Automatically translated. Read the Chinese original.
The conclusion is first: after four rounds of testing, both models have their strengths and weaknesses in "speaking human language". Claude Opus 5.5 is better in terms of sense of proportion and "non-fabrication" of SMS messages; Longer articles in Gemini 4 Argon are smoother, but the default writing style is heavy. Neither of them escaped the AI-driven tone when it came to writing long articles.
Introduction: The Wall Street Journal recently revealed a confidential draft of Anthropic’s S-1 prospectus, with a headline that is nothing short of staggering:The net loss for the first half of 2026 amounted to as much as US$80.7 billion. The public sphere was instantly in an uproar, with cries of ‘AI money-guzzlers running amok’ and ‘a massive bubble’. However, if you really look beneath the surface of the financial statements, you will arrive at a startling conclusion that is the exact opposite:Far from being a bad thing, this massive loss is, in fact, a direct consequence of the company’s immense popularity in the primary market; and this company, which on the surface appears to have ‘lost 80 billion in half a year’, has in fact just passed its profitability inflection point and is now enjoying the healthiest period in its commercial history.
In 2026, the Reserve Bank of Australia (RBA) raised the cash rate from 3.60 per cent to 4.60 per cent, returning it to its highest level since 2011. How will this round of monetary tightening affect Australian households, businesses and tenants? With the interest rate differential between China and Australia now standing at 3.6 percentage points, where is the AUD/CNY exchange rate headed? This article brings together an analysis of the RBA’s rate-hiking cycle, macroeconomic and social impacts, global interest rate differentials and multi-scenario analysis of the AUD/CNY exchange rate, concluding with recommended timeframes for currency exchange and key indicators to monitor.
Around 17 October 2026, large-scale street protests erupted almost simultaneously in several countries, including France, Spain, Italy, Portugal and Belgium, with a partial curfew imposed in certain areas of Paris. Was this simultaneous wave of protests across multiple countries the result of ‘contagion’ via social media, or did everyone receive the same bill? This article examines the September inflation figures for each country, the European Central Bank’s interest rate hikes and fiscal space, to analyse the common causes of this wave of protests, the cases of Spain and France, and possible scenarios for how the situation might unfold this winter.
This article was written on 21 October 2026; the data is current as at 20 October 2026. The dates of the two Federal Reserve meetings mentioned in the article—27–28 October and 8–9 December—are subject to the calendar in effect at the time of publication.
Let’s look at three things first.
Firstly: Brent crude is currently trading at US$109 per barrel. Following military action against Iran by the US and Israel at the start of the year, oil prices have risen steadily; in September, US petrol prices rose by 34.2 per cent year-on-year.
Secondly: US non-farm payrolls rose by just 29,000 in September. By comparison, the figure for August was 133,000. The unemployment rate, meanwhile, rose to 4.2 per cent.
Thirdly: The Federal Reserve raised interest rates just last month, bringing the federal funds rate to 3.75 per cent–4.00 per cent, and the market is still betting on another hike in December, with the probability having been priced in at as high as 94 per cent at its peak.
With these three things happening at the same time, something just doesn’t seem quite right. Prices are rising, jobs are becoming scarcer, and the central bank, far from rescuing the economy, is actually slamming on the brakes?
If you find this unreasonable, congratulations—your intuition has picked up on a genuine contradiction. This article explains exactly why this contradiction exists, why the Federal Reserve finds itself in a bind, and how it relates to your and my mortgage, petrol costs and savings.
I liked Opus 5.5 at first. A stronger model at a lower price was welcome news. I subscribe to both Claude Pro and ChatGPT Plus and use Claude Code and Codex, so I had reason to hope that competition would eventually benefit users.
But after looking at my own usage records and AA's tests, then reading Fuli Luo's post on inference cost reductions alongside Anthropic's capacity expansion and incident reports, I have become increasingly unwilling to take the "stronger and cheaper" story at face value.
I am now more inclined to see this as price competition ahead of an IPO: cutting prices to win users and demonstrate growth while still dealing with expensive compute capacity and service delivery problems. That is my interpretation; I do not have access to Anthropic's internal decision-making documents. I want to know how much of the engineering economics behind these cuts has been explained, and how much of the growth investors welcome will translate into benefits for users.
My Laptop is a very ordinary gaming Laptop: a Ryzen 7 8845H, 16GB of RAM, an RTX 4060 Laptop, and only 8GB of video memory. According to many people's impression, this configuration can only be used to "play around" with local large models.
But now, when I double-click a startup file every day, the browser opens a chat page: I can view images, watch videos, switch the depth of my thinking, and call native tools. My text generation speed is approximately... 31 tokens/sThe entire process involves no internet connection and no API fees. The one running inside is Qwen3.5-9BThe reasoning framework is llama.cpp。
This article is divided into three parts: How I chose the framework and model; An installation tutorial for beginners, written for every command; And the pitfalls I've encountered and the corresponding solutions.
Yesterday, Google officially announced Gemini 4 Argon。
Strictly speaking, this is not a regular release that "everyone can use it now". Argon is being handed over to a group of trusted cybersecurity defenders through the Fairwind Program, and Google is also participating in the US government's pre-release model access process. Paid API users and Google AI Ultra users will be next in line, but Google has not given a specific date.
However, if we only understand it as a cautious grayscale launch, we would underestimate this update. From Google's announced design goals and internal use cases to the third-party testing that has already emerged, Argon is trying to answer a question more important than "Is chatting smarter?":
Can a cutting-edge model work continuously for tens of minutes, hours, or even longer, identifying problems, calling tools, reflecting, modifying, and actually getting things done?
This is exactly what I've always loved about Gemini. Rather than simply pursuing a sense of amazement in conversation, I prefer Google's direction of integrating multimodality, long context, tool calls, and real workflows into a complete system. Argon may not be number one in every test yet, but it is currently the next-generation Gemini that most closely resembles the final form of this approach.
