shalmanese

"The three effects are not separable knobs. You cannot dial in one without moving the other two.

The browning claim is the contested one, so it is worth pinning down. "

Is classic Claude-speak.

The paradox is that we love reading our own AI generated writing and hate reading anyone else's AI generated writing.

On a recent weeklong trip to the Philippines, I generated over a 500 page novel's worth of content from AI around various aspects of Filipino history, culture, social dynamics etc. and actually went over it at least 3 times to fully absorb the material.

But if someone handed me even a 3000 word essay on the Philippines clearly written by AI, I would not be able to get to the end of it.

The entire "blame" paradigm is unproductive. Does Lyft "blame" Uber for it's lowered market share?

The entire system (including nurses and technicians) are just agents making semi-rational decisions in their own self interest. Is it important to judge people within an existing system or is it important to look at locus points that, when pressure is applied, can make durable changes to the system?

Americans when describing their ideal car interior: "Imagine a burger ordering button".

> But, for scientists, I find that these tools address the problem of the exploding complexity barrier in the frontier. Every day, it grows harder and harder to contain a mental map of recent relevant progress by simple virtue of the amount being produced.

AI is going to both help and hinder this process though. At the end of the day, mathematics is mostly a social process at this point. The goal is not raw number of theorems proven, it’s how proving theorems affects the working operational models of mathematicians. Only a rare few new theorems in mathematics nowadays have direct real world applicability.

If AI produced legitimate theoretical breakthroughs at a pace mathematicians are unable to absorb, then the impact will be neutral to negative.

That’s not an analyst, that’s a pundit. An analyst can have a clear point of view that is different from yours and, very far off the consensus in any direction. But the value of an analyst is they have a consistent point of view that they apply to any situation and flag as their point of view evolves.

A pundit starts from a pre-declared conclusion and works backwards to generate the argument. An analyst lets the conclusion be dictated by the analysis.

I fully support the government reading my odometer during every single emissions check of my EV ;).

This is the wrong level of analysis. Disney owns ABC, ABC owned 538. The relevant decisions were made by ABC’s leadership.

And the firing of the staff happened years ago and people broadly understood even if they did not agree with it.

The recent decision was to take down an archive that cost $8 in server resources and was still bringing in page views and ad revenue.

Apple's problem might be they were right too early which is sometimes worse than being wrong. The original vision of Siri was substantively correct in how AI would supercharge our phones but huge parts of the vision got forgotten when Siri was acquired by Apple and the original founders left. The original technical choices around Siri constrained it from evolving into something useful.

A funny story that happened the other day: A friend knew he had to be at dinner at a place across town but he forgot why he had to be at that dinner. While we were waiting for his rideshare to come, he was flipping through every kind of app trying to reconstruct the original context for his appointment.

In theory, this is where AI should shine. He should have been able to say "Hey Siri, pull up all of the info that references tonight's dinner appointment" and AI should be the unified interface into a bunch of app-specific data pools.

But of course he's never in 1 million years would have thought about using Siri to do that because of how bad Siri is.

> I think when LLMs first came out people thought they could just say something like, "Make a Facebook clone". But now we're realizing we need to be more exact with our requirements and define things better. That has always been the bottle neck in software.

This was substantially predicted by Fred Brooks in 1986 in the classic No Silver Bullets [1] essay under the sections "Expert Systems" and "Automatic Programming".

In it, he lays out the core features of vibe coding and exactly the experience we are having now with it: Initial success in a few carefully chosen domains and then a reasonable but not ground breaking increase in productivity as it expands outside of those domains.

[1] https://worrydream.com/refs/Brooks_1986_-_No_Silver_Bullet.p...

This is all substantially correct and gives us hints as to where to focus for AI to make the processes go faster.

Eg: I had a product manager say to me that he envisions a future where any meeting with stakeholders that does not result in an interactive prototype by the end of the meeting would be considered a failure. This feels directionally correct to me.

