Douwe Osinga

@HenrM_Cruz Remember Apple's Ping?

Remember Google Bard? Google closed at $99 the day Wall Street decided it had lost the AI race. The stock has 4x'd since then. Market vibes.

@cosminnegruseri I think you see this a lot all over. We had some test that was more performative than qualitative but worked because it was proof of work. Now the AI can do the work, so the proof is gone. Same with resumes and linkedin posts being throughleadery -- ugh.

as much as they are hated now, people will miss leetcode interviews

2.6B open source model generating up to a minute high quality video from one frame? That's insane! Maybe there is something to these world models after all. Can't wait to get my hands on the actual model nvlabs.github.io/Sana/WM/

well, I'm off google maps API now: theregister.com/ai-ml/2026/05/… -- it seems so weird to me that the default is that a maps API key can also be used for expensive inference and that @Google doesn't allow keys without billing and no budget

I wrote a post: blog.douwe.com/2026/05/the-we… Every LLM contains a wildly lossy “compression” of the web. Even for URLs that never existed. I built a browser for it: the Web of Babel.

The three most valuable companies today ($NVDA, $AAPL, $GOOG) raised around 50M together. Now Anthropic raises 10x that and it barely makes the news. Interesting times. techcrunch.com/2026/04/20/ant… /cc @Julie188

I'm going to be in SF from Wednesday on - come and hangout at luma.com/75auv0v7 - I'll talk a bit about some of the new ideas about automatic wiki writing

goose moved! we're now part of the linux foundation: github.com/aaif-goose/goo…

This is what I'll be working on now!

@jorgelison Have you seen block.github.io/goose/ ?

Oh, yeah, also to update you, @NeptyneHQ got acquired by @blocks and I now am leading the Open Source Goose team (github.com/block/goose)

In 1955 it would have cost 1.5M dollars to buy solar cells to power your home in Palo Alto. Now that is 8 000 dollars. Of course now the home costs 1.5M dollars and in 1955 it was ... 8 000 dollars.

This seems like a crazy story: sherwood.news/power/who-died… But then again, if you leave your money to a charity, are you really sure they do a better job doing something useful with it than the government? And if that is so, shouldn't the government funnel more money to charities?

@davemcclure The trade-off is always (also w military service) that forcing people to work on something for low pay is economically almost always worse than hiring people and taxing the people that are now more productive since they work on something they know about.

imagine: instead / in addition to mandatory #military service (S Korea, Brazil, Nordics, SG, Taiwan, Turkey, Israel), what if countries rqd #civic service for ~2-4 years? 🤔🪖👷🏼‍♀️ public/civic service examples: - education - healthcare - police / fire - transportation - power / water / garbage (utilities)

One way to show that upcoming election is really almost a toss-up, is to treat all the swing states as toss-ups and count the number of combinations that will have either party win. 55% vs 44% is the outcome. Similar to @NateSilver538 's model: #gid=0" target="_blank" rel="nofollow noopener">docs.google.com/spreadsheets/d…

@ashevat But that quote starts with: "if they insist on trying to time their participation in equities" I think Warren Buffet would mostly want you to buy the stocks you like and ignore the short term markets

For tomorrow's market, remember: When people are fearful - be greedy. When people are greedy - be fearful.

Who's the weird candidate for the presidential elections now? bbc.com/news/articles/…

@Felix3S This doesn't seem convincing. The cost of debt is determined by the interest rates countries pay which is set by demand and supply of money. Are you saying climate change will make me earn more money on my savings accounts? Could be nice in a graying society

Rising Temperatures, Falling Ratings: The Effect of Climate Change on Sovereign Creditworthiness pubsonline.informs.org/doi/10.1287/mn…

@wildiris19 That seems like a peculiar argument. Nature is really efficient for sure, but it has to given evolutionary pressure. Humans building AGI face no such restrictions, so why not blow 10B on it if it gets us there faster?

• Hopefully most can agree that a human qualifies as a true autonomous general intelligence. • The human brain is about 1.25-liter in volume, around 1.5 kg in mass, and runs on about 25 watts. • The human genome represents a maximum of about 725 megabytes of information. • Starting with this paltry amount of data storage, nature by a recursive process of cell division and subsequent specialization, over the course of several dozen generations, can build up a true autonomous general intelligence. • True autonomous AGI will arrive that day that someone finally sees the way through to recreate using technology what nature does with biology. • This is not a matter of $100M+ training sets running on systems with the power requirements of a small country. It won’t be LLMs that become the nursery for AGI. Meta-AI and Open-AI will not be part of the AGI revolution. • The path to true autonomous AGI goes through embodiment. And it will be the robotics community that finally pushes AGI over the finish line.

Tried to cancel my @Verizon internet. Took two hours on the chat and then they pointed me to a website where I can do it myself. Which gave me an error saying to contact Verizon. They asked me twice if I didn't want to transfer my account to a friend. Anyone interested?