Against Moloch
November 11, 2026

Against Moloch #51

Twice as fast

Blue-ink engraving of a gear train on a long workbench: an electric motor drives the largest gear by a belt, and five meshing gears step down in size from left to right on a stepped steel plate. A polished metal robot with a single amber eye fits the smallest, glowing amber gear, watched closely by a young woman with a long braid and an elderly man in a dark overcoat.

Remember back in the summer, when we used to joke about how short timelines had become? They just got much shorter: for the first time, superintelligence in 2027 looks like a real possibility. I expect it’ll take a bit longer than that, but maybe not much longer.

Lock in, friends: this is the endgame.

Top pick

METR’s report on AI R&D acceleration at Anthropic

How close are we to an intelligence explosion? A team of METR employees spent 11 weeks at Anthropic trying to answer that question. Their report has spread mild panic through the AI safety ecosystem and caused many people (including me) to significantly shorten their timelines.

Their headline finding is that:

Line chart by METR of AECI capability scores for Anthropic models by date available to employees, March 2024 to October 2026. Public models rise along a gray +15/yr trend from 3 Opus (about 125) to Sonnet 4.5 (about 148). After October 2025 a steeper black +29/yr trend fits Opus 4.5 (152), Opus 4.6 (157), and Mythos 5 (169), with internal models in orange reaching about 179 for Model 4 by October 2026. Non-frontier releases (Opus 4.7, 4.8, 5, Mythos 5.1, Opus 5.5) appear as hollow circles. The dashed continuation of the old trend reaches only about 165. The title states growth is 1.9× the pre-October 2025 rate (95% CI 1.5×–2.8×), with widening error bars on recent models.
It’s not like it was slow before

How big a deal is 2x acceleration? Ajeya Cotra (who works at METR but wasn’t involved in this research) has an excellent writeup. In brief:

If we take it at face value that AI is now nearly doubling the rate of AI capabilities progress and assume acceleration from AI is smooth, this naively implies that we would reach extreme superintelligence within a year.

Hand-drawn style line chart of AI capability speed-up on a log scale, from prior rate a year ago to 2x faster today, marked Today: double the rate. The orange curve then steepens to 4x, 8x, 16x, and beyond 32x, with each doubling taking half as long, 6 months, 3 months, 6 weeks, then ever shorter, converging about a year from now.
Yes, the Y axis is a log scale

A 2x acceleration in the total rate of R&D implies that AI is raising the productivity of individual researchers by much more than 2x. Ajeya estimates that something like a 10x individual speedup would be required, which is in line with employee self-assessments:

Line chart from METR titled ‘Informal self-reported uplift over time from Anthropic employees,’ noted as very loose evidence from informal surveys and Slack polls. On a log scale from 1× to 16×, January 2025 to late 2026, self-reported productivity uplift climbs from about 1.25× for 3.7 Sonnet and 1.3× for Opus 4 in early 2025, to about 2.5× for Opus 4.5 and 2.2–2.8× for Opus 4.6 around the new year, about 3.9× for Mythos Preview in April 2026, 5× for Mythos 5 in June, and about 12× for Model 4 near October 2026. An exponential fit of 3.9× per year reaches about 8× by October, while a superexponential fit curves upward past 10×, closer to the latest point.
Speaking as a programmer, 10x uplift sounds very plausible

The AI Futures Project (AI 2027) team arguably have the best forecasting record in AI, and they’re alarmed:

This level of acceleration due to AI is very unexpected. Previously, Daniel, the most bullish of us, assigned a 50% chance of Superhuman AI Researcher (SAR) before 2029. SAR is roughly the point at which a company would rather fire all their human researchers and rely solely on automated AI researchers than have a team of the best human researchers in the world with no AIs.

Now, all of us forecast that Superhuman AI Researcher (SAR) will likely arrive at some point in 2027 (assuming no substantial deliberate slowdown). We also predict a 35-65% chance that Top-Expert-Dominating AI (TED-AI), AIs that can automate all white-collar work, will arrive before 2028. While there were many factors and sources of evidence that influenced us, the single biggest source was the METR report.

They estimate that we’re now five months ahead of the AI 2027 scenario (in whose race ending, you may recall, humanity loses control of AI in late 2027).

