iiinova — where your digital self keeps evolving

Anders Sandberg @arenamontanus.wsocial.eu · 19/10/2026

Claude and me had a lot of fun making this. Everything is based on real data or simulations. Fixing heartbreak is easier when you can recompile it.

120 2

Anders Sandberg @arenamontanus.wsocial.eu · 06/10/2026

I think @andymasley.bsky.social has a valid take over on the other site:

Tweet: "Basically any belief you pick about AI has a bizarre amount of money behind it and you kinda gotta pick your poison."

010 1

Anders Sandberg @arenamontanus.wsocial.eu · 06/10/2026

The closest I got to seeing this kind of messy dynamics was when I did an OPSEC seminar at FHI in 2017, noticing that as AI was becoming strategic we should be a bit concerned... but most focus was elsewhere.

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 21/09/2026

I suspect these add-ons will be equally popular.

000 0

Anders Sandberg @arenamontanus.wsocial.eu · 20/09/2026

This goes back to an older paper of ours, where we noted that lab automation may make the risk chain change character. We were concerned with low-power human actors, but now we are getting AI actors (of increasing power too!) that can aid them. journals.sagepub.com/doi/full/10....

Figure 4 of (Sandberg & Nelson 2020), showing probability curves of successful bio-attack as function of agent power as the number of steps in the risk chain are reduced. The 7 step base case is a rough normal distribution around mid-power agents; as the steps are reduced more and more probability moves over to the weakest agents, until they eventually dominate.

011 1

Anders Sandberg @arenamontanus.wsocial.eu · 10/09/2026

Vi planerarar ju för tänkbara konflikter. Även om man bara tror det är 10% eller 1% av risk för ett krig är det ju rationellt med ett försvar. Och en substantiell försvarspolitik. manifold.markets/SimoneRomeo/...

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 10/09/2026

Frågan är hur mycket sannolikhet som behövs för att börja arbeta på planer för mer radikala scenarios. Om man visste att 8.3% högskoleutbildade blivit arbetslösa 2030 (en garanterad massiv politisk kris) så skulle man göra det. Om det var 50%? Fortfarande rimligt. 5%?

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 10/09/2026

Och många har åsikter om kapitalvinsterna relativt lönerna. Vilka går upp rejält i det substantiella scenariot, och mycket mer i extremscenariot.

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 10/09/2026

I deras modell skjuter arbetslöshet bland knowledge workers i höjden i extremscenariot (medan andra grupper inte påverkas lika mycket). Även deras substantiella scenario känns. Givet Sveriges ekonomi, och att knowledge workers har mycket politisk röst, så känns det som en väldigt relevant risk.

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 10/09/2026

I de mer extrema scenarios dubblar ekonomin var 4.5e år. Det skulle nog ha en del effekter på statsfinanserna.

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 10/09/2026

Det är synd att svenska valkampanjen inte tar upp så mycket AI. Även om man ignorerar xrisk är ekonomi relevant. Om man kollar in den här modellen så är det en märkbar chans för en rejäl förändring av tillväxt och knowledge work *under mandatperioden*. www.anthropic.com/institute/ec...

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 26/08/2026

I really liked the paper, but I was curious about the data. So I quickly vibe-coded an interactive interface.

110 1

Anders Sandberg @arenamontanus.wsocial.eu · 26/08/2026

There are however synergistic factors that can be improved: governance quality (democratic institutions, low inequality, state capacity, low corruption, and social cohesion are protective), decentralization, preparedness, food system resilience.

Summary of GCBR Resilience Factors
In addition to the general features identified in Section 3.1.6, GCBR resilience could be increased by the following factors:
Better health system in general in terms of financing, capacity and infrastructure.
Universal healthcare.
A higher value of the Global Health Security Index (the highest values are in Canada, the United States, Spain, France, United Kingdom, Germany, Denmark, Finland, Norway, Sweden, Thailand, Australia, New Zealand, Japan and South Korea).
A more healthy population.
Being an island or very remote, especially if tight border control is enforceable.
Previous experience with biological threats.
Public confidence and respect for social rules and institutions.
Less corruption and solid rule of law.
Secured supply chains.
Decentralized systems (in most cases).
These factors conflict less with the general factors from Section 3.1.6 than for the other GCRs but there is still some tension:
Universal health care and strong health systems require substantial resources, which are usually not available in countries which are remote.

110 1

Anders Sandberg @arenamontanus.wsocial.eu · 26/08/2026

The key point is that there is no best place to be in a global disaster: some are better in some circumstances, but none are good against all GCRs.

Diagram showing the top 5 countries by resilience factors - Australia, New Zealand, Japan, Switzerland, Argentina - and that all of them have positive resilience factors and negative vulnerabilities.

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 17/08/2026

As a Swede, I know what happens when the king starts redesigning ships. Vasa sailed one mile on its maiden voyage, then it sank. www.theguardian.com/us-news/2026... www.simscale.com/blog/vasa-sh...

The Guardian headline: Trump orders Pentagon to redesign US aircraft carrier to use steam catapults to launch jets.

