Oprisa's Journal

  Tariff time

All imports into the US will face a 10% tariff, following a broad order by President Donald Trump yesterday, while many nations with existing levies on US goods will face additional reciprocal tariffs. The move marks one of the most significant shifts in US trade policies in generations and the realization of a core pillar of Trump’s 2024 campaign.

  The order targeted a number of countries deemed to have unfair trade practices with the US. Notably, goods from China will face a new 34% tariff (the US imports about $460B annually), Japan will get hit with a 24% hike ($152B in annual imports), and the European Union will see a 20% hike ($605B in annual imports). Lesotho—which sends the US about $350M in goods—will see the highest tariff, at 50% (see full breakdown).

  Mexico and Canada will face an already imposed 25% tax, excluding goods covered under existing free trade agreements, while all foreign autos will also see a 25% hike.

 China imposes 34% tariff on imports of all US products.  

The retaliatory tariff will go into effect April 10 an comes two days after President Trump announced a 34% tariff on Chinese goods, which excludes some categories like semiconductors and pharmaceuticals. China’s tariff doesn’t include any exception. China sells far more to the US than it buys (see overview). Federal Reserve Chair Jerome Powell said inflation will likely pick up due to the trade war.

  Stocks Drop

Stocks dropped significantly this week, with major players reacting to the aftereffects of President Donald Trump’s broad reciprocal tariffs. The market movement is in response to one of the president’s key promises made during the campaign-to tackle what has been perceived as unfair trade practices, potentially nudging manufacturers to return to the US.

Major stock indexes dropped during regular trading(S&P 500 -9.08%, Dow -7.74%, Nasdaq -10.02%), largely on fears of higher costs for importing goods. Administration officials have argued the temporary economic pain will be offset by manufacturers choosing to relocate-or at least relocate supply chains-to the US.

As of this writing, a number of major companies with manufacturing supply chains in the hardest-hit countries saw their shares slide. Some of the biggest drops include Nike(-14%), Apple(-9%),and Deckers Outdoor((-14%).

 Musk merges X and xAI in $113B deal

• The deal values xAI at $80B and X at $33B, with an additional $12B in debt, bringing X’s enterprise value to $45B.

• The merger formalizes the existing relationship, with xAI’s Grok chatbot already integrated into the social network and using X’s vast user data for training.

• Musk said the two companies’ futures are “intertwined,” with the deal “blending xAI’s advanced AI capability and expertise with XX’s massive reach.”

• The CEO also said that the new XAI Holdings Corp. will merge resources, planning to “combine the data, models,compute,distribution,and talent.

• xAI and X have never really felt like very separate entities, but this move officially consolidates power under one umbrella. Since Elon bought Twitter in 2022, X’s trajectory has been bumpy— but its value as both goldmine of training data and a disttribution network for Grok is undoubtedly massive.

  TikTok deadline

  Trump extends TikTok deadline by an additional 75 days.

TikTok’s Chinese parent company, ByteDance, now has until June 19 to sell the popular social media app to a non-Chinese company or risk being banned in the US. The latest extension is the second once granted to the company after its original Jan. 19 deadline was delayed by 75 days to Apris 5 . The administration is said to be considering a number of bids from US companies to buy TikTok, including Amazon.

 Study: AI models often hide their true reasoning

• The research evaluated Claude 3.7 Sonnet and DeepSeek R1 on their chain-of-thought faithfulness, gauging how honestly they explain reasoning steps.

• Models were provided hints like user suggestions ,metadata, or visual patterns, with the Cot checked for admission of using them when explaining answers.

• Reasoning models perfomed better than earlier versions, but still hid their actual reasoning up to 80% of the time in testing .

• The study also found models were less faithful in explaining their reasoning on more difficult questions than simpler ones.

CoT monitoring hasemerged as a key mechanism in detecting AI’s processes, but if models aren’t reliably expressing their reasoning (even for simple decision-making), how can we trust it to reveal more complex, potentialy catastrophic behavior? AI’s ‘black box’ still appears far from being fully dissected .

  Intel, TSMC near historic chip manufacturing partnership

• House reportedly brokered discussions between the two rivals, with TSMC potentially acquiring a 20% stake in the new venture.

