boc

Or if model training is more of a rollercoaster, where spending gets you to the top of the hill where you create a massive internal model which can then build the next version of itself for cheaper and cheaper amounts relative to human R&D costs. If Anthropic is first over that hill, they can race far ahead.

If you go that route, then the US is ahead of China for manufacturing output per capita. Which probably isn't what you wanted to hear.

There's nothing like that in the US, not by any measure.

The whole reason you have entrepreneurs in the US is because people DON'T judge for you failing. You can always start again.

> I tend to notice the poorest, most desperate tend to be the proudest

I think this is where folks go astray. There are tons of well-off or even wealthy people who love America, give to their churches, cheer on their kids at sporting events, and have what appears to an outsider as an unbelievably high amount of nationalism. It's not really the poorest, most desperate slice of society by any measure. Have you lived outside of a major city in the US?

May I present to you the humble briefcase full of cash?

The US manufactures more goods annually than Germany, Japan, and S. Korea combined.

Best part is that he did this all with a controller + pretty slow sensitivity settings. Aka this probably took 2X as long as someone on a PC doing the same thing

> I still remember the day I found out game theory was a thing - it was like finding out that someone successfully systematized common sense.

Game Theory as a field is actually very complex and often reveals dominant strategies that would never occur to someone as the "common sense" approach to a problem. It was arguably created to solve for problems where there was no obvious correct answer, even for a very intelligent person.

I was talking about this with a friend this weekend. We both work in the field and test new models within minutes of them being released. We both immediately clocked Opus 5.5 as being cracked within the first hour. Went on HN and the launch announcement was full of people whining and pointing at cost/token charts vs Chinese models. It was like the upside-down world.

We were both sad that HN has become a negative signal news source on AI lately - you're much more likely to be misled by this website in 2026 on the topic of frontier AI. If you're reading this comment, you should do your own research vs trusting the "Astra is 1000% the best" or "Deepseek is the $/tk KING" comments swarming these announcement posts.

Yeah men with guns can always take it, but there's this thing called the stock market and bond market that would freak out if free enterprise was crushed by state coercion with armed men, with billions worth of IP forcibly exfiltrated by state actors. Trillions get wiped out overnight.

That massive economy is what pays the guys with guns. You can win that battle but you will absolutely kill the golden goose and lose the war.

> and can't seem to fathom that men with guns always override the opinions of nerds in SF.

Those nerds in SF are holding the proverbial gun up to the head of the economy right now. Maybe men with physical guns are bad at understand that type of thing.

The reason I did it the way I did was so that I can still use the codex and claude code subscriptions vs paying the API cost. Can you do that via OpenCode?

So far in the past 20 minutes it sounds much better in my sessions. Way better than 5.0 so far.

I've been getting a ton done with Fable as the supervisor and astra as the implementer, with opus for adversarial reviews of the astra PRs. You can use terminal multiplexers with custom harnesses to allow Fable to start codex sessions and send instructions / read instructions / allow/deny actions. It's pretty cool!

So China's brand new recon satellite that was placed specifically in orbit over Iran just happened to self-destruct weeks after China was accused of giving Iran high-res satellite imagery that was used to kill American soldiers?

And the US just publicly told everyone they have weapons in space?

Hell of a coincidence.

Small reminder that the US government rug-pulled Fable, not Dario. Lots of the safety guards that users find annoying/objectionable were the results of negotiations to get the model back online after the US government forced them to take it down.

Maybe Dario should have just "donated" $1M to Trump's inauguration fund like Altman, Meta, Amazon, Microsoft, Tim Cook, Elon, and Google. There's a reason they are the odd man out with this current Administration.

I set my env up so I can see the exact context used in CC CLI, and then once I get over about 40% ctx used I have it handoff to a new, fresh session. Nothing good comes from running above, say 60% of your context window. Coincidently, I usually have good results with CC. I never compact a session ever.

That assumes they release their models publicly. The future is leaning towards these labs air gapping their best stuff (Model 2, etc) and using it internally to snipe their competitors and charge insane amounts for monitored use in consulting environments.

