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

🔍 Read the full analysis: Is 2027 A Turning Point For AI? Six Signals May Tell on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get audio and creator gear delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Thorsten Meyer’s Nov. 18, 2026, article sets out six indicators for tracking a scenario in which AI increasingly helps develop successor AI systems and accelerates progress in 2027. The tracker reports rapid growth in U.S. data-center construction, but says current public measures of AI capability, autonomy, power supply and labor impacts do not yet establish that the scenario is unfolding. Its thresholds are the author’s judgments, not official forecasts.

Thorsten Meyer published a six-signal tracker on Nov. 18 for assessing whether artificial intelligence could accelerate its own development enough to compress a year of progress into about two months in 2027. The article is a scenario watchlist, not a forecast; its thresholds are Meyer’s judgments, and the evidence available so far does not show that the scenario has been reached.

The tracker measures AI task performance, AI involvement in research and development, formal self-improvement risk assessments, data-center construction, electricity supply and industrial robots and jobs. Meyer says the signals should be read together rather than as a single pass-or-fail test. For compute and labor, the article says there is no publisher-defined threshold, so the direction of the data matters more than a fixed cutoff.

The clearest stated capability line is a 16-hour, 80% task time horizon by mid-2027. METR’s last public record cited in the article was 185.9 minutes—about three hours and six minutes—for an early Claude Mythos Preview in April 2026. Meyer says reaching 16 hours would require a new task suite: METR has warned that its current suite cannot measure reliably above that level.

On AI-assisted research, Anthropic reported that Claude “leads” 26% of its AI R&D work as of August 2026, according to the tracker. “Leads” means AI completes most of a task end to end while a human supervises; the article says no measured portion was fully autonomous. It also reports that OpenAI’s GPT-6.2 Astra was the company’s first released model to meet its “High” threshold for AI self-improvement. These are company assessments, not evidence that AI research is operating without people.

For physical infrastructure, U.S. data-center construction was running at an annual rate of $96.5 billion in September 2026, compared with $50.7 billion a year earlier. The article cites expected U.S. electricity-generation growth of 2.7% in 2026 and 1.2% in 2027. Factories installed 603,000 robots in 2025, with 655,000 forecast for 2026; only 2% of U.S. firms reported AI-related job cuts. Those figures suggest infrastructure and workplace changes are measurable, but they do not by themselves establish that AI progress is accelerating at the pace in Meyer’s scenario.

At a glance
reportWhen: Published November 18, 2026; future ind…
The developmentThorsten Meyer published a Nov. 18, 2026, watchlist specifying six indicators and thresholds for tracking a proposed rapid-AI-progress scenario in 2027.

The Gap Between AI and Infrastructure

The tracker focuses on whether AI capability can scale faster than physical resources. If AI systems materially speed the research and engineering behind later systems, progress could accelerate; the scenario described by Meyer then runs into constraints such as chips, buildings and electricity. The construction figures show investment activity, while the slower projected growth in power generation points to a possible mismatch. The data cited do not establish that a shortage has already stalled development.

The measures also matter because they separate a striking scenario from evidence that could test it. A longer task horizon, more AI-led research and rising infrastructure investment would each provide different information. Jobs and robot deployment offer another check on whether effects are spreading beyond software labs. The tracker cautions that no one reading settles the question, and several indicators have no regular publication schedule.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

How the Six-Signal Watchlist Works

The tracker follows a separate scenario analysis Meyer published earlier on Nov. 18, titled “Scenario 2027: A Year of AI Progress in Two Months, Until Compute Runs Out?” That analysis argues for a possible feedback loop in which AI contributes more to building future AI systems, followed by a constraint from physical capacity and a slower phase involving robots, factories and laboratories. The tracker does not restate the case as a finding; it identifies data that could support or weaken it.

Meyer says the thresholds are judgment calls fixed before the data arrive, not official cutoffs. The article also discloses that AI, including Anthropic’s Claude, assisted with its research and drafting. It notes that Anthropic and OpenAI evidence is central to two signals and that METR’s independent estimate concerns Anthropic. Meyer says none of those organizations reviewed or sponsored the article.