On the 16th of January, 1625, that King Gustav II Adolph of Sweden directed Admiral Fleming to sign a contract with the shipbuilders of Stockholm (Hendrik and Arend Hybertsson) to build four ships. Two smaller ships (108-ft keel length) and two bigger ships (135-ft keel length) were to be built over the course of four years. In the months that followed, King Gustav changed his orders several times, leading to total chaos and confusion for the builders.

290 9

Anders Sandberg @arenamontanus.wsocial.eu · 06/08/2026

The current AI-cybersecurity situation is worrying: that frontier AI identifies lots of vulnerabilities, allowing patching, is good. But that assumes the patches get used by users - if they lag, they become exponentially more vulnerable. We are in the summer holidays... epoch.ai/data-insight...

Plot of serious cybersecurity vulnerabilities disclosed over time, rising dramatically in 2026. From epoch.ai

110 1

Anders Sandberg @arenamontanus.wsocial.eu · 05/08/2026

Interesting data from a survey showing it is the US political left that currently supports free trade deals between nations. www.theargumentmag.com/p/who-suppor...

Graphics from https://www.theargumentmag.com/p/who-supports-free-trade-everyone showing popularity of free trade agreement between countries across the US political parties, showing that the most positive are to the political left.

000 0

Anders Sandberg @arenamontanus.wsocial.eu · 02/08/2026

OK, I might not be entirely neutral here.

Anders Sandberg wearing a "Clean up Clacton" Count Binface campaign t-shirt.

041 4

Anders Sandberg @arenamontanus.wsocial.eu · 02/08/2026

British politics and British comedy are very close to each other.

Bar chart showing 33% of the public supporting Count Binface, 21% Nigel Farage, 32% Neither, 13% dont't know. From Ipsos.

Public prefer Count Binface to win in Clacton over Nigel Farage

· When asked about the Clacton contest, 1 in 3 British adults (33%) would prefer Count Binface to win, while 21% prefer Nigel Farage and 32% choose neither. 13% don’t know.
· 69% support Count Binface’s policy of restoring the price of a 99p flake to 99p, and 53% back capping croissant prices at £1.
· However, other proposed policies have more middling support, with 36% opposing forcing rule-breaking cyclists to ride unicycles. Perhaps surprisingly, just 32% support abolishing VAR, compared to 21% who support abolishing the football technology.

230 3

Anders Sandberg @arenamontanus.wsocial.eu · 26/07/2026

I am not a serious conjecture-hunter. But I think we are going to scour away a lot of conjectures that could have been solved with existing tools over the next months, leaving us with the ones that actually do require new math.

Plot of distribution of dwell times in the three body problem showing a heavy tail, histograms of ejection velocity and syzygy word lengths for launcher configurations.

000 0

Anders Sandberg @arenamontanus.wsocial.eu · 26/07/2026

I also found an old one, "Can any symbol sequence be realized in the planar three-body problem?" that had actually been answered by Moeckel & Montgomery since it was posed... but Fable and me realized it has implications for some of the stellar rearrangement in Grand Futures.

Graph of a three body problem solution, and what syzygy words the orbit spells out over time.

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 26/07/2026

It is fun to scout around for conjectures to throw AI at. Here is one that got away so far: are there convex polyhedra without nonoverlapping unfoldings? The Barvinok–Ghomi pseudo-edge construction almost disproves it, but Sol failed at turning it into an edge construction.

Part of a convex polyhedron that might not be unfoldable onto a plane without overlaps, based on Barvinok-Ghomi's model updated by ChatGPT Sol.

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 25/07/2026

Strolling around on the MIT campus. One thing I notice is the number of tourists. Families coming to see where mom or dad studied - or where the kid likely will go. Looking at cool machinery or architecture. I have never seen this at other universities; I wish it was more common.

Anders Sandberg in front of the Infinite Corridor of MIT.

120 2

Anders Sandberg @arenamontanus.wsocial.eu · 23/07/2026

Incidentally, a good opportunity to dust off this oldie but a goodie tweet: x.com/anderssandbe...

Tweet saying ""It is only doing it in a few percent of cases."
"Sure, but it is not agential enough to actually succeed."
"Sure, it found the escape hatch but it was fake."
"Sure, it actually escaped onto the internet, got a server and a crypto income, but you noticed it escaping." " by Anders Sandberg, Dec 6 2024

000 0

Anders Sandberg @arenamontanus.wsocial.eu · 23/07/2026

One of the delights of the Jacobian conjecture affair is that I got an excuse to peek into algebraic geometry... and quickly sneak out again, because that stuff is seriously complex!

100 1

Anders Sandberg @arenamontanus.wsocial.eu · 23/07/2026

This is on the level of Lander and Parkin’s paper disproving the sum of powers conjecture by Euler. I suspect counterexamples are particularly suited for tweet-length math. But I look forward to the tweet-length proofs even more.

110 1

Anders Sandberg @arenamontanus.wsocial.eu · 11/07/2026

Normally IVA reports like this are done at a leisurely pace, but here we aimed at a rapid output: it is URGENT. (Next, we will do the deeper work for a more thorough analysis... but first do the obvious!)

000 0