How Much AI Progress Can Two Months Bring In 2027?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: How Much AI Progress Can Two Months Bring In 2027? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials 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 November 18 scenario asks whether AI systems could compress a year of progress into two months in 2027 as they help build their successors. The claim is not a forecast: benchmark data show rapid gains, but the benchmark has limits, an acceleration estimate is far below the scenario’s required pace, and compute and deployment constraints remain.

Thorsten Meyer published a scenario analysis on November 18 arguing that AI systems could, in a possible 2027, compress a year of progress into about two months as they help develop future systems. The article presents this as a conditional scenario, not a forecast established by current evidence; it points to fast benchmark gains while identifying limits in measurement, computing infrastructure and real-world adoption.

The scenario’s central mechanism is a software feedback loop: AI systems do more of the coding, testing and experimentation used to build later AI systems, potentially speeding up subsequent development. Meyer says the proposed pace would amount to roughly six times the historical rate of progress. That is substantially faster than an estimate from METR, a nonprofit that evaluates frontier AI agents: METR estimates Anthropic is improving models on well-specified tasks at around twice the earlier rate, while saying it would not be surprised by a rate between 1.3 and 3 times.

METR’s public time-horizon measurements show a large rise in the length of software tasks models can complete with 80% reliability. In the figures cited by Meyer, the measure rose from 0.89 minutes for GPT-4 in March 2023 to 185.9 minutes, or about three hours, for an early Claude Mythos Preview in April 2026. METR marked its public page as no longer actively updated on September 8, 2026. The benchmark measures performance on a specific set of tasks, not general intelligence or the full range of economically valuable work.

Meyer’s own curve-fitting illustrates how much projections depend on the selected period. A fit to METR observations from 2024 produces a doubling time of 3.4 months and crosses a 16-hour horizon in January 2027; a fit from 2019 produces a 6.4-month doubling time and crosses that mark in June 2028. These are extrapolations by the article’s author, not METR forecasts. The article also says a hyperbolic curve performed worst in its backtests, reaching an implausible infinity date when applied to earlier data.

At a glance
analysisWhen: Published November 18, 2026; scenario c…
The developmentThorsten Meyer published a scenario analysis on November 18, 2026, testing whether AI development could accelerate enough in 2027 to deliver a year’s progress every two months.
Crypto market snapshot
Fear & Greed Index
11/100 — Extreme Fear
Bitcoin BTC$64,720▲ 0.2%
Ethereum ETH$1,812▼ 0.9%
Tether USDT$0.9991▼ 0.0%
BNB BNB$592.33▲ 0.0%
USDC USDC$0.9995▼ 0.0%
XRP XRP$1▲ 4.2%
Solana SOL$75.01▲ 3.1%
TRON TRX$0.3088▲ 1.0%
Live data · CoinGecko · alternative.me (24h change)

The Gap Between Benchmarks and Deployment

If AI systems substantially speed up the work of developing their successors, the pace of capability gains could affect research, investment and planning across technology industries. Meyer’s scenario goes further: once physical computing limits ease, more capable and cheaper systems could spread through businesses and reshape work. Those outcomes depend on several steps that have not been demonstrated together.

Current adoption figures in the source material caution against treating benchmark progress as equivalent to broad economic change. It cites survey data showing 18% of U.S. firms used AI in a business function in late 2025 and early 2026, while 2% reported AI-related job cuts. These figures describe different measures and do not establish what adoption or employment will look like in 2027. They do underline the distinction between systems becoming more capable in tests and organizations deploying them at scale.

Infrastructure is another limit. The source reports U.S. data-center construction at an annual rate of $96.5 billion in September 2026, compared with $50.7 billion a year earlier. It also cites Epoch AI’s view that power may become the first constraint on scaling, with new power supply taking three to five years after a decision. A faster software loop would not by itself remove delays in chips, buildings or electricity.

Products Worth Considering

Apple MacBook Pro 16.2-inch with M5 Pro, 24GB Memory, 1TB SSD, Wi-Fi 7, Space Black

Apple MacBook Pro 16.2-inch with M5 Pro, 24GB Memory, 1TB SSD, Wi-Fi 7, Space Black

Apple MacBook Pro 16-inch (M1 Pro, 10‑core CPU, 16‑core GPU, 16GB RAM, 512GB SSD) - Silver

Apple MacBook Pro 16-inch (M1 Pro, 10‑core CPU, 16‑core GPU, 16GB RAM, 512GB SSD) – Silver

As an affiliate, we earn on qualifying purchases.

What METR’s Time Horizon Measures

METR defines an AI time horizon as the human-expert completion time for tasks at which an agent is predicted to succeed at a specified reliability. The 80% horizon means the system succeeds four times out of five at that task duration. The tasks cited by Meyer primarily concern software engineering, machine learning and cybersecurity.

