iiinova — where your digital self keeps evolving

Jeffrey Lewis @armscontrolwonk.bsky.social · 04/10/2026

I can look it up in Thrusty, but I am not sure I believe the numbers, Have been rebuilding the whole program to deal with those stupid chines/wings whatever they are.

220

Jeffrey Lewis @armscontrolwonk.bsky.social · 04/10/2026

This looks like a two-stage booster to me.

2411

Jeffrey Lewis @armscontrolwonk.bsky.social · 03/10/2026

DPRK Missile Administration (left) vs. Thrusty (right). Assumes a booster that burns out at about 2.2 km/s.

2342

Jeffrey Lewis @armscontrolwonk.bsky.social · 23/09/2026

This offer valid only for selected countries.

1291

Jeffrey Lewis @armscontrolwonk.bsky.social · 23/09/2026

It’s a Hwasong-11 with a glider. BORING.

3557

Jeffrey Lewis @armscontrolwonk.bsky.social · 18/09/2026

I did not realize the DF-15A had thrust termination ports!

0283

Reposted by Jeffrey Lewis

Aaron Stein @aaronstein1.bsky.social · 12/09/2026

Colleagues from around the world joined us @fpri.bsky.social in Philadelphia for a TTX to study nuclear escalation dynamics in space. Thanks to @slair.bsky.social @duitsmanms.bsky.social @armscontrolwonk.bsky.social and Emily Holland for support. Game report coming soon!

1101

Jeffrey Lewis @armscontrolwonk.bsky.social · 11/09/2026

It’s nice because I can ask Claude to generate complicated CLI strings, which I would routinely mistype.

1181

Reposted by Jeffrey Lewis

thinkingbayonet.bsky.social @thinkingbayonet.bsky.social · 11/09/2026

Houthis: what the fuck, you told us it was economical to build the IR sensor with aluminum foil instead of mercury cadmium telluride Claude: That is exactly the right challenge to raise, and the honest answer is that the claim was asserted rather than demonstrated. Inshallah it won't happen again

4761114

Reposted by Jeffrey Lewis

Stanford CISAC @stanfordcisac.bsky.social · 10/09/2026

Please join us in welcoming Jeffrey Lewis (@armscontrolwonk.bsky.social) and Or Rabinowitz to CISAC! Both scholars bring expertise in nuclear weapons policy, nonproliferation, open-source intelligence, and international nuclear diplomacy to Stanford. cisac.fsi.stanford.edu/news/cisac-w...

cisac.fsi.stanford.edu

CISAC Welcomes Nuclear Security Scholars Jeffrey Lewis and Or Rabinowitz

Jeffrey Lewis and Or Rabinowitz bring expertise in nuclear weapons policy, nonproliferation, open-source intelligence, and international nuclear diplomacy to Stanford.

0183

Jeffrey Lewis @armscontrolwonk.bsky.social · 10/09/2026

That’s true! But I still suspect that Turing correlates with p(doom).

020

Jeffrey Lewis @armscontrolwonk.bsky.social · 10/09/2026

It probably also correlates with whether one is more John Searle than Alan Turing.

2230

Jeffrey Lewis @armscontrolwonk.bsky.social · 10/09/2026

+1. A lot of the construction of risk has to do with underlying priors about existential and catastrophic risk more so than the particularities of a given technology. That doesn’t mean the view is invalid; but it does imply one should interrogate those assumptions.

210715

Reposted by Jeffrey Lewis

Ankit Panda @nktpnd.bsky.social · 09/09/2026

I think people with no exposure to effective altruism/online rationalist communities are having a very hard time making sense of the doom-talk coming out of the frontier labs. A bit of orientation with those worlds is helpful context.

1128447

Reposted by Jeffrey Lewis

Sam Lair @slair.bsky.social · 09/09/2026

Curious about how to model a constellation of space-based interceptors? Well @aaronstein1.bsky.social, @armscontrolwonk.bsky.social, & I have a report for you! We tailored SBI constellations to genuine adversary missile threats to get a more realistic, flexible sense of their size & cost

11911

Jeffrey Lewis @armscontrolwonk.bsky.social · 02/09/2026

This is a good answer and it beats the hell out of the alternative.

612013

Reposted by Jeffrey Lewis

Foreign Affairs @foreignaffairs.com · 01/09/2026

Read @armscontrolwonk.bsky.social on why Trump’s plan for a Golden Dome missile shield might accelerate the arms race with China and Russia:

foreignaffairs.com

The Powder Keg in Outer Space

How to head off a new arms competition.

0195

Reposted by Jeffrey Lewis

Sam Lair @slair.bsky.social · 27/08/2026

I return to the Arms Control Wonk Podcast to talk Golden Dome and SBIs with @aaronstein1.bsky.social and @armscontrolwonk.bsky.social Come for our explanations of math, stay to hear what everyone's favorite SBI countermeasure is! www.armscontrolwonk.com/archive/1222...

armscontrolwonk.com

Calculating the Cost of Golden Dome

0295

Jeffrey Lewis @armscontrolwonk.bsky.social · 24/08/2026

That @wellerstein.bsky.social’s book. He’s a friend!

