• To reassess these expectations, investors are awaiting further data on inflation, employment, and wage growth in Japan.

Gold and silver prices rose sharply during Friday’s trading, as escalating geopolitical tensions and market uncertainty pushed investors toward safe-haven assets, driving both precious metals to unprecedented record highs.

The gains came amid ongoing disputes between the United States and NATO over Greenland, as well as growing concerns surrounding the independence of the Federal Reserve.

Separately, media reports indicated that the administration of US President Donald Trump is considering a plan to impose a naval blockade around Cuba in an effort to control its oil flows.

In trading activity, February gold futures settled up 1.35%, or $66.3, at $4,979.70 per ounce, marking the sixth record close in 2026. The precious metal also posted a weekly gain of 8.4%, its strongest weekly performance since the onset of the pandemic crisis in 2020.

Meanwhile, March silver futures surged 5.2% to $101.33 per ounce, closing above the $100 level for the first time ever and recording weekly gains of 14.45%.

“AGI is here… now.” With that phrase, Sequoia Capital announced this week — one of Silicon Valley’s most storied venture capital firms and a major investor in OpenAI — that we have crossed the threshold into artificial general intelligence (AGI).

In its post, the firm stated, plainly and explicitly, that it was “not bogged down by details at all.” When Sequoia speaks, the tech world listens. The claim dominated discussions across the AI developer community for days.

As someone who is simultaneously a developer, a venture capitalist, and an AI researcher, I see this declaration as deeply useful in one sense — and deeply dangerous in another.

What Is Useful About Sequoia’s Argument?

Sequoia offers a practical definition of AGI: “the ability to discover solutions. Nothing more.” Under this framing, AI systems today can search vast bodies of information, determine a course of action, and then execute it. The core shift, according to Sequoia, is that AI has moved from “talking” to “doing.”

The firm points to concrete examples. It says platforms like Harvey and Legora “act as legal associates,” Juicebox “acts as a recruiter,” and OpenEvidence’s Deep Consult “acts as a specialist.” These are literal descriptions. While I am skeptical of this conceptual framing — more on that shortly — the provocation itself matters.

What Sequoia is doing here is directly challenging developers, and that is important. AI systems can already review contracts clause by clause and engage meaningfully with prospective customers in real time. This is a reminder that we need to think bigger about what is now possible, and that the frontier has expanded dramatically in just a single year.

I sent Sequoia’s post to my co-founders not to debate philosophy, but to push us to rethink the “execution versus conversation” framework it proposed. We need to rise to that challenge.

But Why Is Calling These Systems AGI Dangerous?

Labeling these systems as “artificial general intelligence” causes real harm — both to the credibility of the AI revolution and to the safe deployment of these technologies. It obscures what so-called AI agents can actually do today — and they are certainly not general superintelligence — while offering no guidance on how humans should interact with them. The short answer: do not trust them blindly.

Three examples illustrate these limitations clearly.

First: AI Systems Fail Outside Their Training Distribution

I addressed this in a previous article, but the Greenland crisis provides a live, evolving example. I tested whether generative AI tools — including ChatGPT 5.2 with maximum “reasoning and research” enabled — could analyze this rapidly developing geopolitical event. If these systems are truly AGI, could they help me understand what was happening?

The answer was no. They could not even conceive that the events were possible.

I presented screenshots from Wikipedia documenting the crisis. Every model told me the story was fabricated, “nonsense,” and impossible. When I continued pressing, citing real news sources, ChatGPT repeatedly urged me to “calm down,” insisting that “this is not a real crisis.”

These models are so tightly anchored to traditional Western alliance frameworks that they cannot generate context that contradicts their training data — even when confronted with primary sources. When reality moves outside their training distribution, AI “reasoning” collapses. Instead of expressing uncertainty, the system confidently misleads the user and continues reasoning while wrong. If policymakers or politicians are relying on these tools to understand Greenland right now, that is a genuine risk.

Second: AI Systems Reflect the Beliefs of Their Builders

A study published in Nature two weeks ago made this explicit. Researchers found that large language models reflect the political ideologies of their developers. Chinese models were strongly positive toward China, while Western models were clearly negative.

Even within Western models, bias is evident. Grok, developed by Elon Musk’s xAI, showed negative bias toward the European Union and multiculturalism, reflecting a right-leaning agenda. Google’s Gemini, widely seen as more liberal, was more positive toward both.

This is now widely accepted within the AI community: language models reflect the ideology of the labs that build them. So how can we trust that an “agent” with a supposed blank slate can neutrally “discover solutions,” especially when analyzing complex, large-scale data?

Declaring the existence of AGI implicitly assumes neutrality — or at least gestures toward it — while the evidence points in the opposite direction.

Third: Deterministic Systems Versus Non-Deterministic Systems

Generative AI is inherently non-deterministic. The same input can produce slightly different outputs, or radically different ones.

