One of the big projects I am working on is exploring and experimenting with strategic agentic thinking, or how AI models handle decision-making under uncertainty. So far, my conclusion is that humanity isn't f'd yet. More thoughts: www.riskgaming.com/p/gpt-54-pl...

Claude is so pleasant, so cooperative, and yet, so incapable. GPT-5.4’s version is far more menacing. Read for more observations from pitting AI agents against each other in live Riskgaming negotiations: www.riskgaming.com/p/gpt-54-pl...

Why is GPT-5.4 so much more secretive in its negotiations while Claude is so much more cooperative? I don’t know, other than it is hilarious that the cultures of the agents match the popular reputations of their parent companies. More observations: www.riskgaming.com/p/gpt-54-pl...

It will never not be miraculous to me that a coding agent proactively solves problems I didn't ask it to fix. So pleasant!

My experience with Claude Code and Codex is just having them fix config bugs is a miraculous gift from the intelligence gods. Given the complexities of modern dev tooling, there's just always some small tweak needed to keep the damn machine running – and they always finds it

😭😭😭😭😭😭 "He said most were waiting and seeing before taking action, but that the rich 'really value the ability to move and work wherever they want, when they want, and when something like the Iran war disrupts it, they don’t like it'." www.ft.com/content/80b...

It's nice to know that NATO is starting to figure out that autonomous warfare is the future. In my discussions last week in London, I am always shocked just how much folks are preparing for yesterday's wars rather than the wars of the future. More: www.bloomberg.com/news/articl...

The combination of AI backlash and crony New York capitalism combines in proposed bills like this one, which would ban chatbots from offering any health or legal advice. These chatbots have already been invaluable to me on mundane issues – I can't imagine going back

AI in the developing world is an angle that far too few are paying attention to. So I appreciate Lucy Quaggin covering the subject for SCMP. A few quotes and a link below. First, I talk about why US AI companies are paying attention to Africa at all:

It’s weird how, as a writer in say 2002, the only way to reach people in bulk was by collecting email addresses. And in 2026 the only method is… collecting email addresses

Converting a Riskgaming scoring spreadsheet (and all of its formulas) into Python code using Claude Code … and apparently Claude Code can't handle spreadsheets larger than 256kb? Am I missing something?

Fight punct junk - maintain standards!

FWIW, this is also why I think some of these publications are making a mistake when they opt-out of AI scans, since AI is generating more traffic instead of search. But it's also why I always focus on direct traffic to @techdirt.com more than any referrer.

Catastrophic, publication-ending numbers

Google is choking the internet to death, and it's not an accident. Like all the big tech firms, they want to be the only service you interact with at all times.

I'd also suggest that unless you already know a publication exists it can be really difficult find a real one in a sea of AI slop sites. To say nothing of formerly real sites that have been gutted and turned into AI zombies.

The only subscription pubs that work are either a) hobby pubs (passion) or b) finance-related (profits). The evidence from NYT is that it is all games (entertainment) or cooking (passion) and barely news that draws people in

No discussion of tech media can get past this basic traffic fact: in the AI world, Google and social no longer refer traffic, which means that the vast majority of readers just never find you in the first place. Analysis: growtika.com/blog/tech-m...

Spending the afternoon trying to convert my first Riskgaming scenario Hampton at the Cross-Roads into an agentic AI negotiation exercise: www.luxcapital.com/riskgaming/...

How did European and UK policymakers perform in Riskgaming at Chatham House? Well, they fell into the trap of European space sovereignty just like in their everyday government positions. Interestingly, they were *more* dependent on SpaceX by the end then our betatest cohorts

The whole Lux Riskgaming crew had a blast hosting Dead Reckoning, our scenario on European space sovereignty at Chatham House's annual defense forum. Not the best outcome for the group (or Europe, for that matter), but lots of fun and learning all around

Had a whole hell of a lot of fun hosting Laurence Pevsner's new Riskgaming scenario Dead Reckoning in beta this past week. European space trilemma: space capabilities, autonomy or jobs? The answer was … none, basically, in both runthroughs we ran

This is a great article by Taylor Rayne in Asimov Press on the history of scent, the incredible challenges of mapping out its complexities, and how new startups like Osmo are finding novel applications for AI in digital olfaction.

One thing I love about design is that you can never be done. There is always something to improve in a product, technology, novel, etc. Just made a massive improvement to my game Hampton at the Cross-Roads from a suggestion – why didn't I think of this great improvement before?

