iiinova — where your digital self keeps evolving

To give an example I know: Trump is claiming an "energy emergency", and using it to order coal plants to stay open. When the order expires, he simply issues a second. Defining "emergency" as an actual power shortage, eliminating the extension option, etc. are all simple legislative fixes.

060

We also need to think of Trump II as the equivalent of IT penetration testing, but for democracy. Someone's just gone through and identified lots of possible exploits. How can we shut them down while preserving the useful stuff?

1100

Needhi Bhalla 💅🏽 @needhibhalla.bsky.social · 29/09/2026

Thinking about this a bit: is imposter syndrome the antithesis of Dunning-Kruger?

261

Reposted by John Timmer

The wrong fucking people have imposter syndrome

28816

John Timmer @jtimmer.bsky.social · 28/09/2026

One of the likely results of the midterms will be that we won't have to listen to this yahoo mindlessly lying to us for at least two years. And I am here for that.

180

I don't know what this says about how Anthropic identified these in the first place, but I'd be surprised if the failure to repeat isn't significant in that regard.

030

John Timmer @jtimmer.bsky.social · 28/09/2026

An interesting thing that may be relevant here: If you go to Anthropic's pre-print, they say that they re-ran the analysis and "the array was missed in every rerun." (The array being a diagnostic series of repeats.) So, for whatever reason, the AI mostly misses these.

www-cdn.anthropic.com

161

Reposted by John Timmer

Carl Zimmer @carlzimmer.com · 27/09/2026

Four days ago, a scientist was surprised by the news that Anthropic discovered new virus genes for making DNA. He says he’s been studying them for years—and feeding his data to Anthropic’s AI as part of his own research. Coincidence? Here’s my story. nyti.ms/3VSWayc

nyti.ms

Did Anthropic’s A.I. Really Make a Scientific Discovery on Its Own? (Gift Article)

An expert at the University of Copenhagen said his team had been sharing its research with the company’s A.I. model, Claude, and that its new finding matched their work.

882123873

John Timmer @jtimmer.bsky.social · 27/09/2026

A great story about one of the cooler organisms on the planet - Paulinella, which I didn't even know existed as recently as a decade or so ago.

1121

Wind is also about two years away from matching coal. Two major offshore wind farms (including VA's 2.6GW farm) should come online later this year, which should give it a boost. So, there's a good chance that coal will drop to 5th place behind nat gas, nuclear, and the two big renewables.

040

Looking over trends a bit more carefully: At growth rates similar to this year, two years from now, utility scale solar alone will be very close to total coal generation. With rooftop thrown in, it will be producing more energy (though not all of that will show up on the grid).

140

(I'm not sure how many people outside of biology realize how nightmarish it is to face the all-too-common situation of "here's a protein where we have absolutely no clue about its function, and you have to figure it out.")

080

None of those are very easy questions to answer, which is probably why Anthropic has not done any of them. But if they want this story to be anything more than "AI spots weird pattern", then they'll either have to dive in, or wait for actual biologists to extend their work.

140

So, as a lot of people are saying, there is a lot of biology to do here: Are these repeat RNA's processed? If so, by what? Is this actually a reverse transcriptase? Are the accessory proteins part of a system, or just happens to be nearby due to genome constraints? What do the accessories do?

(I know nothing about this phage family, so don't know how compact or rigorously structured their genomes tend to be. So I don't know how unexpected finding these genes or larger non-coding regions is here. Anthropic's draft doesn't tell us, either.)

There are at least four separate types of accessory proteins. Only one of these has sequence homologies that suggest a potential function. None of them have been tested for any function.

The repeats are associated with a reverse transcriptase homolog and an associated protein. They have not confirmed that this is a functional RT except by noting the preservation of catalytic residues.

The repeats are found in a group of phages, and get made into processed RNAs after infection. Not clear if that's functional or if the secondary structure of the RNA just leads to consistent digestion products.

That's the sort of thing that AIs are almost certainly much better than humans at. But if they repeated the analysis, the AI generally missed them. So not obvious why it got lucky here?

130

Ok, I've quickly gone through the technical doc Anthropic released about its newly-found repeat associated protein. A short summary thread: The AI did a great job of identifying a pattern that humans could easily have overlooked (relatively short repeats with lots of intervening DNA).

180

Aaron Rupar @atrupar.com · 24/09/2026

For the first seven months of 2026, the US saw photovoltaics (both utility and small-scale) generate 72 percent of the electricity produce by burning coal. Barring a dramatic change in trends, we're probably about two years from solar passing coal for annual generation. Data from US EIA.

1171

Reposted by John Timmer

OMG -- under questioning from Hassan, Trump's FDA nominee Heidi Overton refuses to acknowledge that Trump can ever be wrong about anything even after she's confronted with a specific example of something Trump is clearly wrong about

59350241758

John Timmer @jtimmer.bsky.social · 24/09/2026

Exactly why I mentioned objective standards.

020

John Timmer @jtimmer.bsky.social · 24/09/2026

Between him and Feinstein, it's pretty clear that we need a rough equivalent of the 25th amendment for Congress. Except one with some objective standards so that a congressional majority can't just ignore the problems.

2132

John Timmer @jtimmer.bsky.social · 23/09/2026

So, I turned my chaotic little developmental biology rant into an actual article that is hopefully somewhat less chaotic and a bit more approachable.

68432

Yeah, I expect that this will be blocked by the lower courts (which is what that sentence is about) and, if we're lucky, will take long enough to get to the Supreme Court that we'll be looking at the end of the term.

110

I should hope so - my personal favorite among my handful of papers was published there.

021

Appreciating Bluesky this morning after having written a developmental biology lecture in the form of a rant and waking up to find it's one of the more popular things I've ever posted here.

1150

Oh, god, I had forgotten all about rhombomeres.

030

Relief to hear that, given I've been out of the field for an embarrassing number of years now.

010