METR cautions that these tasks are cleaner and more bounded than many real jobs. It says a two-hour task may resemble work a new hire or freelance contractor could complete in two hours more than work handled by an experienced colleague. The benchmark also has a measurement ceiling: according to the source, METR says results above 16 hours are unreliable with its current task suite. Of 31 tasks estimated to take humans at least eight hours, human times were measured for only five; the remainder were estimates.

The scenario also distinguishes fast software development from a slower physical feedback loop. Meyer calls the latter embodied self-improvement, referring to automation in robots, factories and laboratories. The article argues that physical systems could help expand computing capacity, but does not establish that this cycle will arrive on the proposed schedule.

“We reach ASI either in 2027 or shortly after. The effect: in two months we see the progress of a whole year, and then further progress until we hit a ceiling in available compute.”

— Thorsten Meyer

The Evidence Still Leaves Open Questions

The central claim—that a year of progress could arrive in two months—remains unverified and conditional. The source’s benchmark series ends with an April 2026 model result, and METR’s public page was no longer actively updated as of September 8, 2026. The reasons for that status and whether a comparable public series will resume are not specified in the supplied material.

It is also unclear whether results on METR’s software tasks translate to faster AI research, broader work performance or measurable economic gains. The benchmark has task-suite limitations, while the scenario’s projected dates depend on which historical observations are used in the curve fit. The source does not provide enough information here to independently assess every model, data point or calculation.

Meyer’s article says OpenAI’s GPT-6.2 Astra was the first model released externally by OpenAI to meet its “High” threshold for AI self-improvement. That is a company-defined threshold as presented in the source; the material does not provide the evaluation details needed to establish what the result predicts about future development. Meyer also discloses that the article was researched and drafted with AI assistance, including Anthropic’s Claude, and says Anthropic, OpenAI and METR did not review or sponsor it.

Signals to Watch Through 2027

The scenario will be easier to evaluate if independent, regularly updated benchmarks show whether AI agents keep extending the length of tasks they can complete reliably. Any such comparisons need to account for changes in task difficulty and measurement methods, rather than treating every increase as directly comparable.

Other signals include evidence that AI systems are speeding up the research and engineering used to create subsequent systems, as well as whether this effect approaches the sixfold acceleration the scenario requires. Public details about evaluations such as OpenAI’s self-improvement threshold would help readers distinguish a stated capability rating from demonstrated gains in development speed.

On the physical side, data-center construction, chip supply and access to electricity will indicate whether infrastructure can keep pace with demand. Business adoption and employment data can show whether capability changes are reaching workplaces. The article sets out these indicators but does not give a confirmed date for a decisive test; the proposed 2027 timeline remains a scenario to track, not an announced milestone.

Key Questions

Is the two-month estimate a confirmed AI forecast?

No. Meyer presents it as a scenario for 2027, not a forecast established by the evidence. Its central acceleration claim remains uncertain.

What does METR’s 80% time horizon measure?

It estimates the duration of tasks, measured by human expert completion time, at which an AI agent succeeds with 80% reliability. The cited tasks focus mainly on software engineering, machine learning and cybersecurity.

How fast has the benchmark increased?

The figures cited in the article rise from 0.89 minutes for GPT-4 in March 2023 to 185.9 minutes for an early Claude Mythos Preview in April 2026. METR cautions that measurements above 16 hours are unreliable with its current task suite, and its public page was no longer actively updated as of September 8, 2026.

What could prevent the proposed acceleration?

The scenario could fail if AI systems do not substantially speed up AI development, or if limits in chips, data centers and power constrain computing capacity. The source cites an estimate that new power supply can take three to five years after a decision.

Does faster benchmark performance mean AI will quickly replace jobs?

No. Benchmark results do not by themselves establish workplace adoption or job losses. The source cites survey figures of 18% of U.S. firms using AI in a business function and 2% reporting AI-related job cuts in late 2025 and early 2026; those observations do not determine what will happen in 2027.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

Bloomingdale’S NYC Pop-Up With Flamingo Estate: a Must-Visit

The Bloomingdale’s NYC pop-up with Flamingo Estate promises a delightful fusion of sustainable luxury—discover the exciting offerings waiting just for you!

A Top Microsoft Executive Foresees AI Reshaping the Future of Wealth Management

Incredible advancements in AI promise to revolutionize wealth management, but what unexpected changes could this mean for your investment strategies?

Lightning Liquidity Pools: The New Frontier for Yield Seekers

Jump into Lightning liquidity pools to unlock new earning opportunities and discover how they are reshaping the future of Bitcoin transactions.

Software-Defined Warfare: How Ukraine’s Delta Turned the Battlefield Into a Shared, Real-Time Map

Ukraine’s Delta is a cloud-based, browser-run battlefield system that fuses real-time intelligence, revolutionizing modern warfare and strategic agility.