iiinova — where your digital self keeps evolving

Sophisticated statistics can produce extremely precise nonsense when the design fails to identify the causal mechanism.

0348

It turns out that the novel proximity method they used in the papers is the real reason for the results. The results are positive even with negative synthetic data. They should have completed more internal testing and baselining before using it for published results.

2150

Using the method specified in the papers, Deric Tilson and I found that every landmark we tested can be attributed to increases in cancer, most at much higher values than nuclear power. Harvard alone accounts for almost as many attributable deaths as all US nuclear power.

1120

A recent series of papers suggested that nuclear power plants are responsible for more than 100 thousand cancers. Using their method, every landmark we tested causes cancer

2121

Yes, clusters far from the centroids tested did skew the results. The original paper's sensitivity test for this was defined in a way that would not find the issue and falsify the hypothesis. The distance of the power plant centroid from the centroid of the host county also has a large effect.

020

Their positive association is supposed to indicate increases in cancer deaths with greater proximity. Everything should not have an increase relative to everything else or you only see net increases (relative to real recorded values). That's one of the problems they didn't test.

020

Hi! Thanks for actually reading the article instead of just the figure 😄. There will be a full journal article, but felt it was important to publish this before the long journal publication cycle. The code and some data will be available with that paper, but the CDC data is not distrubitable.

150

The cost estimate was done before anyone went underground to evaluate. That was intentional at this stage, to reduce from 5 sites to 3 that warrant further evaluation. But the estimates were so uncertain they couldn't be used for ranking. The sites could have been in any order.

The cost error that reverses each pair of sites, computed from DOE’s Table 4-6 estimates. Each percentage is a move in that site’s own base-case cost: the higher-ranked site’s rising, the lower-ranked site’s falling by the same percentage. Dollar swing is the two movements combined, in 1986 dollars.

010

Many attributes are a functional tie. Others vary, but have too little weight to affect the outcome. Cost carried almost all of the discriminating power.

Each attribute’s share of the weighted between-site variation, reconstructed from DOE/RW-0074, Tables 4-7 through 4-9.

Promisingly, all five sites passed on the attribute a repository must get right—safety—with large margins. In fact, most site attributes are decision-irrelevant; they cannot change the outcome across their full range.

I rebuilt DOE’s 1986 model for siting a nuclear spent fuel repository, then tested it in ways that DOE never did. What the ranking was actually driven by was uncertain cost estimates. This becomes visible only when the model is taken apart.

America chose its nuclear waste site based on a cost estimate, by accident. There are lessons that must be applied before the next site evaluation.🔌💡

The same five sites under five readings of DOE’s own 1986 numbers; the last column is the ranking that went forward. Rank 1 is the most favorable site, and the shaded blocks are functional ties.

120

What actually helps: Trace constraints to their source, not where they surface. Reduce seams where you can, guarantee sequencing where you can't. Stop adding decision layers to your own projects.

000

So why do we keep missing this? Incentives are stratified too. Every actor sits inside one layer, so nearly every proposed fix is introspective to each layer. It is rarely the layer where the root of the problems exists, and proposals almost never cross layers

Risk isn't reduced when you move it. It's converted. Cost risk → contractor → counterparty risk Financing risk → public → political risk Compliance risk → suppliers → supply chain risk By the time risk is untenable, it looks nothing like the original source.

100

Gas turbines are sold out past 2030 or 2035, counting expansion already underway. One of the most mature, least regulated supply chains in the country.

100

You can see a live test of this. Data centers relieved the constraint on demand for new electricity faster than anything in decades. Projects initially accelerated. But, the constraint moved — to interconnection, transmission, supply chain.

100

The implication is very tangible and stark. These approvals are non-fungible. Regulatory approval can't offset a missing buyer. Financing can't substitute for a grid connection. A project that has cleared every condition but one isn't 90% done. It's at zero.

100

Dr. Adam Stein @astein.bsky.social · 17/07/2026

Large infrastructure projects don't fail because of a single bottleneck. They fail because approval is stratified across independent layers, with different definitions of risk, and non-substitutable approvals. My new report explains why solving one often just moves the problem. 🔌💡

121

I fully electrified my previous home. It ended up being an unfortunate choice for resilience reasons. But, it was very efficient and clean when the electricity distribution system cooperated.

130

Dr. Adam Stein @astein.bsky.social · 17/07/2026

Ok, I'm listening. Say more about which layer you think the intentional valuation of information between a buyer and seller loses meaning.

100

Dr. Adam Stein @astein.bsky.social · 16/07/2026

Is there a "Reverse Information Paradox", as Satya Nadella suggests? He is half right, but only if you set a narrow scope—include firms, but exclude people.

