w10-1

Willa Cather?

Homer, or the Homeric tradition, arose as rootless people displaced by drought and war in ~1,000 BC needed to sell the idea that entertaining stranger guests who had only stories from afar was worth doing, and that they shared common gods (versus the city-specific gods of earlier). This mind-set supported looser city-state alliances and cultures open to trading and ideas, who then outgrew the more xenophobic cities (like the Etruscans). So they semi-accidentally created a legitimizing belief system and thereby stumbled into founder roles.

America as the land of opportunity has flourished by acquiring more land and capturing value better, mainly by diminishing obligations to country and god and fellow humans. Yes, many writers bore witness to the frictions this involved with varying degrees of acuity or tragi-comedy, but no American writer wrote in a way that actually established this American mindset.

I think instead the mindset is perpetuated by each era building achievement heroes, stimulating belief in unbounded personal possibility. American literature is reduced to pointing out the vacuous consequences. As William Carlos-Williams wrote looking into his mirror in middle age, "I am lonely, lonely. I was born to be lonely, I am best so! [...] Who shall say I am not the happy genius of my household?"

If I had a vote, I'd nominate Willa Cather as the writer who portrayed that knowing moment of the powerless, non-achieving people enthralled with the beauty of nature as it was being destroyed. It's not what guides Americans prospectively, but it might best reflect history's perspective and our own retrospectives.

(I should add that I'm glad America has hoisted individuals and done so much to lift the global standards of living and care, even if it was more by greed and necessity than charity.)

Apple could take the ChatGPT app down

No. If Apple did that without a basis in their own App Store policies, they'd spawn antitrust cases. OpenAI damages are a droplet compared to the App Store revenues.

It's insane to cross Apple

... on the question of employee confidentiality. Because the IP landscape is so muddled, they try hard to keep employees happy (most don't leave), and they're very, very good at writing, securing, and enforcing confidentiality agreements.

motivated reasoning: study in advance of nursing union contract negotiations

The sad thing is that there are many issues, but the study didn't really address them. One would hope for more effective critics.

The AI application I'd like to see would be to identify the elective hyper-utilizers (so we could try to demotivate them by addressing what's really bothering them). That would improve quality of care for them and quality of life for providers, and create time for outreach to the under-utilizers who need preventative care.

Sounds like Correnti didn't negotiate well

Are you sure this is negotiable?

Note that the objection is as much to the abruptness as the scale of the price increase.

Large customers sinking millions into (here aviation) assets rightly expect that their vendors respect their market lifecycles.

Starlink's behavior creates a credibility gap that could drive such customers to the 2-3 upcoming competitors with poorer service and even higher costs, but perhaps more reliability as partners.

I'd see this as an opportunity.

pandoc purportedly can run as wasm: https://github.com/haskell-wasm/pandoc-wasm

Loved it. I would change the timer from text (which I'm constantly reading as it changes) to a progress bar or set of dots which don't require thinking. Also I would use 3 strikes instead of one you're out, because a single miss isn't really representative. (similarly, any valid word should work.) Letter display order can offer clues but seems random. Perhaps vowels on top row, then consonants on bottom, in alphabetic order? Would love a hard mode for multi-lingual (latin-text) speakers to use multiple languages.

If decisions are actually being made based on analogies instead of analysis, the whole thing is brittle.

The entire military is predicated on physical possession and distance; that's the main reason Ukraine is a quagmire, Iran was a no-go, and Taiwan has been relatively safe.

But the next real war for the US is not for territory but to destroy its economic and financial leverage, and destroy its ability to produce those -- by cutting academic research, firing military and intelligence leadership, alienating allies, creating divisions that paralyze democracy, crashing the market, overloading government debt, and binding the Fed, so any response would be muddled and capital and people find opportunity elsewhere. Hence the political enlistment of the poor and unemployed on one hand, and single-minded capitalists on the other, to the same ends.

The trillions of dollars in AI spending and on the military do nothing to address this, and indeed make it worse by exhausting resources for real solutions.

also, asbestos is too small to interact with T-cells, so it must be safe.

"Fair use" was always fuzzy. To be honest, I care a lot less about slurping up the public internet and private books to make models than about every profession on the planet being forced by their employers to create skills that automate their knowledge work. The latter is much more directly an expropriation, legitimized only by the shortage of work, i.e., market power.

As prior art, see also djot (~2023), from the pandoc folks: carefully crafted as more well-formed than markdown. Notwithstanding the provenance and engineering, it had minimal uptake.

- https://djot.net

- https://github.com/jgm/djot

In wikipedia-land, I read "primary source" as "motivated source", given their need to prune biased edits.

Paper title: Disinhibitory signaling enables flexible coding of top-down information in cortical networks

(should be qualified as in-silico visual systems)

Method: replicate fMRI findings of visual abstraction using simple networks to model what's essential

Gist: in tasks 'Inhibitory neurons that suppress other inhibitory neurons seem to pass key information from the “thinking” part of the system to the “sensing” component of the system'

I've heard the same for motor control: it's not that the cortex aims for one action; it aims for a bunch, but most are inhibited. (You see this in chaotic movement when inhibition fails).

So it's not really "think and see" but "what you see when you're doing a task".

(There's some analogy in there wrt (AI) exuberance effacing selectivity in investment decisions...)

Hey investors: companies will pay big to save salaries!

Hey people: not to worry, new jobs will come!

1 + 1 = 3 because growth!

