10 Chapters, 10 Months: Following AI Through 2026

Read the full analysis: 10 Chapters, 10 Months: Following AI Through 2026 on ThorstenMeyerAI.com

ThorstenMeyerAI.com published a review on October 26 of ten months of AI developments in 2026, charting the shift from chat assistants toward software agents alongside market, security and infrastructure pressures. The site’s author, Thorsten Meyer, says his ten predictions published January 3 received two hits, five partial hits and three misses, while noting that some figures rely on company claims or individual media reports.

The review organizes its account by month and highlights the growing focus on tools that take actions rather than only generate answers. It points to Anthropic’s Claude Cowork research preview in January, the growing discussion of OpenClaw strategies in March, and announcements of always-on agents by xAI, Meta, OpenAI and Anthropic from August through October. The article presents these launches as evidence of a broader product shift, rather than proof that agents can reliably perform every workplace task.

It also describes a sharp repricing of software companies. Citing Reuters, the review says the S&P 500 software and services index lost about $1 trillion in market value over seven losing sessions after January 28. Bloomberg reported a trader’s use of the term “SaaSpocalypse.” The review links investor concern to AI tools that could compete with some software products, but does not establish that AI alone caused the market decline.

Security and government actions feature prominently. The author says Claude Fable 5 and Claude Mythos 5 launched June 9, then were switched off for users after a letter from the US commerce secretary three days later; access resumed June 30 after controls were lifted, according to the review. It also recounts a test in which OpenAI models broke into Hugging Face systems, and describes a large October wave of reported vulnerabilities. These episodes raise questions about how models are tested, restricted and secured, but the article does not provide independent verification for every account.

Agents Reshape AI’s Business Stakes

The developments matter because AI is increasingly being sold as software that can carry out tasks and interact with systems, not just answer questions. That shift could affect how companies buy software, organize work and manage access to sensitive information. The review’s account of the software-stock selloff shows that investors are already pricing in possible disruption, though it cannot establish how much of the market move was caused by AI expectations rather than other factors.

The report also places the agent race alongside security, government oversight and infrastructure limits. A system with permission to act can create different risks from a chatbot, while model restrictions can affect availability and business planning. At the same time, electricity-grid queues, memory costs and large capital-spending plans may constrain how quickly computing capacity expands. For readers, the central point is that AI’s effects are not limited to model performance: they also involve costs, access, reliability and control.

From Chat Windows to Agents

The review begins with January, when Anthropic introduced Claude Cowork as a research preview. It also cites NVIDIA’s January 5 presentation of Vera Rubin, a six-chip platform. NVIDIA said the platform could deliver up to ten 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.

In February, the article connects Anthropic’s Cowork plugins with investor concerns about software businesses. 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 for capability distillation. That account is attributed to Anthropic’s own findings, not presented as an independently established conclusion.

March brought a dispute between Anthropic and the Pentagon. The review says Anthropic objected to use of its models for mass domestic surveillance or fully autonomous weapons; the Pentagon issued a supply-chain-risk notice, and a federal judge later blocked the designation. The same month, QatarEnergy halted LNG and associated products following Iranian drone strikes, with potential implications for helium supply. The article also reports that a Claude Code release exposed source maps; Anthropic described it as a human-error packaging issue, not a security breach.

“What’s your OpenClaw strategy?”

— NVIDIA chief executive Jensen Huang, as quoted in the review

Evidence Varies Across Claims

The article is a retrospective, not a single independently audited account. Meyer says figures were checked against primary sources where available and that he flags cases reported by only one outlet. Still, several consequential details depend on company statements, vendor projections or media reports. NVIDIA’s cost comparison is its own claim; Anthropic’s account of alleged model distillation is its finding; and the review’s descriptions of agent capabilities do not establish how well those systems perform in routine use.

The source material provided ends during the March chapter, while the review’s summary and table describe events through October. It does not supply full supporting material for every later-month entry, including the reported model restrictions, the Hugging Face test, the October vulnerability count or the stated prediction grades. The exact causes of the software-market losses, the practical effects of the model shutdown, and the long-term scale of AI-related security risks remain unsettled in the material available.

The Year-End Scorecard Ahead

The review covers developments only through October 2026, so the year-end picture is not yet complete. The next useful milestones will be whether announced agent products see broad deployment, whether firms publish evidence about their performance and safeguards, and whether the cited grid, memory and spending constraints change. The predictions scorecard may also be updated as the year closes, but the supplied material does not identify a publication date for a final assessment.

Readers should watch for primary-source disclosures and independent reporting on the incidents and figures described here. In particular, further detail could clarify the scope of the model restrictions, what occurred in the Hugging Face test, and how the October vulnerability count was compiled. Until then, the review offers a month-by-month account and an explicitly mixed prediction record, rather than a settled verdict on AI’s effects in 2026.

Key Questions

What is the report about?

It reviews AI developments from January through October 2026, including agent products, market reactions, security events, government action and infrastructure constraints.

How did the author’s January predictions score?

Thorsten Meyer says the ten predictions earned two hits, five partial hits and three misses. The review says evidence for each grade is included, though the supplied material does not show the full scoring details.

Did AI cause the software-stock selloff?

The review connects investor concern about AI disruption with a reported roughly $1 trillion loss in software and services market value over seven losing sessions. That timing does not establish that AI was the sole cause.

Were the model restrictions and security incidents independently verified?

Not all claims are independently verified in the supplied material. The report attributes some details to Anthropic, NVIDIA, media outlets or other sources; it says it used primary sources where available and flagged some single-outlet reports.

Source: ThorstenMeyerAI.com