You can't distill or catch up if you can't access the models. You'll basically have a situation where nation-states will need to try and steal the models Oceans 11 style.

> More importantly, what's the point of building monster-scale data centers on unprecedented amounts of debt when a more than good enough model runs on a GPU from a couple years ago?

Probably because the future "monster" models will be insane. 100T+ param models might be the type of things that can independently run a small business, which means anyone not using them is at a distinct disadvantage to their competitors.

The top model from 2025 looks silly compared to the top model of the first half of 2026. Do you feel like progress has stalled?

Ducklake supports postgres for the catalog, so you get the postgres concurrency benefits + duckdb engine to read the parquet files in the bucket.

It's not 2025 anymore my friend.

Chinese open weight models are great for this turn, but American private models generate orders of magnitude more cashflow. This cashflow = investment in training future models. It's unclear how Chinese open weight companies are going to compete in future rounds if they can't raise the same capital for training runs.

The American business model is exceedingly efficient at building large businesses from zero. I wouldn't dismiss it as just a jobs creation thing.

Kansas genuinely has a ton of paved and gravel roads all across the state. It's like a gigantic grid. Oddly high amount of street-view coverage too on google maps: https://maps.app.goo.gl/EyBFj9BNQa8nhUcE6

The Kansas landscape is very underrated. Just infinite sky in all directions.

One under-discussed topic is whether Americans hate data centers in their neighborhoods, or whether they actually just hate gas turbine generators running 24/7 in their neighborhoods.

You could argue that cheaper/safer local rideshare would boost traffic to local bars and restaurants in denser areas though. If you want to keep the $ local, just add an AV tax that goes to the city.

> Here (NL) contactless payments works with every debit card, physical or stored Apple Google/Pay, in every store

That's how it works in the US as well. You might be conflating Apple Pay with Apple Wallet, which stores your credit cards for tap-to-pay. Most people in the US use Apple Wallet/Google Wallet to digitally store their credit/debit cards, and use those cards to pay at the store by tapping the POS terminal.

For example, I use Apple Wallet for nearly every transaction, but I never use Apple Pay/Apple Card as the payment method.

cmux + Claude Code / Codex with a custom agents.md file.

With cmux your agents can access the browser, spawn full shells, and you aren't locked-in to a specific vendor. It's really powerful but you need to update your instruction file so your coding harness knows it can use the cmux tools to test.

You can also download Handy to use voice so you can just give your feedback via voice into the CLI during the planning phase of a project.

Seems really good so far using it in Claude Code CLI - it gave me a new flag when I asked a question:

"I don't have a reliable way to read that number, so I'd be guessing if I gave you one — and this is exactly the kind of question where a confident guess is worse than none.

What I can tell you is what I actually observe:"

I really like this update - gave me a clear sense of the facts but didn't give me a guess just for the sake of guessing.

One oddity is that it appears to only have a 200K context window right now via CC. Hopefully the 1M version will appear soon!

Interested in how the critics of approaches like this defend an agentic session (with Fable, for example) that stops and runs a multi-hour ML training session. It's a script, so the actual LLM convo goes stale, but then when the results get returned to the main thread you get an expensive cache hit without doing anything.

You would have avoided that cache hit if the LLM session was kept "alive" for those few hours. Why not automate the part where you keep the large main thread alive until you're ready to analyze the results?

The fun part about this is that we can see who is right in about a year. If the leading labs continue making progress at hardening their models against distillation, and then they start pulling away again, we see who is right. If China is able to pass the US and release an independently better model than anything the US has, then your theory is correct.

Both sides have extremely smart people. One side has more $$$ and exclusive access to the best chips. For progress to converge without a corresponding breakthrough suggests there's something else at work.

You're making the point even stronger for US companies to just block EU traffic entirely. If you have no EU customers, the mere act of letting EU visitors on your site might cause an EU country to "make an example" of you for incorrect data handling, but the alternative of just blocking all traffic would let you travel to the EU without issues.