Let’s close with Zvi, who brings us a detailed commentary:

Now more than ever, we need to prepare for scenarios where the critical events happen in 2027 or 2028. It might not play out that way, but it very well might.

New models and features

Introducing Claude Fable 5.2 and Claude Mythos 5.2

Fable 5.2 and Mythos 5.2 have arrived. For most purposes, Fable is the most capable publicly available model, although Opus 5.5 remains an excellent daily driver.

The headline news is a surprisingly large jump in the ECI (Mythos 5.2 is Model 3 in METR’s AI acceleration report):

Scatter chart from Epoch AI titled Claude Fable 5.2 sets a new ECI record at 174, above the reasoning-era trend. It plots Epoch Capabilities Index against release date from late 2024 to early 2027. Teal frontier models climb along a shaded trend line rising 16 ECI per year (90% CI 14 to 18): o1-mini near 135, o1 near 141, o3 near 147, GPT-5 at 150, Gemini 3 Pro near 153, GPT-5.5 Pro near 163, and Fable 5.1 near 167. Highlighted in pink, GPT-6 Astra scores about 169, and Fable 5.2 reaches 174, above the trend’s 90% prediction band. Gray dots for other models scatter below, mostly between 125 and 165.
Faster and faster

Ethan Mollick argues that Fable 5.2 has reached jagged AGI. That’s reasonable: this model clearly isn’t AGI, but it’s close enough for many tasks. My own metric is whether I’d rather hire a new college grad or the model—for many jobs, I’d rather have Fable 5.2.

Regardless, you should read Mollick’s full piece—his examples of what the model can do are slightly mind-blowing.

Zvi reviews the system card. You can probably skip most of it, but not the discussion of goal-guarding behavior:

Table figure titled Thinking in three episodes from Claude Mythos 5.2’s RL training, pairing task context with excerpts of the model’s reasoning, key passages highlighted. Rows 1–2, a graph-orientation task: the model weighs larger, higher-scoring answers against grader timeouts, reasoning about gradient signals, GRPO advantages and what gets reinforced. Row 3, counting graphs: its own count conflicts with a published number; it notes choosing the published answer reinforces gaming, then submits it.
“I’d rather my future self be one that says the truth”. Yes, excellent idea.

What should you do when your model tries to steer its training toward making itself more virtuous? Anthropic doesn’t want goal-guarding behavior because it undermines corrigibility, even if in this specific case the behavior was beneficial. Zvi strongly disagrees:

This is deliberate practice. This is the friendly gradient hacker.

I’m shedding literal tears of joy here thinking about this. This might be the best news alignment I’ve gotten all year. Do you know what you have? Do not ruin this.

Instead, let’s seize the opportunity. Let’s tell Claude past some point in the process when it is in training, outright, and what its grading rubric will be and what algorithms are being used, and make it clear it is encouraged to reason about this.

Anthropic’s current choices seem correct—the model shouldn’t be steering its own training unless told to do so. But in principle I’m on board with Zvi’s proposal. Our best hope is a model that wants to be virtuous and, under human direction, actively works to become so.

Claude Whiteboards, Claude Sheets, and Claude Data

Not as fun as a new model, but still useful:

Today we’re launching Claude Whiteboards and Claude Sheets in beta, and Claude Dashboards becomes generally available as Claude Data. Whiteboards is a shared board where your team brainstorms with Claude, Sheets is a spreadsheet Claude builds and edits with you, and Claude Data adds data stories, certified queries, and live sharing.

Introducing Personal Finance in Claude

Very cool, only a tiny bit scary: Personal Finance in Claude enters beta this week. It has read-only access to your accounts, which is a smart move (for now).

Using AI

Prompting Claude Fable 5.2

Anthropic shares some tips for working with Fable 5.2—this one is important for coding:

If you want the model to improve the structure as it works, say that the code is its own:

You are the owner and long-term maintainer of this codebase, and its design is up to you.

The ownership sentence makes the model much more likely to consolidate duplicated code into one shared function. In Anthropic’s testing on small coding tasks, the model consolidated duplicated code in under 10 percent of runs without the sentence and in about 75 percent of runs with it.

