AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

A November 18, 2026, article sets out six signals to track whether AI progress could accelerate sharply in 2027, from task performance and lab automation to data centers, electricity, robots and jobs. Its author calls the thresholds judgment calls for a scenario tracker, not a forecast; current evidence is mixed and several key measurements have no scheduled update.

Thorsten Meyer AI published a six-signal tracker on November 18 for assessing whether AI progress could accelerate in 2027, including measures of model capabilities, AI-assisted research, computing infrastructure, electricity, robotics and employment. The author says the tracker monitors a scenario rather than making a forecast or investment recommendation, and sets its thresholds as judgment calls.

The tracker’s largest gap is between current measured AI performance and the pace implied by the scenario. The article reports that METR estimated an 80% task time horizon of 185.9 minutes for an early Claude Mythos Preview model from April 2026. The scenario’s threshold is 16 hours or more by mid-2027. METR’s current task suite, the article says, cannot reliably measure beyond that point, and the organization’s time-horizon page is no longer actively updated.

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

For automation inside AI labs, the article cites Anthropic’s August 2026 estimate that Claude “leads” 26% of the company’s AI research and development work, with no measured work fully autonomous. It also cites OpenAI’s November 17 statement that GPT-6.2 Astra reached the company’s “High” threshold for AI self-improvement. These are company-specific measures and claims; the tracker says no schedule has been stated for updated readings.

Infrastructure and deployment readings offer a different picture. US data-center construction spending was running at an annual rate of $96.5 billion in September 2026, compared with $50.7 billion a year earlier, according to figures attributed to the Census Bureau. The article cites a forecast for US power generation growth of 2.7% in 2026 and 1.2% in 2027. It also reports 603,000 industrial robots installed in 2025, with a 655,000 forecast for 2026, and says 2% of US firms reported AI-related job cuts in a Census Bureau survey spanning November 2025 to January 2026.

At a glance
reportWhen: Published November 18, 2026; next sched…
The developmentThorsten Meyer AI published a six-signal watchlist on November 18, 2026, to track whether its rapid-AI-progress scenario for 2027 is gaining support.
Log-scale chart titled The gap: about five times today's record. METR's measured 80% time horizons of record-setting AI models rise from 0.89 minutes for GPT-4 in March 2023 to 185.9 minutes for an early Claude Mythos Preview in April 2026, the last public record. A dashed line marks 16 hours; a highlighted target marks our gains line of 16 hours by mid-2027. A dotted line marks today, November 18, 2026, with no newer public record
Signal 1: METR's 80% time horizon records and our gains line. Source: METR, Time Horizon 1.1 results file and time horizons page (last updated September 8, 2026). Chart: Thorsten Meyer AI.

The Signals Test Different Bottlenecks

The tracker matters because it separates rapid software progress from the physical limits that could constrain deployment. A rise in model task horizons or AI-led research would support the scenario’s claim that systems can help build better systems. Data-center construction, power forecasts and robot installations address whether computing capacity and physical automation can expand at a comparable pace.

The labor figure offers a check on claims about near-term workplace disruption: the reported 2% of firms with AI-related job cuts is an observation for a defined survey period, not evidence that AI has caused a particular share of total job losses. Taken together, the measures do not establish that a rapid feedback loop is underway. They show what the author plans to monitor and where the evidence would need to move for the scenario to gain support.

Products Worth Considering

Samsung Odyssey OLED G8 34-Inch Curved Ultra WQHD Gaming Monitor

Samsung Odyssey OLED G8 34-Inch Curved Ultra WQHD Gaming Monitor

As an affiliate, we earn on qualifying purchases.

A Tracker for a 2027 Scenario

The watchlist follows an earlier scenario analysis, titled “Scenario 2027: A Year of AI Progress in Two Months, Until Compute Runs Out?” The article describes that scenario as a loop in which AI systems take on more of the work of developing future AI systems, potentially compressing progress before constraints in chips, buildings and power intervene. It then posits a slower phase involving robots, factories and labs.

The November 18 piece says it does not repeat the case for that scenario. Instead, it records current readings, thresholds and data sources, with dated updates planned as new figures appear. Its author discloses that AI tools, including Anthropic’s Claude, assisted with research and drafting, and says Anthropic, OpenAI and METR did not review or sponsor the article. METR’s assessment cited in the tracker is dated November 5, 2026.

“METR estimates that Anthropic now improves its models on well-specified tasks at around twice its earlier rate, and “would not be surprised” by 1.3x or 3x.”

— Thorsten Meyer AI, describing METR’s assessment

Key Measures Lack Update Dates

Several of the most direct measures of AI capability and lab automation have no announced publication schedule. METR’s time-horizon page is no longer actively updated, according to the tracker, and it says the current suite cannot measure reliably beyond 16 hours. That limits how clearly the proposed threshold can be tested with the existing measure.

The article also does not establish that the six signals will move together, or that any one threshold would prove the scenario. The reported figures come from different sources, periods and methods. Anthropic’s internal measure of AI-led research is not a sector-wide estimate, while OpenAI’s self-improvement classification is the company’s own threshold. Whether power capacity, construction and robot deployment can keep pace with AI demand remains uncertain.

December Brings Infrastructure Readings

The tracker lists the next scheduled data releases as US construction spending on December 1 and power-related figures on December 8, 2026. Those readings can show whether data-center building continues to rise and whether electricity growth expectations change.

No calendar is given for new METR or lab automation figures. The author says the tracker will receive dated notes as data arrive, so readers will need to watch for updates from the organizations that publish those measures as well as the scheduled infrastructure releases.

Key Questions

Is the article predicting that AI will take off in 2027?

No. The author describes it as a scenario tracker, not a forecast, and says its thresholds are judgment calls.

What are the six signals?

They cover AI task time horizons, AI’s share of AI research, independent estimates of acceleration, data-center construction, industrial robot deployment and firms reporting AI-related job cuts.

What is the current AI task-horizon reading?

The tracker reports METR’s last public 80% time-horizon record as 185.9 minutes for an early Claude Mythos Preview model from April 2026. The scenario threshold is 16 hours by mid-2027.

When are the next updates due?

The listed scheduled releases are US construction spending on December 1, 2026, and power-related figures on December 8. The article gives no publication dates for new METR or lab automation readings.

Source: Thorsten Meyer AI

You May Also Like

Latest Food Trend: The Viral Popularity Of Taco Bell’s Ice Cream Taco

The viral popularity of Taco Bell’s Ice Cream Taco highlights a new food trend, gaining rapid attention on social media and consumer platforms.

The Critical Role Of Canada’s Energy Infrastructure In AI

Canada’s hydro power limits and provincial restrictions challenge its role in AI infrastructure, affecting global data center growth and energy policy.

DeepMind’s WeatherNext Model Achieves Breakthrough Forecasting Cyclones

DeepMind’s WeatherNext model has demonstrated unprecedented accuracy in predicting cyclones, marking a significant advance in weather forecasting technology.

Muse Spark 1.1

Meta has launched Muse Spark 1.1, an updated AI model aimed at improving language understanding and generation, with new evaluation metrics and capabilities.