Models are trained on human outputs. It’s not super surprising to me that inputs following encouraging patterns product better results outputs; much of the training material reflects that.

I think we need a way to verify the specs. A combo of formal logic and adversarial thinking (probably from LLMs) that will produce an exhaustive list of everything the program will do, and everything it won’t do, and everything that is underspecified.

Still not quite sure what it looks like, but if you stipulate that program generation will be provable, it pushes the correctness challenge up to the spec (and once we solve that, it’ll be pushed up to the requirements…)

Fewer and fewer people can make a decent living with traditional work. Hence, my theory goes, the rise of actual lotteries along with influencers, injury lawyers, and schemes like New Orleans.

Something is seriously wrong when family members hope an elderly relative will die on the hospital so they can get a payout, or when people are crashing into trucks or promoting BS snake oil on instagram.

It’s an indictment of the people involved for sure, but our social and economic systems have created the perverse incentives that these people are betting on. And it seems to be accelerating.

But I lean the other way with app stores. The companies hire reviewers, the listings appear in the App Store trade dress, it feels more like a museum or magazine than an ISP. But I get how reasonable people can disagree.

Maybe we need some formal choices: is this a curated App Store that reflects editorial judgment (in which case it must be possible to ship alternatives on equal footing), or is it a common carrier (in which case you can be the only game in town).

The ambiguity doesn’t help, and of course megacorps love shifting the frames depending on context.

https://www.justice.gov/atr/case/us-and-plaintiff-states-v-g...

https://www.justice.gov/opa/pr/department-justice-prevails-l...

https://www.justice.gov/atr/case/us-and-plaintiff-states-v-a...

https://en.wikipedia.org/wiki/Antitrust_cases_against_Google...

https://www.cnbc.com/2025/03/19/eu-orders-apple-to-open-up-a...

…and there are many more.

You can say those aren’t enough, but it is 100% fallacious to say there has been zero antitrust actions against Apple and Google since 2000.

That and I don’t see how Google and Apple can both be monopolies in mobile. Is this the “Ford has a monopoly on Mustangs” argument? Never found that persuasive.

Now, reframe as duopoly, and maybe layer in that a platform owner who curates their App Store must allow alternative app stores on equal footing, and I’d be with you.

But people who go > 5 minutes between prompts and see no cache, usage is eaten up quickly. Especially passing in hundreds of thousands of tokens of conversation history.

I know my quote goes a lot further when I sit down and keep sessions active, and much less far when I’m distracted and let it sit for 10+ minutes between queries.

It’s a guess. But n=1 and possible confirmation bias noted, it’s what I’m seeing.

Claude code uses a bunch if best practices to maximize cache hit rate. Third party harnesses are hit or miss, so often use a lot more tokens for the same task.

Or maybe RAEND

There might be a market for your alternative though. Should be easy enough to build with Claude Code.