jgalt212

That's the irony of EVs. They're most useful in cities--short driving distances and less emissions in dense areas, but easiest for the consumer to use in the burbs where driving distances are longer and emissions are less of an issue.

Matt Levine discussed this relevant NY Times article.

https://www.nytimes.com/2026/09/19/business/draftkings-ai.ht...

Pull quote:

> Ms. Kalani said that company leaders made a “collective decision” not to use predictive technology for problem gambling. “We evaluated that it wasn’t evidence-based,” she said.

Mackie acted in Hacks, Tales of the City, and The Love Boat.

https://www.imdb.com/name/nm0533496/

you must have tokens to burn.

That rapture is coming too. So I guess it will be a footrace.

> so I can better understand the change and alternatives considered. I don’t do this very often,

# YOLO mode

The European Commission strikes again!

> it'll keep running as long as anyone anywhere is willing to make money by renting hardware to it.

You just admitted it's controllable.

fair enough, but in my experience fast models seem to be 5-10X faster than slow models.

> in terms of functionality and value to users, software ought to be malleable and composable.

For technically competent users, sure. but for everyone else the UI should just follow the principal of least surprise.

fast results that you need to verify are better than slow (allegedly better) results that you still need to verify. REPL vs batch.

Information wants to be free. But other than AI slop (where the tokens have become too cheap to meter), information costs money to produce. Pay with you info or your money. It's amazing how many people just cannot come to grips with that concept.

Sounds a bit disproportionate. If you do that, you better hope the vendor's CEO and your company's CEO are not golfing buddies.

And in my mind, one email alert or pop-up to all current users letting them know the service is going away (in case they had not been told which is as I've said before often the case in large orgs) is very far from the definition of spam.

I don't see a problem with this. B2B2C is a tricky dance with many stakeholders whose business needs are often ignored. If procurement cancels a service that users need without telling the users, which they have been known to do, that's a big problem.

If I use Claude or OpenAI swarm to commit crimes, and the vendor knows I'm up to no good, what's their liability? Do they plan to invoke the phone company defense?

> This isn't even remotely true

That's quite an extraordinary claim there which I doubt would hold up to an even cursory level of scrutiny. But if you yell it loud enough, maybe people will be afraid to challenge your assertion.

There is really is no excuse for Washington here. Even for the layperson+, it doesn't take much use of an LLM to figure out where they produce good and usable results and where it's of specious of value.

+ every layperson is an expert in 1 or more areas.

is it easier for an LLM to figure out a CLI or a REST API?

> I had Claude port CADO-NFS to run on GPUs. Then it orchestrated a fleet to run on scavenged idle capacity

Is it easier to find unused GPUs than unused CPUs?

> Isn’t the value in being able to verify the design before implementing?

Fair enough, but to be truly safe you really need the whole pipeline connected. model -> validation -> emitted code.

> Several of the new admissions, however, run counter to OpenAI’s fair use defense, particularly the rule’s requirement that use doesn’t substitute or harm the market for the original work.

Sounds about right for the people and orgs involved.

> as very confusing for me to understand how you can trust the implementation unless you can compile TLA+ to C.

Preach. I feel like I'm taking crazy pills every time someone claims TLA+ as the sine qua non of provably correct systems. They should sub probably for provably.

> It takes less time for local model to take the first action on your task than it does for Claude to validate your login, put you into queue and start issuing the commands.

This is only true if your local model is already resident in RAM / VRAM.

> They’re hyping up their products as being able to destroy the human race then are somehow surprised at the level of anti-AI sentiment out there in the world.

You hit the proverbial nail on the head there!

Yes, these are very clever fuzzers. But the real problem is not the cleverness, but the willingness to spin up 100s or 1000s of subagents no questions asked. Must make those token numbers go up!

Geographic Determinism

https://www.litcharts.com/lit/guns-germs-and-steel/themes/ge...

That's true, but then again, CNN is a mere husk of its former self. Something needs to change, but maybe it's too late for profit-driven news media.

> All Oracle is doing is unpacking Nvidia servers and plugging them in. This is a commodity business

But it's not if you can jump to the front of the GPU allocation line like Musk has done.

RL works better / easier to implement with static typed languages.

book pitch:

On being sane in insane places: A tale from inside the marketing department of a frontier AI Lab.