The Reverse Information Paradox Depends on the People it Leaves Out.

100

Dr. Adam Stein @astein.bsky.social · 01/05/2026

Licensing microreactors and other low-consequence reactors took a major step. The NRC published a new proposed framework, called Part 57. I outline some of the major provisions here. The rule is open for public comment for the next 45 days. 🔌💡 thebreakthrough.org/issues/nucle...

thebreakthrough.org

Part 57: Right-Sizing Nuclear Regulation for Low-Consequence Reactors

042

Dr. Adam Stein @astein.bsky.social · 02/04/2026

Eventually, policymakers listened. It took a few blackouts to make it clear that reliability is a necessity, while also considering clean energy. This outcome will benefit CA with clean, reliable, and economic power for decades to come.

020

Dr. Adam Stein @astein.bsky.social · 02/04/2026

The community around DCPP is the strongest supporter of the plant. Organizations like @moms4nuclear have worked on this project for a decade youtu.be/Xzz1WOZw47c?si… mothersfornuclear.org

youtu.be

Mothers for Nuclear's Fight at Diablo Canyon | Earth Focus

YouTube video by PBS SoCal

100

Dr. Adam Stein @astein.bsky.social · 02/04/2026

The NRC renewed the license for Diablo Canyon Power Plant for another 20 years! This is a major milestone. Just a few years ago DCPP was set to retire under CA policy. Thanks to the work of many people, and the undeniable need for reliable power, DCPP was renewed instead. 🔌💡

191

Dr. Adam Stein @astein.bsky.social · 26/03/2026

Find the tracker here: thebreakthrough.org/issues/nucle...

thebreakthrough.org

NRC Rulemaking Tracker

000

Dr. Adam Stein @astein.bsky.social · 26/03/2026

The Nuclear Regulatory Commission is working on a large volume of rulemakings. I created an NRC Rulemaking Tracker to help the public stay informed and engaged. It updates multiple times a day, and opportunities to submit public comments are highlighted. You can even download a pdf!

Screenshot of the NRC Rulemaking tracker

Dr. Adam Stein @astein.bsky.social · 05/03/2026

The Nuclear Regulatory Commission voted to issue the Construction Permit for TerraPower's first plant in Wyoming. The review was completed well ahead of schedule. This is a very positive step forward. 🔌💡 Read our comment here: thebreakthrough.org/press/releas...

Timeline of NRC application review for Kemmerer Unit 1.

163

Dr. Adam Stein @astein.bsky.social · 02/03/2026

Two new papers claim nuclear plants increase cancer risk. That sounds alarming. But they are not studies of radiation exposure. They are studies of geographic proximity and do not provide valid results. 🔌💡

Landscape of a community near a nuclear power plant.

284

Markets don’t form around spending. They form around contracts.

010

Until the government actually buys fuel, enrichment capacity will remain hesitant—and improvised workarounds will persist.

100

That’s why Congress created the HALEU availability program: to aggregate demand and act as the first buyer.

Nuclear fuel markets are thin, capital-intensive, and timing-sensitive. Announcements don’t reduce risk. Contracts do.

100

The real problem is offtake. No one will produce fuel at scale without durable, early purchase commitments.

100

Why Billions for Uranium Enrichment Still Won’t Deliver Nuclear Fuel

DOE’s $2.7B for uranium enrichment is necessary, but it won’t solve the HALEU problem. The constraint was never technical capacity. It’s the absence of early, durable offtake. The U.S. can already produce HALEU at pilot scale. New piece www.breakthroughjournal.org/p/why-billio... 🔌💡

breakthroughjournal.org

Despite new federal spending, the missing ingredient for advanced nuclear power remains a buyer willing to commit early

241

Dr. Adam Stein @astein.bsky.social · 09/01/2026

Ho Nieh was designated as the new Chairman of the Nuclear Regulatory Commission. thebreakthrough.org/press/releas...

010

Dr. Adam Stein @astein.bsky.social · 02/01/2026

A credible nuclear expansion will not be constrained by physics or safety data. It will be constrained by whether our decision frameworks can stop chasing statistical ghosts—and start governing real risk.

240

Dr. Adam Stein @astein.bsky.social · 02/01/2026

Complex energy systems succeed when regulation, science, and capital allocation are aligned around risks that are measurable, material, and tractable. When they aren’t, we delay deployment, inflate perceived risk, and erode public trust—without improving safety.

Dr. Adam Stein @astein.bsky.social · 02/01/2026

The opportunity cost is real. A multi-decade, multi-billion$ national cancer study would divert resources from hazards we know how to reduce—like radon exposure, which causes ~22,000 lung cancer deaths annually in the U.S.