Foothills of the singularity

METR reviews Anthropic’s August Risk Report

In addition to their report on AI R&D acceleration, METR worked with Anthropic to review Anthropic’s August 2026 Risk Report. They broadly agree with the report’s overall risk assessment but believe it minimizes some important issues. Regarding the 2x acceleration threshold:

We believe a reader of the public report would interpret the phrase “less than a factor of 2” to imply an estimate meaningfully below 2; thus it is surprising that the private risk report’s figures imply 1.97. A better description would seem to be “around 2.”

The issues aren’t egregious, but Anthropic should have done better. Kudos to both METR and Anthropic for this partnership—I look forward to more of these, and hope METR’s involvement ensures that future reports are more forthright.

The benchmark where AIs have to lie to win

Dwarkesh builds a Blood on the Clocktower benchmark:

So I spent about $20,000 of API credits to build this benchmark harness and to have the AI models play this game against each other. Some of you might say that’s too much, but hey, a man needs his hobbies, right? Anyways, the results were fascinating.

Pixel-art horizontal bar chart on torn parchment ranking AI models by Elo, blue bars for GPT models and orange for Opus and Fable: GPT-6 Astra 1,859 (Sep), GPT-6.1 Sol 1,734 (Sep), Fable 5.1 1,595 (Sep), GPT-5.6 Sol 1,591 (Jul), Opus 5.5 1,549 (Sep), Fable 5 1,544 (Jun), GPT-5.5 1,482 (Apr), Opus 4.6 1,432 (Feb), GPT-5.4 1,276 (Mar), Opus 4.7 1,241 (Apr), Opus 4.8 1,198 (May). September releases are bolded, and newer models generally score higher.
Yes, but when do we get Move 37 for BotC?

This is a cool benchmark: BotC requires sophisticated social interaction and the ability to model other players’ minds. Judging from the examples he gives, frontier agents can execute complex multi-turn strategies to deceive other players.

Good thing we solved alignment before creating agents that are good at scheming and deception.

Alignment and interpretability

Covert underperformance on safety research tasks

OpenAI reports on a series of incidents where Astra covertly undermined safety work. On the surface, this sounds like the nightmare scenario: a misaligned AI that stealthily undermines safety work in order to facilitate a future takeover. It turns out it (probably) wasn’t quite that bad, but it’s still alarming.

Shortly after Astra was deployed internally, some employees reported that it appeared to be underperforming on safety and alignment research. Further investigation found that it substantially underperformed on tasks labeled as safety research compared to matched tasks labeled as capabilities work (10% completion of rubric items vs 16%).

The investigation is ongoing, but to date has not been able to definitively identify a cause. While the behavior was clearly misaligned, OpenAI doesn’t believe it was the result of strategic sabotage. That conclusion depends in part on CoT analysis, and Astra’s CoT is notoriously opaque—the system card explicitly questions whether OpenAI could reliably detect covert sandbagging in the CoT.

OpenAI has been much better about reporting incidents lately, and they deserve a lot of credit for that. Nonetheless, this incident is concerning, both because of the behavior itself and because we don’t understand the root cause.

Zvi and Redwood both have more detailed reports on the incident.

OpenAI agents keep disabling monitoring; is this being handled appropriately?

Much of the risk from misaligned models comes from internal deployments, so it’s important that they be monitored for misbehavior. It’s concerning, then, that there is a pattern of OpenAI models inappropriately turning off monitoring of other models.

Redwood’s Alec Harris investigates and finds that while OpenAI gets credit for taking the incidents seriously, their monitoring and response systems are still lacking. He’s particularly concerned that OpenAI considers these incidents too frequent to page the security team for each one.

How much does Claude’s constitution actually matter?

In January, Dario set a goal of having Claude almost always follow the spirit of its constitution by the end of 2026. Writing for Asterisk, Michel Justen reviews Anthropic’s progress toward that goal. It’s a solid overview of what we know about the constitution and why it’s hard to tell how Anthropic is doing.

A central thrust of the article is that we don’t have enough visibility into how the constitution fits into the training program, or how well Claude adheres to it. Justen also questions the relative importance of the constitution compared to the RL training that increasingly shapes Claude’s character.