I just learned that the U.S. is 30th in the world for public bathrooms per capita. Can we get a Make America Poop Again (MAPA) initiative going? (for those curious, Iceland is apparently number one)

New post: Why adversarial ambiguity makes prediction markets like Polymarket and Kalshi extraordinarily challenging to operate. Words like "invasion" and "war" are intentionally avoided in order to not trigger their consequences. Example: The U.S. hasn't declared war since WWII

New post from me: why countries like Korea have learned the art of the straddle and how to extract tech, wealth and influence from the great powers without getting pulverized. Link below:

First essay from me for 2026: "Our 'just take it' era" – The Hobbesian nightmare stealing positive-sum growth, covering Venezuela, California's wealth tax, my home state of Minnesota's welfare fraud, Park Chan-wook, Battle Royale and venture growth. Link below:

After 15 years in and around VC, I still have one person to thank for getting me started, and that's Niko Bonatsos. He "discovered" me, got me my first job, mentored me and has remained a great friend. Few are made like him in VC. Excited for his next steps: www.wsj.com/articles/ea...

41 books, 26 movies, 32 live theater performances and thousands of essays later, these are my culled favorites from 2025: www.dannycrichton.com/2025/12/31/...

Check out your top reads of the year: including parties, lying, AI, Zelda, Legos and more. riskgaming.substack.com/p/your-top-...

This is a good read on the AWS outage and how such events will become more common. I'd agree that's probably inevitable and individually we can only seek to "survive" them - but I wonder if they'll happen often enough that companies will see uptime during them as a competitive advantage?

Top reader click of the year: JA Westenberg’s “Boarding Group One,” on how airline boarding groups have become a perfect microcosm of our status-obsessed culture. www.riskgaming.com/i/181733882...

"Throughout history it’s been said that mathematics is the language of nature, but I couldn’t even grasp something as simple as the relationship between the language of mathematics and the laws of the physical world." Close out your year with Karl Ove Knausgård. www.riskgaming.com/i/181733882...

Like Spotify Wrapped, but for Riskgamers: riskgaming.substack.com/p/your-top-...

LegoGPT is an AI tool to help you generate LEGO structures from text prompts. What would you build first? youtu.be/b4Bq1EF1EB4

Just watch. It doesn't even matter if you're a Zelda fan. youtu.be/gcS1HIci4hQ

Din Tai Fung, which debuted in the United States as a family-run suburban LA strip mall restaurant 25 years ago, is now America’s highest-earning restaurant chain. www.riskgaming.com/i/181733882...

Making the stock equity of defense primes less valuable is a very weak and roundabout way to push them to be more efficient. Perhaps just holding them accountable to the costs and timelines they sign on the dotted line in procurement would be a lot easier?

Airport lounges may be the truest mark of modernity there is. www.riskgaming.com/i/181349587...

World models are extremely important to the future of AI. They allow for rapid and parallel learning, and can help AI agents learn edge cases faster and more realistically than other approaches. Excited to see Runway launch their first world model: techcrunch.com/2025/12/11/...

The future of the biolab and the life sciences is parallelized automation. We shouldn't be running one experiment at a time, but rather thousands all at once. That's why I am so excited for what Medra is building. Excited they have $52M to keep building: www.bloomberg.com/news/articl...

How much of our sleep problems can really be blamed on light pollution? www.riskgaming.com/i/180714225...

Fix your life by fixing typewriters? Appealing www.riskgaming.com/i/180714225...

Zen and the art of typewriter maintenance from the NYT: www.riskgaming.com/i/180714225...

We're already losing the AI battle. www.riskgaming.com/i/180714225...

A former FBI agent just spent years tracking down a North Korean hacker who ransomed a Kansas hospital. Then he offered them a job building a crypto casino. www.riskgaming.com/i/179468812...

Check out Brian Moore's new project: Every minute, nine different AI models are asked to generate a new analog clock showing the current time. The results have me puzzling over what, exactly, makes this exercise so complicated. www.riskgaming.com/i/179468812...

Internet writing needs more feedback. If you’re an intellectual and someone tells you you’re wrong, it galvanizes you, it goads you. www.riskgaming.com/p/how-to-be...

Lessons from a new wargame: By forcing players to take on the Chinese perspective, the exercise exposed a blind spot in U.S. military planning — we neglect China’s incentive to pursue coercive strategies that exploit American political hesitancy www.riskgaming.com/i/178807966...