The other thing I expect to see is Vibecoding being the "Excel 2.0" where it allows significant self-serve of building interactive apps that's engaged in a continual war with IT to turn them into something with better security guarantees, proper access control & logging, scalability, change management etc.

But the larger historical point here is that every revolutionary transition produces, in the early stages, "Steam Horses". The invention of the steam engine had people imagining that the future of transportation would involve horse shaped objects, powered by steam, pulling along conventional carts. It wasn't until later developments that we understood the function of transportation as divorced from the form.

I started talking about Steam Horses originally in the context of MOOCs, which was a classic Steam Horse idea.

FYI: the author of this piece is the eugenicist Cremieux who was responsible for using hacked data to attack Zohran Mamdani for checking Black and Asian on his college application.

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

No, you’re thinking of .NET. Or maybe 365. Or perhaps you’re thinking of Live. Or maybe it was ActiveX.

Not streets but an outsider to the Bay Area might be confused that a trip from Richmond to Inner Richmond is like an hour’s drive.

Do you know that Adobe has an entire martech side of the business that is 25% of revenues and growing? I feel highly skeptical of any analysis that try’s to translate their personal experience into a prognosis of business prospects. There are tons of products of a business you never encounter but have enduring competitive moats.

Only for commodity goods does the cost of production impact the price. As substitutionality lessens, the price more and more approaches the value delivered.

But if you wanted to buy an OSX machine with up to 1.5TB of memory, you only had a brief window between when the new Mac Pro was announced and the old Mac Pro was discontinued to snag one before that option went away forever. The M-series Mac Pro only ever supported 192GB.

San Francisco is one of the most brutally hard places to run a business, as evidenced by how competitive the landscape is.

What would have been actually interesting about this publicity stunt is if it demonstrated if/how AI could have dealt with some of the SF specific, non-sexy parts of running a business. Filing the relevant permits, co-ordinating inspections, negotiating with landlords, interfacing with locals at planning meetings.

Those are things SF business owners report as empirically unpleasant parts of running a business and a sufficient financial drag that they meaningfully affect business success. But my feeling is they had humans clear the way of all these thorny issues ahead of time so the AI could focus on the "sexy stuff".

> Even basic email has more lock-in than any of the model provider.

History has proven the average person has very little ability to discern which products have lock-in.

Everyone was confidently predicting Uber would dominate over all the regional ride sharing apps because ride sharing is a commodity and subsidies were enough to shift user behavior.

The thesis from AI providers about lockin have always been coherent: increased personalization and learning of workflows over time would increasingly make using a new AI much worse than an AI that already knows you. If you look at the human virtual assistant world, stickiness is incredibly high once are happy with your onboarded assistant because onboarding a new person unavoidably sucks.

Is this thesis correct? We don’t know, it took Uber billions of burnt cash to discover their thesis was incorrect.

This is the classic tech person trying to sound smart in a domain outside of their expertise.

Fixed civilian infrastructure has never been safe. The way you protect it is diplomatically, by having that region not be attacked in the first place or putting incentives such that certain bits of infrastructure are considered off limits.

Secondly, people whose only experience of war comes from fictional media have a drastically distorted view of the impact of chemical explosives. The science of chemical explosives became a mature discipline pre-WWII. Our best chemical today are maybe 1.7x more dense than TNT and our standard chemical explosives are often sometimes less energy dense because we prioritize safety over explosive power.

The science of rocketry became mature in the 60s, for a given fuel quantity, we can roughly lob X mass Y distance at Z speed.

The only area we've made massive advancements is precision, relevant if you want to put a small bomb through a window to kill one guy, not so relevant if you want to hit an oil refinery that is several square km wide.