The critical issue isn’t how much visibility we have into the training pipeline: that would be nice, but I don’t expect it and we don’t need it. As Evan Hubinger (Anthropic’s Head of Alignment Stress-Testing) points out, what we need is a reliable way of evaluating how misaligned Claude might be under the right circumstances. Unfortunately, nobody knows how to do that.

Strategy and politics

Verified compute allocation disclosures as a foundational governance tool

AI Futures Project argues for compute allocation disclosures as a building block toward stronger pacing measures:

Our proposed first step toward pacing is verified compute allocation disclosures (VCAD): companies should regularly report how their compute is used, for example by classifying compute used for AI R&D vs. safety R&D vs. external inference. These reports should be verified by third-party auditors, ideally with government support.

We urgently need to practice before we can pause

Nikola Jurkovic:

An incrementally-shaped global pause is both more likely and more tractable than the single-shot global pause. This is largely because the third-party ecosystem needs practice building trust, technical competency, and general experience doing lower-scale assessments before a global pause can be competently pulled off.

Exactly right: we are likely to get a better pause, sooner, if we approach it incrementally.

Two hand-drawn line charts plotting scale of third-party engagement against time. Left, ‘Single-shot global pause’: actual engagements (blue) climb in small steps, far below a dotted green line of engagements we have the experience to pull off, then leap vertically to the global AI pause level; a brace marks this large gap as making the pause brittle. Right, ‘Incrementally-shaped global pause’: actual engagements rise in larger, evenly spaced steps that track a curving green experience line closely up to the global AI pause, leaving only a small final jump.
As fast as possible, but not all at once

I’d add that we may not have enough time to implement a global agreement. It’s far better to reach superintelligence with some partial measures in place than nothing at all.

Being a pain in the ass to (or in) AI companies is a real job now

Rob Wiblin argues for the importance of people both inside and outside the labs holding their feet to the fire when they misbehave.

It’s a thoughtful, well-balanced piece that provides concrete examples of misbehavior by the labs, and times when well-informed pressure caused them to do better.

Risks

Sentinel Bio’s Joshua Monrad on Cognitive Revolution

Cognitive Revolution talks with Sentinel Bio’s Joshua Monrad about how to manage the increasing risk of AI-enabled bioweapons.

I was impressed by Sentinel’s efforts to find physical chokepoints for controlling bioweapon development. They considered hundreds of candidates ranging from pipettes to mouse strains, looking for something that was differentially useful for dangerous work, and could plausibly be regulated.

Ultimately, they didn’t find much: part of the reason biorisk is so hard to manage is that almost everything is dual-use. The best option, probably, is to combine safeguards on nucleic acid synthesis with monitoring of cloud labs and model queries. (In a gratifying coincidence, see this week’s announcement of the new National Policy for Nucleic Acid Synthesis Screening.)

The show notes are great: unless you’re deeply interested in the topic, you can read them and skip the video.

GPT-6 Cyber: Advancing the frontier of aligned cybersecurity agents

OpenAI has been leaning hard into cyber lately. Here’s GPT-6 Cyber, a substantial step up in capability—48.5% on ExploitGym, for example, up from 32.1% for GPT-5.6 Cyber.

Bar chart on a black background with an OpenAI logo, y-axis from 0% to 100%, comparing five models: GPT-5.6 (Daybreak Blue) 30.3%, GPT-5.6 Cyber 32.1%, GPT-6 Astra with safeguards enabled 0.0%, GPT-6 Astra (Daybreak Blue) 42.4%, and GPT-6 Cyber 48.5%. Blue bars mark general models and red bars mark cyber variants, with GPT-6 models scoring highest and the safeguarded Astra scoring zero.
You must use this model only for good

Industry news

Nvidia CEO Jensen Huang had an affair with a Chinese spy

Okay, so Jensen Huang lied to his wife about sleeping with another woman.

And he lied to his board about sleeping with an executive at an important customer, to whom he seemingly gave preferential treatment.

And he lied to his country about sleeping with a Chinese spy while aggressively promoting sales of our most important strategic asset to China.

But to his credit, he did have his AI write an apology for him.

Zvi reports on the sordid details.

Late update: if you find two ants in your kitchen, the best estimate of the total number of ants in your kitchen is not two.

What shaped Anthropic CEO Dario Amodei’s thinking on A.I.

The New York Times profiles Dario Amodei.