That classic argument for step up is that a farmer’s son inheriting the farm suddenly owes a lot of taxes which he can’t pay without selling the farm.

Is this a real problem? And how do we fix that?

It is a real problem but its presented in a misleading way.

If you inherit a farm worth 1m and you have to pay estate taxes on that of 100k. You do not have to sell the farm, you can instead take out a mortgage to cover that 100k.

When we frame it that we its still very fair to the person inheriting the farm because who wouldnt want a $1m farm for 100k.

Except you don't pay any estate tax unless your estate is larger than $15 mil

In American maybe thats the case. Im talking about situations where we are trying to implement an estate tax and the discussion is framed as basically forcing the receiver to sell the estate to pay the tax instead of presenting it accurately as taking a small loan to pay the tax.

You fix it by sidestepping it. Don't do it, even though it is the simplest/easiest to implement, and instead create 5 other laws that surround it and accomplish the same goal. Loop the hole back a them.

The key flaw in the argument is that high quality interfaces do not spring to life by themselves. They are produced as an artifact of numerous iterations over both interfacing components. This does not invalidate the entire article, but interface discovery phase has to become the integral part of disposable systems engineering.

I agree with your comment. but in defense of the articles scope: as someone trying to build longterm utility and developer infrastructure, I find the vibecoded prototypes to help me dial in on interfaces. Instead of waiting to hear back from only downstream user experiences (because you previously didn't have time to experience the various consequences of your choices yourself), you can get the various experiences under your own belt, and better tune the interfaces :)

Which is to say, the interface exploration comes part-and-parcel with the agent tooling in my experience

This comment is a little orthogonal to the content of the article, but my experience made what they wrote click with me

I found it a lot safer to cherry pick into a new temp branch, test things out, rename the old branch to an archive name, and rename the new temp branch to old name.

That is until I started using graphite, that solved the problem completely for me. The only trick is to never mix graphite and git history editing.

If you have not already, try Graphite. You will be delighted as it serves that exact purpose.

I use magit which afaik is still undefeated for this workflow. Particularly with this snippet to “pop down” individual ranges of changes from a commit: https://br0g.0brg.net/notes/2026-01-13T09:49:00-0500.html .

If a company choses Cloudflare they would have great service everywhere except Italy. If they chose a service with lower quality / reach, they will suffer degraded service across the board. If they try to use more than one CDN that’s a lot of hassle.

It’s not clear which way the decisions will go in reality. Past experience suggests that tech companies eventually accommodate local laws, trading complexity of explaining this to customers for complexity of implementing targeted blocking tools.

A lot depends on the next couple of months and the US's continued belligerence against - former? - allies. If that isn't toned down, and drastically so then I expect there to be many more consequences than just for CDN providers.

> When a model exhibits hallucination, often providing more context and evidence will dispel it,

I usually have the opposite experience. One a model goes off the rails it becomes harder and harder to steer and after a few corrective prompts they stop working and it’s time for a new context.

Once it’s in the context window the model invariably steers crazy. Llms cannot handle the “don’t think of an elephant” requirement.

It depends.

It's a natural inclination for all LLMs, rooted in pre-training. But you can train them out of it some. Or not.

Google doesn't know how to do it to save their lives. Other frontier labs are better at it, but none are perfect as of yet.

Blending is socially is very useful for the individual, i don’t agree with the author here.

Society itself may benefit from cohesion or from truth depending on circumstances.

> maybe you have examples of historical breakthroughs in AI Alignment that didn't involve doing / understanding the mathematical concepts I mentioned in the previous paragraph?

Multi agentic systems appear to have strong potential. Will that work out? I don’t know. But I know the potential there.

How was your experience with Xbox? I heard it was rather watertight?

Why would I ever pay for anything microsoft made?

It’s not just career. Any artistic endeavor suffers from competition with the past, which eventually becomes a guaranteed loss. Negating the frame becomes the primary way to leave a mark.

But for actual art, such as music, constantly doing novel things is kind of the point. Every generation needs its own music that runs counter to what came before.

User interfaces are not art.

> User interfaces are not art.

Do UI designers think that way?

I imagine some see it as engineering - make things work efficiently for the users. Others see it as art. The outcome will depend on which group gains the upper hand.

There's some linguistic ambiguity here if we just say "art", because it includes things we might divide into "artistic choice" versus "craftsmanship", ex:

1. "Picasso, that's the wrong way to depict a human nose."

2. "Picasso, that's the wrong material, that vibrant paint is poisonous and will turn to black flakes within the year and the frame will become warped."

I interpret parent-poster's "interfaces are not art" as meaning they're mostly craftsmanship.

It may not be quantifiable enough to be labeled "engineering", but it's still much less-subjective and more goal-oriented than the "pure art" portion. All these interfaces need to be useful to tasks in the long term. (Or at least long enough for an unscrupulous vendor to take the money and run.)