The long and the short of it is that there's a massive, misunderstood gap between temporarily disabling something and destroying it. Take the Crimea bridge for example, back in 2022, there was a massive, high profile truck bombing that completely destroyed the center part of the bridge. However, it was fully repaired within 4 months. Subsequent, there have been 3 more attacks, and repaired every time and still transporting vital war materiel [1]. Concrete and steel are heavy. Even if left 100% undefended, the amount of weaponry needed to totally destroy the bridge, such that it could not be used for 3 years+ is a substantial chunk of Ukraine/US weapon's arsenal. Same goes for every power plant, oil refinery, airplane runway, tank factory etc. There's a reason why Ukraine still has reasonably reliable power after 6 years of Russian bombardment and the allies were never able to degrade German production abilities all the way down to zero despite near saturation bombing campaigns over cities like Dresden.

Bombs just aren't that powerful and our ability to produce them definitively peaked at the tail end of WWII. Currently, globally we're about at 1/20th of of that ability to produce that quantity of explosives and we likely never will reach that amount again.

The only way to actually truly destroy civilian infrastructure energetically is tactical nukes but that's an entirely other ball of wax.

In short, any time you see an expert in one field confidently expound in another field, you should be wary because, while they might be high IQ, their priors could be wrong in drastic ways that make any analysis foolish. The entire essay is arrant nonsense and would be laughed at by anyone with any degree of military analysis.

(Disclaimer: I did use AI to help me generate two numbers: the energetic ratio of WWII vs modern day high explosives and estimates of global military explosive production over time. All writing was my own).

[1] https://en.wikipedia.org/wiki/Crimean_Bridge#Attacks_after_t...

I thought Signal didn’t show message previews by default and you had to go in and enable it? I’ve never had message previews in my Signal and I don’t remember changing anything. Maybe when they introduced the feature, you could pick but they strongly suggested it not showing?

Did you know Back had an interest in Japan? A rare and unusual trait within the cypherpunk community! /s

You can mark a video as not to influence your recommendations or delete it from your YouTube history.

This is $19 Polishing Cloth slander!

> for example the V2 was technically a hypersonic missile.

The V2 was not a hypersonic missile, it was a ballistic missile that had a predetermined flight path that was easy to predict. The distinguishing factor of hypersonic missiles is that they do not require a ballistic trajectory to achieve their speeds (and are hence, much harder to detect) and they maintain maneuverability throughout their entire flight path.

https://www.sipri.org/commentary/topical-backgrounder/2022/m...

> The ancestor of ballistic missiles, Germany’s [...] V-2 was first launched in the 1940s. During ascent, it could reach a speed greater than Mach 5 and could do so again momentarily on its way back down. But, no one would claim that the V-2 was a hypersonic missile. In a similar vein, should one apply this label to modern intercontinental ballistic missiles that reach speeds beyond Mach 20 at ascent and re-entry?

> Certainly not, and there are other characteristics commonly cited when defining ‘hypersonic missiles’. However, while a combination of defining characteristics is increasingly adopted among experts, hypersonic missiles are often not well understood within public discussions in politics and the media [...]. The US-based Missile Defense Advocacy Alliance states that ‘hypersonic weapons refer to weapons that travel faster than Mach 5 (~3800mph) and have the capability to maneuver during the entire flight.’

Almost nobody uses websites in China. You're expected to interact with them via WeChat Mini-app where the experience is generally fine.

That hasn't stopped them from periodic stalinist redesigns in an attempt to purge other ports from their machines. But the audio jack always survives.

tbf, that was because a lot of people abused it by being permanently outside of the US and relying exclusively on the roaming for all their data. I know because I was one of those people for 6 years.

What's interesting is that Apple was the one responsible for removing the jack from their phones but they've stubbornly kept them on all their computers.

The only port in common between a 2015 Intel Macbook Pro and a 2026 Macbook Neo is the 3.5mm headphone jack. But also, the only port in common between a 2015 Intel Macbook Pro and a 1991 Powerbook 100 is the 3.5mm headphone jack.

Typically, Kiwis would leave after graduating university or in their mid-20s. That Kiwis 30 - 50 are leaving now is a relatively recent phenomena (18 -> 43K in 4 years).