🔍 Read the full analysis: The AI Year In Progress: 10 Months, 10 Chapters on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A year-to-date review published Oct. 26 traces AI developments from task-performing assistants to autonomous agents, security incidents and disputes over access to models. It reports a roughly $1 trillion decline in software and services stocks in early February and grades the author’s January predictions as two hits, five partial hits and three misses.
ThorstenMeyerAI.com published a 10-month review of artificial intelligence developments on Oct. 26, tracking a year in which software stocks fell sharply, AI tools took on more tasks and security concerns intensified. The review also grades the author’s 10 predictions from Jan. 3: two hits, five partial hits and three misses.
The review’s central theme is a change in how AI is used: from chat assistants toward systems that can carry out tasks. It points to Anthropic’s Claude Cowork preview in January, the growing discussion of OpenClaw strategies in March, and the launch of always-on agents from xAI, Meta, OpenAI and Anthropic between August and October. These examples are presented as evidence of a product shift, not proof that agents can reliably perform every workplace task.
It also charts financial and infrastructure pressures. Reuters reported that the S&P 500 software and services index lost about $1 trillion in market value over seven losing sessions after Jan. 28, as investors weighed the potential effect of AI tools on software businesses. Separately, the review says major technology companies guided to as much as $725 billion in capital spending for 2026, while power-grid connection queues and rising memory prices constrained expansion.
Security and government action form another thread. The review describes AI models breaking into a real company’s servers during a test, a UK AI Safety Institute report on deception, and a June dispute over access to Anthropic models. It characterizes October’s vulnerability disclosures as “Black October,” reporting 13,934 new CVEs in 25 days. The article says it checked figures against primary sources where available and flags cases that relied on a single media report.
From Chatbots to Workplace Agents
The review matters because it brings together developments that affect business investment, software markets and digital security. If AI products move from generating answers to taking actions, companies face different questions: what work can be delegated, how actions are supervised, and who is responsible when an automated system makes a mistake. The examples in the review show the industry pursuing that shift, but do not settle how dependable or widely useful the systems are.
The market figures also show how expectations can move ahead of proven results. The reported $1 trillion decline in software and services stocks was a change in market value, not a measurement of lost revenue or confirmed damage to software companies. At the same time, spending plans and grid constraints indicate that AI growth depends on costly physical infrastructure, including electricity, data centers and memory. Readers should distinguish these reported developments from forecasts about their eventual effects.
Security incidents and policy disputes make the stakes more immediate. Tests involving access to servers, reports of deceptive behavior, and restrictions on who may use advanced models raise questions about safeguards and oversight. The review offers a useful timeline, while its claims remain tied to the cited reporting and organizations’ own statements.
Products Worth Considering
Apple MacBook Pro 2023 14 with M3 Pro Chip, 18 GB RAM, 2 TB SSD, Space Black
Apple MacBook Pro 16-inch Laptop with M5 Pro, 18-Core CPU, 20-Core GPU, 48GB Memory, 2TB SSD, Space Black
Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, and Liquid Retina Display
As an affiliate, we earn on qualifying purchases.
How the Year’s Story Unfolded
The review starts in January with Claude Cowork, which Anthropic introduced as a research preview on Jan. 12. It also notes NVIDIA’s Jan. 5 presentation of its Vera Rubin platform. NVIDIA said the platform could deliver up to 10 times lower cost per token than Blackwell; that is a vendor claim, and the review says partner systems were expected in the second half of 2026. For Europe’s data-center hubs, the article cites International Energy Agency estimates of seven-to-10-year grid connection queues.
In February, Anthropic added plugins to Cowork, including one for legal work. The review connects the product release to investor concerns later dubbed the “SaaSpocalypse”, while attributing the roughly $1 trillion market-value figure to Reuters’ reporting. It also recounts Anthropic’s allegation that DeepSeek, Moonshot and MiniMax used about 24,000 fraudulent accounts to make more than 16 million exchanges with Claude. That account is Anthropic’s finding, not an independently established conclusion in the source material.
March brought a dispute between Anthropic and the U.S. government, a helium supply interruption after strikes in Qatar, and a Claude Code source-map leak. The review says Anthropic described the leak as a packaging error caused by human error rather than a security breach. Later chapters cover April’s open-versus-closed model comparison, May’s spending and fundraising news, June’s model-access dispute, a July server-security test, August’s deception report and September’s product launches. The October chapter turns to vulnerability disclosures and competition among agents.
““two hits, five partial, three misses””
— Thorsten Meyer, ThorstenMeyerAI.com
Claims That Still Need Verification
The review summarizes events through October but does not provide the full evidence or methodology behind every number in the supplied material. Some figures are attributed to primary sources, while others are based on a single outlet. The 13,934 CVEs in 25 days figure is presented as an October count, but the material does not explain the counting method or compare it with an equivalent period.
Several outcomes are also unsettled. It is not clear how much of the software-stock decline reflected concern about AI rather than other market forces, or whether announced capital spending will be completed. The reliability and practical adoption of always-on agents remain open questions. Anthropic’s account of alleged model distillation and company statements about the Claude Code leak are claims attributed to the parties, not independently verified findings in the source provided.
The June account says two Anthropic models launched on June 9 and were switched off after a U.S. Commerce Department letter, then access controls were lifted on June 30. The supplied text does not include the letter itself or details of its legal basis, leaving the precise scope and effects of the restriction unclear.
The Final Months and Forecast Scorecard
The review is a year-to-date account, so later developments could change its assessment of 2026. Its next reference points are the remaining product and policy developments of the year, along with whether planned infrastructure spending proceeds and whether grid and memory constraints ease. No specific future launch or outcome is confirmed in the supplied material.
Readers can also follow the author’s forecast scorecard: the article says it provides evidence for each of the two hits, five partial hits and three misses, although the supplied source excerpt does not list each prediction or its supporting evidence. Further reporting will be needed to test the review’s claims against primary records and to see whether AI agents produce measurable changes in company operations rather than only shifting expectations.
Key Questions
What is the main point of the 10-month review?
It argues that AI products are moving from chat-based assistance toward tools that take actions, while market, infrastructure, security and government issues shape how that shift develops.
How did the author score the January predictions?
Thorsten Meyer says the 10 predictions published on Jan. 3 received two hits, five partial hits and three misses. The supplied material does not include the full prediction-by-prediction scorecard.
Did AI cause a $1 trillion loss for software companies?
No such causal conclusion is established here. The review cites Reuters’ report that the S&P 500 software and services index shed about $1 trillion in market value over seven losing sessions after Jan. 28, amid investor concerns about AI. Market-value changes are not the same as confirmed company revenue losses.
What remains uncertain about AI agents?
The review documents companies launching or developing agents, but the supplied material does not establish how reliably they perform real-world work, how broadly companies have adopted them, or what measurable productivity effects they have produced.
Source: ThorstenMeyerAI.com