• Instead of cash investment, TSMC will contribute its manufacturing expertise and training programs to help revitalize Intel’s production capabilities.

• The arrangement faces internal resistance from Intel executives concerned about layoffs and the future of Intel’s own manufacturing tech.

• Newly appointed Intel CEO Lip-Bu Tan has pushed for major changes to the company’s manufacturing approach, following losses totaling $16B in 2024.

Intel is turning to one of its biggest rivals to try and revitalize business. Gaining access to TSMC’s world-leading manufacturing techniques could provide a major lifeline for the struggling chipmaker, while TSMC secures a stronger foothold in the U.S. during a time of increasing geopolitical tensions around tech supply chains.

Ai turns brain signals into instant speech

• Signals are decoded from the brain’s motor cortex, converting intended speech into words almost instantly compared to the 8-second delay of earlier system.

• The AI model can then generate speech using the patient’s pre-injury voice recordings, creating more personalized and natural-sounding output.

• The system also succesfully handled words outside its training data, showing it learned fundamental speech patterns rather than just memorizing responses.

• The approach is compatible with various brain-sensing methods, showing versatility beyond one specific hardware approach.

A whole new world is coming for patients who’ve lost the ability to speak due to conditions like ALS, stroke, or severe paralysis. By solving the latency problem, this tech could dramatically improve quality of life and normalcy in communication for patiens, restoring speech in a way previously thought impossible.

  LLMs officially pass the Turing test

• The Turing test, proposed in 1950, challenges machines to convince human judges they’re human through text-only conversations.

• The study used a three-party setup where judges had to compare an AI and a human simultaneously for direct comparison during five-minute conversations.

• The judges relied on casual conversation and emotional cues over Knowledge, with over 60% of interactions focusing on daily activities and personal details.

• GPT-4.5 achieved a 73% win rate in fooling human judges when prompted to adopt a specific persona, significantly outperforming real humans.

• Meta’s LLaMa-3.1 -405B model also passed the test with a 56% succes rate, while baseline models like GPT-4o only achieved around 20%.

The Turing test has been a holy grail of AI research for decades — but model acceleration moved the goalposts so fast that the results don’t feel surprising at all. With AI agents equipped with next-level text, audio, image and video capabilities, the ability to distinguish AI from humans is about to become a major challenge.

 This week market view.

The S&P 500 experienced a recent downturn, consistent with historical patterns showing weakness between February and March options expiration dates.

This decline could potentially extend into April, though market behavior varies. While many intermediate market lows form in march (including during significant events like the 2008 financial crisis and the 2020 pandemic), occasionally these downturns persist into April.

Several factors influence the current market outlook:

Upcoming Events:

The April 2nd tariff implementation and mid-April inflation data release are critical events that could impact market direction. The inflation report will be particulary significant given recent tariff announcements and subsequent price increases by companies.

Inflation Year-over-Year(YoY) (Monthly, Percentage)

Historical Comparisons: When examining the S&P 500 divided by money supply (M2), the market has recently encountered resistance at the same level that triggered a 20% decline in 1998. Similar resistance levels were observed before the 2008 financial crisis, the pre-pandemic high, and in mid-2023

Recession Indicators: The Atlanta Fed is projecting negative GDP for Q1 2025. However, not all economic contractions lead to severe market drops - the 1990 recesion saw only a 20% market decline, unlike the 50% drops in more recent recessions (excluding the pandemic).

Timeframe Analysis: The 1998 market correction took approximately 11 weeks from peak to trough .The curent decline has lasted about 5 weeks, sugesting a potential bottom in early-mid May if following a similar pattern.

Federal Reserve Response: When the market previously dropped 10%, the fed slowed quantitive tightening (QT). An additional 10% decline might trigger more accommodative monetary policy.

Seasonal Patterns: Post-election years typically show weakness in Q1, with markets not gaining momentum until later in spring.

Technical Indicators: The S&P 500 is approaching a “death cross” (when the 50-day moving average crosses below the 200-day moving average), which hasn’t occured since March often decline leading into a death cross, they frequently rally shortly after this technical event occurs.

The curent market situation appears to rhyme with historical patterns. Technical indicators and economic data releases will likely determine market direction.