🔍 Read the full analysis: Will AI’s Growth Accelerate In 2027? Six Clues To Follow on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
ThorstenMeyerAI.com published a watchlist on Nov. 18, 2026, for six signals that could indicate whether AI progress accelerates in 2027. The piece describes a scenario, not a forecast: current evidence is mixed, key thresholds are the author’s judgment calls, and several AI-specific measures have no scheduled updates.
The central scenario is that AI systems could take on more of the work involved in building future AI systems, potentially speeding development before constraints such as chips, data-centre capacity and electricity limit the cycle. The tracker does not repeat the full case for that outcome. Instead, it identifies what to measure, who publishes the data, when figures may arrive and what changes would support or weaken the scenario.
The largest gap highlighted by the source is between current capability measurements and the proposed pace of acceleration. METR’s last public record for an AI system’s 80% task time horizon was 185.9 minutes, or about three hours and six minutes, for an early Claude Mythos Preview in April 2026. The tracker’s threshold for the scenario to gain support is a measured horizon of 16 hours or more by mid-2027—about five times that record. METR has cautioned that its existing task suite cannot reliably measure results above 16 hours.
On AI involvement in AI research, Anthropic reported that Claude “leads” 26% of its AI research and development work, as of August 2026. In the source’s description, “leads” means completing most of a task end to end while a human supervises; it does not mean the work is fully autonomous. The tracker also says OpenAI’s GPT-6.2 Astra was its first released model to reach the company’s “High” threshold for AI self-improvement. Those assessments are lab-reported measures, and the companies have not set regular publication schedules.
The gap between capability and acceleration
The watchlist matters because its signals test different parts of the same proposed feedback loop: whether AI can handle longer tasks, contribute more to AI research, and operate in a growing infrastructure base. A sharp change in one measure would not establish that AI progress has accelerated; the source says the scenario would be more persuasive if several indicators moved together.
The infrastructure figures point to both expansion and limits. US data-centre construction was running at an annual rate of $96.5 billion in September 2026, compared with $50.7 billion a year earlier. At the same time, the source cites expected US power-generation growth of 2.7% in 2026 and 1.2% in 2027. Those figures do not by themselves show how much electricity will be available to AI facilities, but they illustrate why building capacity and power supply are tracked separately.
For readers, the distinction is practical: the scenario is not a prediction that AI will reach a particular capability by a set date. It is a framework for checking whether new measurements support the proposed acceleration or point instead to slower progress and physical constraints.
Products Worth Considering

DJI Mini 3 Drone with DJI RC Controller

Amazon Kindle Scribe Colorsoft 64GB with Premium Pen

Apple AirPods Pro (2nd Generation) Wireless Earbuds with MagSafe Charging Case (Lightning)
As an affiliate, we earn on qualifying purchases.
What the six indicators track
The source groups the watchlist around AI capability and research automation, along with the physical and economic conditions that could support or restrain further growth. The six indicators named in its accompanying scenario analysis are time horizons on longer task suites; AI’s share of AI research; compute and data-centre capacity; electricity supply; industrial robots; and labour-market effects.
For the first indicator, METR defines a time horizon by the length of a task, measured against the time a human expert needs, at which an AI agent is predicted to succeed at a specified reliability level. METR’s public results page was last updated Sept. 8, 2026, and the source says no date has been given for a new public measurement. METR’s February-to-March 2026 Frontier Risk Report judged the labs’ internal frontier “likely” to have an 80% horizon of at least three but under four hours.
The source also cites 603,000 industrial robots installed worldwide in 2025 and a forecast of 655,000 installations in 2026. Separately, it reports that 2% of US firms said they had made AI-related job cuts. These numbers describe different measures and time periods; the tracker does not treat either as proof that automation has broadly displaced workers.
The article was published as a scenario tracker and opinion piece, not an independent forecast. Thorsten Meyer disclosed that the article was researched and drafted with AI assistance, including Anthropic’s Claude. He also said that Anthropic, OpenAI and METR did not review or sponsor it, although their evidence is central to some of the indicators.
“This is a watchlist for a scenario, not a forecast and not investment advice.”
— ThorstenMeyerAI.com
Key measurements remain incomplete
There is no scheduled release date for the next public METR time-horizon measurement, and the public page is described as no longer actively updated. The source says the existing suite cannot reliably measure above 16 hours, while some results may be sensitive to how attempted cheating is scored. It reports that METR did not regard the cited GPT-5.6 Sol results as robust measurements.
Regular publication schedules are also absent for Anthropic’s R&D automation measure and OpenAI’s model assessments. The tracker’s thresholds are not official cut-offs, and its source says no publisher defines a specific line for compute or labour; those series are to be read for direction. It is also unclear how quickly new data-centre construction can translate into usable power and computing capacity.
The source’s available material gives selected current readings and threshold descriptions, but it does not establish that the scenario is happening. Results from different organizations and measures may not be directly comparable, and the article’s figures reflect information available on Nov. 18, 2026.
The next scheduled data releases
The nearest dates identified by the source are Dec. 1 and Dec. 8, 2026, for the next construction and power figures, respectively. Those releases can update the infrastructure side of the tracker, but they will not answer whether AI agents are taking on longer tasks or more AI research.
For those AI-specific indicators, the next public readings depend on METR and the labs choosing to publish. ThorstenMeyerAI.com says it plans to update its tracker with dated notes as new figures arrive. The key test it sets for the capability scenario is whether a reliable 80% time horizon of at least 16 hours is measured by mid-2027, alongside movement in the other signals.
Key Questions
Is the article predicting that AI growth will accelerate in 2027?
No. ThorstenMeyerAI.com calls it a watchlist for a scenario, not a forecast. Its thresholds are the author’s judgment calls.
What is the main capability threshold in the tracker?
The tracker says the scenario gains support if an AI agent’s measured 80% task time horizon reaches at least 16 hours by mid-2027. METR has cautioned that its current task suite cannot reliably measure above that range.
Does Anthropic’s 26% figure mean AI runs its research independently?
No. The source says Claude “leads” 26% of Anthropic’s AI research and development work, with a human supervising. It also says no measured part runs fully autonomously.
When are the next figures expected?
The source identifies Dec. 1 and Dec. 8, 2026, for construction and power data. It gives no publication date for the next public METR or lab-specific AI automation readings.
What would make the scenario more convincing?
The tracker says no single indicator settles the question. Support would come from several measures moving in the same direction, including longer reliable AI task horizons, greater AI involvement in AI research, and changes in infrastructure capacity.
Source: ThorstenMeyerAI.com