170

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

So TL/DR: 57 seems to low and, in any event, too precise to be an IC estimate. Trump is trying to tell us a secret he heard, but for the life of me I can't figure out what it means.

8733

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

One final caution. Remember Trump's "super duper" missile that Bob Woodward thought was nuclear? We're now pretty sure it was the Common Hypersonic Glide Body. Trump was describing a real program, but mangled beyond recognition. (He also reportedly kept saying "hydrosonic" in private.)

1584

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

Or maybe someone told Trump the DPRK design resembles the B57, which was a fairly light-weight bomb with a yield of 20 kilotons. So he means they have very powerful B57 like nuclear weapons? Really, your guess is as good as mine.

1464

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

Back to Trump. Where did 57 come from? Maybe he means some subset of North Korea's arsenal? Trump referred to "very powerful" nuclear warheads. I might believe that North Korea has ~60 thermonuclear warheads for ICBMs. That would be a plausible increase from 4-6 in 2019.

1452

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

I suppose it's possible that North Korea is just ahead of itself -- that the missiles are available and the fissile material is coming, but they haven't put it all together yet. But it seems more likely to me that we've been slow to update our priors.

1472

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

How does that square with 250 Hwasong-11 launchers (each with 4 missiles)? Or 80 600 mm MLRS launchers (each with 5-6 rockets)? Or destroyers with VLS cells? North Korea is deploying a lot of missiles to arm with just 57 warheads.

2531

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

North Korea claims there are eight distinct different tactical nuclear delivery systems. So that's 4 warheads left for each type? And here's half of them?

1491

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

Indeed, North Korea is also producing MANY nuclear-capable delivery systems including new solid- and liquid-fueled ICBMs. Here's Kim looking at an ICBM production run. I count 25+ missiles. We think there are only ~30 more warheads left to go around?

1541

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

I understand that not every kilogram of material need be weaponized immediately, but given Kim Jong Un's repeated references to the urgency of increasing fissile material production, it's hard for me to imagine that he just going to stockpile it.

1502

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

All told, the DPRK has produced roughly 80-100 kg of plutonium and nearly 3,000 kg of HEU. That is enough for many, many more than 57 warheads -- unless they're building W-47s.

1512

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

Meanwhile, North Korea is piling up fissile material. A new reactor at Yongbyon came online in late 2023 and new centrifuge annexes went up at Kangson (2024) and Yongbyon (2022), and a new enrichment facility appeared at Yongbyon this year.

1502

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

Do the math. 60 in 2017, plus 12 a year for nine years would put the upper bound at around 200 warheads today. Even starting with the more cautious figure of ~30 assembled weapons in 2019, one would have to believe North Korea builds only 3-4 weapons per year.

1544

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

I'd be shocked if the IC offered an individual warhead count. More likely are estimates are derived from how much Pu & HEU produced, divided by how much material each weapon might use. That's how it was reported in 2017 & 2019 stories by @nktpnd.bsky.social and @jobywarrick.bsky.social .

1520

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

One thing I would add is that my friends at @scientistsorg.bsky.social -- @nukestrat.bsky.social and @mattkorda.bsky.social -- and at @sipri.org all lean toward ~60. Disagreement is healthy as long as we're all appropriately uncertain.

1610

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

IC estimates of foreign nuclear stockpiles are basically never point estimates. They're ranges with confidence levels. Declassified documents show rounded numbers—"enough material for X to Y weapons"—not individual warhead counts. "57" is not how the IC talks.

1752

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

Trump says North Korea has "57 very powerful nuclear weapons." 57 is an oddly specific number. A short 🧵on why the real number is almost certainly higher.

1120956

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

He’s a bad President but I think it’s a mistake to organize our practices around catastrophic cases like this:

100

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

The President should be able to tell the people he trusts what he thinks they need to know.

220

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

I don’t care if he’s having an affair, he has original classification authority, and it’s all Calvin-ball anyway. The President can tell the people he trusts what he thinks they need to know.

3330

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

I wish we would take this opportunity to reduce over-classification instead of endorsing over-classification as a pretext to attack Trump. He’s broken enough real laws that we don’t have to endorse the fake ones.

5473

Jeffrey Lewis @armscontrolwonk.bsky.social · 20/08/2026

The code works fine but … yeah. Variable names are a nightmare.

2230

Reposted by Jeffrey Lewis

Aaron Stein @aaronstein1.bsky.social · 04/08/2026

My colleague @armscontrolwonk.bsky.social and I discussed on our podcast the history of our now decade-and-half old show, his future, my future, and our major plans for 2027 and beyond www.armscontrolwonk.com/archive/1222...

armscontrolwonk.com

From Past to Future

161

Jeffrey Lewis @armscontrolwonk.bsky.social · 02/08/2026

“Beltway geniuses” — the people who were smarter than Caddell and took his job.

1200

Jeffrey Lewis @armscontrolwonk.bsky.social · 02/08/2026

Fact check: True. Caddell was talking about Reagan-like landslide for Romney in 2012 but in the end it wasn’t even particularly close as Obama swept Michigan, Minnesota and Pennsylvania, while picking off Florida and Ohio.

1413

Reposted by Jeffrey Lewis

Mike Black @mikeblack114.bsky.social · 02/08/2026

someone needs to stop her with extreme prejudice

719710