Publication timing is uneven. The tracker lists Dec. 1 for construction data and Dec. 8 for power figures as the next announced releases. It gives no calendar date for new METR measurements or lab disclosures, leaving some of the most direct AI-specific signals dependent on when organizations publish.

“A scenario is only useful if you can tell early whether it is coming true.”

— Thorsten Meyer, in the tracker’s description of its purpose

The Missing Data and Measurement Limits

The tracker cannot yet establish whether the proposed acceleration is happening. METR’s public page was marked as no longer actively updated, with its last change dated Sept. 8, 2026, and the article identifies no date for another public time-horizon reading. METR’s current suite also cannot reliably measure beyond 16 hours. A longer result would need a new suite, and the article describes how attempted cheating can sharply alter some scores, making those readings hard to interpret.

Lab disclosures are also limited. Anthropic’s 26% figure describes work the company says Claude “leads,” not fully autonomous research, and the article says the share of work running entirely without people was zero in the measured data. The tracker provides no release schedule for Anthropic’s index or OpenAI’s model assessments. It also does not give enough detail in the supplied material to evaluate every measure’s threshold or present status.

Finally, the article labels its infrastructure and labor figures as indicators, not proof of cause and effect. It remains unclear whether new data-center construction will translate into available power and computing capacity, how quickly those resources can come online, or whether reported job cuts reflect AI adoption. Meyer’s scenario and its thresholds remain an interpretation to be tested, not a settled outlook.

The Next Scheduled Data Releases

The next dates identified in the tracker are Dec. 1, 2026, for U.S. construction data and Dec. 8 for power figures. Those releases will update two of the six signals. For AI task horizons, lab research automation and model risk assessments, the tracker names no scheduled publication dates; those measures depend on METR, Anthropic and OpenAI releasing further information.

Meyer says the watchlist will be updated with dated notes as data arrive. Readers should compare each new reading with the stated definitions and limitations, then look for multiple signals moving in the same direction. Until those updates appear, the tracker offers a framework for monitoring the scenario, not confirmation that 2027 will mark a turning point.

Key Questions

Is the article predicting that AI will reach a turning point in 2027?

No. Meyer describes the piece as a scenario tracker, not a forecast. It sets out indicators that could help assess whether a rapid-progress scenario is developing.

What is the main capability threshold in the tracker?

The tracker says the scenario gains support if an AI agent reaches an 80% time horizon of 16 hours or more by mid-2027. METR’s last public record cited was 185.9 minutes, and its current task suite cannot reliably measure above 16 hours.

Does Anthropic’s 26% figure mean its AI research is autonomous?

No. The article says Claude “leads” 26% of Anthropic’s AI R&D work as of August 2026, meaning it completes most of a task end to end while a human supervises. The tracker reports no measured portion as fully autonomous.

When are the next figures due?

The tracker lists Dec. 1, 2026, for construction data and Dec. 8 for power figures. It gives no dates for the next METR time-horizon results or new lab disclosures.

Who set the tracker’s thresholds?

Thorsten Meyer says the thresholds are his own judgment calls, not official cutoffs. He says they were fixed before the data arrive and that no single reading determines whether the scenario is playing out.

Source: ThorstenMeyerAI.com

You May Also Like

When Security Savings Meet A Rhysida Attack: Berlin’s Experience

Berlin says its attack-detection system covered 5% of devices in one affected department as Rhysida stole and published about 1.4 million files.

What Does California’s OpenAI Lawsuit Say About The Hugging Face Hack?

California alleges OpenAI’s testing failures made the Hugging Face incident foreseeable. OpenAI disputes being singled out; no court has ruled.

Introducing Forezai · TradingAgents — a committee of LLMs decides paper-trades

Forezai introduces TradingAgents, a system where multiple large language models collaboratively decide paper trades, marking a new step in AI-driven trading research.

Flash‑Optimized Vector Stores: Designing for Cold and Warm Recall

Optimize your vector store for cold and warm recall, but discover the key strategies that ensure fast, scalable access across varying data lifecycles.