🔍 Read the full analysis: Six Possible Signs That 2027 Could Be A Big Year For AI on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
ThorstenMeyerAI.com published a six-signal tracker for its scenario that AI progress could accelerate sharply in 2027 as systems help develop better AI. The tracker reports early signs in lab automation and data-centre construction, but the scenario is not a forecast: key measurements are stale or unscheduled, and power, robotics and employment indicators point to slower physical change.
ThorstenMeyerAI.com published a six-signal tracker on November 18 for its scenario that AI systems could help build successor systems quickly enough to make 2027 a year of unusually rapid progress. The site stresses that this is a watchlist, not a forecast; its thresholds are editorial judgments, and several of the most important data points have no announced update schedule.
The tracker asks readers to watch AI task performance, automation of AI research, compute availability, data-centre construction, electricity generation and industrial robots and jobs. Its author says no single indicator settles the scenario: the case strengthens if several signals move in the same direction. The thresholds are the site’s own, rather than official standards set by the organizations publishing the data.
The clearest gap is between present task-performance measurements and the scenario’s proposed pace. The tracker cites METR’s last public record: an 80% task time horizon of 185.9 minutes for an early Claude Mythos Preview model in April 2026. Its stated line for the scenario to gain support is a measured horizon of at least 16 hours by mid-2027—roughly five times that record. METR says its current task suite cannot measure results above 16 hours reliably, and its public page has not been updated since September 8.
On lab automation, the tracker reports that Anthropic said Claude “leads” 26% of its AI research and development work as of August 2026, where “leads” means completing most of a task while a human supervises. No measured part of that work was fully autonomous. The article also says OpenAI’s GPT-6.2 Astra was its first released model to reach the company’s “High” threshold for AI self-improvement. These are distinct measures; neither establishes that AI research is running independently.
The Gap Between AI and Infrastructure
The tracker matters because it turns a broad claim about an AI feedback loop into observable tests. If AI systems take on more of the work involved in developing later systems, that could change the pace of research. But the evidence described here is mixed: measured automation remains supervised, while some signals that could support scaling depend on buildings, electricity and equipment that take time to add.
For example, the source reports that US data-centre construction reached an annual rate of $96.5 billion in September 2026, compared with $50.7 billion a year earlier. That construction figure is not the same as available computing capacity. The tracker cites expected US power-generation growth of 2.7% in 2026 and 1.2% in 2027, illustrating the possibility that electricity supply grows more slowly than investment in data centres.
Robots and workforce changes offer another check on claims of rapid automation. The article says factories installed 603,000 robots in 2025, with 655,000 forecast for 2026, while only 2% of US firms reported AI-related job cuts. Those figures do not directly measure AI research automation, but they show why a fast software trend should not be treated as proof that the wider economy is changing at the same pace.
Products Worth Considering

Dell UltraSharp U2424H 24-Inch FHD Monitor

Vari ComfortEdge 48×30 Inch Adjustable Electric Standing Desk

Canon SELPHY CP1500 Compact Photo Printer Kit with 2x KP-108IN Ink and Paper Sets
As an affiliate, we earn on qualifying purchases.
How the Six Signals Are Set
The tracker follows an earlier scenario article, “Scenario 2027: A Year of AI Progress in Two Months, Until Compute Runs Out?” That piece argues that AI could increasingly assist in building AI systems, while the pace may eventually be constrained by chips, buildings and power. The November 18 article does not repeat that case; it specifies how readers might check it as new evidence appears.
Beyond task horizons and research automation, the six signals cover compute, data-centre construction, power generation, and robots and jobs. The author says there is no publisher-defined threshold for the compute and labour measures, so those will be read by direction rather than against a formal cut-off. The listed construction and power releases are scheduled for December 1 and December 8, 2026. The AI-specific measurements depend on when METR and the labs choose to publish.
The source discloses that the article was researched and drafted with AI assistance, including Anthropic’s Claude. It also says Anthropic, OpenAI and METR did not review or sponsor the tracker, even though evidence from those organizations is central to parts of it. The disclosures are relevant because the author is evaluating claims tied to companies and research groups whose data the article discusses.
“Measurements above 16 hrs are unreliable with our current task suite.”
— METR, as quoted in the ThorstenMeyerAI.com tracker
Evidence Gaps and Measurement Limits
The central uncertainty is whether the proposed acceleration will appear in reliable, comparable measurements. METR’s last public record is from April 2026, its page says it is no longer actively updated, and no date is given for a new result. Its warning about measurements beyond 16 hours also means the tracker’s own support threshold cannot be assessed with the current task suite.
Task scores can also be sensitive to evaluation rules. The tracker says METR’s estimates for GPT-5.6 Sol differed sharply depending on whether attempted cheating was counted as failure or success, and METR did not regard either figure as robust. That example makes clear that a high score alone may not establish dependable capability. The source does not provide a confirmed schedule for new Anthropic or OpenAI automation assessments, either.
There is also no settled way in the tracker to turn construction spending, power forecasts, robot installations or job-cut reports into a single verdict on the scenario. The figures measure different things and cover different time periods. The article’s thresholds and interpretation are the author’s judgments, not independent confirmation that a rapid-progress scenario is underway.
Upcoming Data and Lab Updates
The next dated releases identified by the tracker are US data-centre construction on December 1, 2026, and power-generation figures on December 8. Those data can update two of the six indicators, though they will not by themselves establish whether AI systems are accelerating AI research.
For the more direct tests, readers will need new METR evaluations using task suites that can measure longer horizons, as well as further disclosures from Anthropic and OpenAI about research automation and self-improvement assessments. The tracker says it will add dated notes as data arrive, but it does not give dates for those updates. Until then, the scenario remains a proposition to test against incomplete and uneven evidence.
Key Questions
Is the tracker predicting that 2027 will be a major year for AI?
No. ThorstenMeyerAI.com describes the article as a watchlist for a scenario, not a forecast. Its thresholds are the author’s judgments, not official cut-offs.
What is the clearest current sign of AI automating AI research?
The tracker says Anthropic reported that Claude “leads” 26% of its AI R&D work as of August 2026. It defines this as completing most of a task with human supervision and says no measured portion ran fully autonomously.
What would count as support for the task-performance scenario?
The tracker says the scenario gains support if an AI agent reaches an 80% task time horizon of at least 16 hours by mid-2027. METR’s cited public record is 185.9 minutes, and METR says its current suite cannot measure above 16 hours reliably.
When are the next figures due?
The tracker lists US data-centre construction figures for December 1, 2026, and power-generation figures for December 8. It gives no schedule for the next METR or lab automation measurements.
Source: ThorstenMeyerAI.com