Humans intuitively understand what should be deterministic and what can be creative. Your shirt size when ordering online is deterministic; choosing a pattern or color is subjective. Even the most advanced models still confuse these categories constantly. We have all seen generative AI treat hard facts as if they were creative suggestions.

This exposes a critical gap in metacognition — awareness of the thinking process itself. Without the ability to distinguish between what must be fixed and what can be generative, AI cannot reliably “discover solutions.”

So What Should We Do?

We have clear tools available.

First, choose narrow, well-defined use cases where bias and out-of-distribution failure are less likely.

Second, provide AI systems with full, customized, real-world context rather than letting agents operate in a vacuum. As I have written before, context is king for AI agents. It also clarifies what must be deterministic and what can be generative.

Third, deploy rule-based filters and supervisory agents that trigger human review when necessary.

Finally, we must acknowledge a core reality: large language models will always reflect their training data and the ideologies of their creators. These models — and their developers — are political actors, whether they intend to be or not. AI should therefore remain under the control of individual human users, not imposed on people as an opaque system. Traceability and accountability are essential — the ability to trace every decision back to a human, no matter how many intermediate steps exist — to ensure governance and safety.

Ultimately, I do not care much what we call these technologies — as long as we do not call them AGI. What we have today is extraordinarily powerful AI, capable of speaking and executing effectively within narrow, well-defined domains. With strict safeguards, deterministic filters, and human-in-the-loop systems, these tools can add trillions of dollars to the global economy.

Call it narrow AI. That is where the trillion-dollar opportunity actually lies today.

US stocks fell on Friday, putting Wall Street’s main indexes on track for a second consecutive weekly loss, as shares of Intel slumped sharply following weak guidance, while ongoing geopolitical tensions continued to weigh on investor risk appetite.

Stocks had rebounded over the previous two sessions after a sharp sell-off on Tuesday, triggered by threats from US President Donald Trump to impose tariffs on European allies unless Washington was allowed to purchase Greenland.

Trump later softened his rhetoric on tariffs and ruled out using force to take control of Greenland. Even so, the S&P 500, the Nasdaq, and the Dow Jones Industrial Average remained on course to end the week lower. At the same time, flows into safe-haven assets persisted, pushing gold prices to a new record high.

The biggest drag on markets on Friday came from chipmaker Intel, whose shares plunged 14.9% after the company forecast quarterly revenue and earnings below market expectations, citing difficulties in meeting demand for server chips used in artificial intelligence data centers. Despite the sharp drop, Intel shares were still up about 50% since the start of the year.

The Philadelphia Semiconductor Index fell 1.6%, pulling back from the record high reached in the previous session, while Wall Street’s volatility index, the VIX, known as the market’s fear gauge, rose after declining over the prior two sessions.

Peter Cardillo, chief economist at Spartan Capital Securities, said: “Earnings season has been good, but one or two stocks have issued less optimistic guidance and sold off accordingly as investors reposition. Guidance has now become more important than ever.”

He added: “Investors will remain cautious because we’re not just watching earnings, we’re also focused on the Federal Reserve. We don’t expect a policy change, but the question is what the Fed will say in its statement.”

By 9:48 a.m. Eastern Time, the Dow Jones Industrial Average was down 320.71 points, or 0.65%, at 49,063.30. The S&P 500 fell 14.68 points, or 0.21%, to 6,898.78, while the Nasdaq Composite slipped 36.50 points, or 0.16%, to 23,399.52.

Anticipation of the Federal Reserve decision

The Federal Reserve is widely expected to keep interest rates unchanged in the 3.5% to 3.75% range at its meeting next week. Investors will scrutinize the policy statement and comments from Chair Jerome Powell for clues about the next move. According to the CME FedWatch Tool, markets are pricing in the first rate cut in June.

Preliminary data from S&P Global showed US business activity remained steady in January, as an improvement in new orders offset weakness in the labor market.

Several members of the “Magnificent Seven,” including Apple, Tesla, and Microsoft, are set to report earnings next week. Their outlooks will be closely watched to assess whether the growth narratives supporting their elevated valuations remain intact.

Supported by the strength of the US economy and expectations for interest rate cuts later this year, market gains had broadened beyond mega-cap stocks into other sectors. Both the Russell 2000 small-cap index and the Dow Jones Transportation Average hit record highs on Thursday.

In other moves, shares of Nvidia rose 1.4% after Bloomberg reported that Chinese officials told companies including Alibaba, Tencent, and ByteDance to prepare for potential purchases of Nvidia’s H200 AI chips.

US-listed mining stocks such as Hecla Mining and Coeur Mining also edged higher by 0.6% and 0.3%, respectively, as silver prices climbed to record levels and approached the $100-per-ounce mark for the first time.