This has a lot of potential; true disruption can happen for existing markets only when the transaction cost features change, and CloudFlare is ideally positioned to drive a new standard. Ideally they create a service that can be open and replicated in competition (not just technically but economically), and this creates the right incentives for bots and sites/services.

I dislike stablecoins because they legitimize their cousin coins and because (I think?) they have transaction fees that create the wrong incentives for providers. I'm not sure what the real benefit is over prepaid (policy-driven) fiat currency with (possibly-paid) transaction records.

I can see how selling to bots could become so profitable that no one bothers to present directly to humans, but I look forward to an ad-free, much more capable internet, where paywalls are more like a headwind than a wall.

I like this distinction, but it reminds me to cherish the people who don't require such confidence games.

Feynman has a famous anecdote about sitting around the table with senior scientists in contentious argument where he was perplexed because it was obvious to him who was right. They argued all sides, and ultimately agreed, having proofed the idea and its alternatives.

That's who I want on my team: people who can shake things out without needing to be right or needing others to be humble, and without playing games. After viability, that's my primary criterion for a position.

Apple open-sourced their container tooling and they also support Linux in their Virtualization framework. Anyone can run any compatible Linux now in a VM, with bare-metal speed but limited to virtio devices. To me it's baffling that more people don't run Linux this way.

At Javasoft there were Solaris test machines that had been up for 2+ years, and we had to reboot the windows test machines multiple times a day. It felt really good to leave a large queue of work at the end of the day on the Solaris/Sparc machines, knowing it would be done the next morning.

change units to "natural log"

Great tip, but I didn't see "natural log" specifically. Perhaps "Compounded rate of change" is most applicable? That's still mostly above 0 historically, indicating margin usage is ever-increasing. The helpful graph would be margin usage as a weighted percentage of market participation.

dumb peripheral question: memory as a circuit would seems easy to add directly on-chip for a company like Apple. Is that blocked by time-frame, IP, technology, cost?

re: dissolving into compilation, I think the machine/human separation has been at work for some time. Modern languages (e.g., rust, swift) are already pioneering tracking aspects like effects, lifetimes, regions, etc. and then using whole-program optimization at compile- and link-time, largely based on intermediate languages like LLM IR/SIL, which are surfaced as user-visible features when they compose well with other user-visible language features. LLM training on these languages makes them suitable for generative AI; I doubt LLM's could pick up some new language, particularly if it weren't analogous to existing ones.

The idea of this post - to write separate requirements for each concern and let LLM's integrate them - is much closer to the Leo version of literate programming, which allowed documents to be composed (roughly via scatter/gather operations mediated by sentinels).

But the post entirely lacks the motivation for the AspectJ/AOP join point model: to have principled time/place for concern integration that was statically determined, type-safe, understandable to users -- and suitable for integration.

I've also always hated the specific mechanism that AOP chose to implement it with – something called the "join point model" which basically amounts to runtime pattern matching on a program's call stack and running some code every time a pattern matches.

AspectJ's join point model is only dynamic where Java as a reference-based language could not support the static analysis. At compile-time, the "static shadow" of the pointcuts was calculated and implemented where staticly determinable; only the dynamic residue is deferred to runtime (e.g., is the caller to this method of type X?).

Many of AspectJ's join points and type extensions - method call or execution, exception throwing, field access - largely have been adopted in many languages (python context managers, swift getter/setters/extensions), and the residue are a bit hard to use.

But nothing really matches the power of pointcuts: to combine these predicates and the type-safe state-management - e.g., "when throwing an exception after a transaction, capture the span id along with the user id into a log message"

AOP was great for the 7% of code that it was intended for, but was largely displaced as too complicated. Now with LLM's it's a decent hypothesis that with proper training LLM's could actually handle the more complicated but ultimately cleaner programming model - cleaner because it avoids the scattering of similar code which makes it hard to change.

The key insight is that dominant concerns establish the basic structure of the application, leaving some important but residual aspects to fit themselves to the structure. That means the dominant structure must be suitable for the AOP integration (i.e., support the right pointcuts and type extensions); solve that and you've solved most integration issues. It's especially helpful for feature architectures, where you offer code in open-source to gain API adoption, paid for by closed-source library integrations with additional features.

It's not just the second-guessing. It's the getting in the ballpark but striking out: explaining in detail why they are not correct. A little bit of patient knowledge requires a tremendous amount of doctor time to explain away the ignorance.

It's a 180 for me: While I believe doctors should explain diagnosis or treatment decisions when asked, I don't believe they should be taxed with explaining away alternatives. In my anecdotal 2nd- and 3rd-hand experience, doing that is taking at least a third of their time (on roughly 5% of the patients who think demanding answers will make things better) -- with zero improvement to diagnostic accuracy or treatment effectiveness. Doctors already consult with other doctors, and it makes no sense for them to have to consult with ignorant patients or treat their AI psychosis on top of their disease. It doesn't increase patient autonomy any more than adding a steering wheel for child car seats would help toddlers learn to drive.

It's a great movie and a great scene, in some respects.

But I don't think the realness of being an orphan or being in war or being in love has much to do with the problem of AI slop, nor would I rely on some human essence to privilege human agents.

AI slop is just the aesthetic end of a deeper problem more closely related to the so-called banality of evil: how normal social and governance systems can have horrid effects notwithstanding high participation requirements. We rely on the unlikelihood of collective evil in juries, representative governance, and reputation to discipline markets, but AI and unlimited anonymous political contributions have changed that likelihood even more than the proverbial self-interest (attributed to Upton Sinclair, something like: It is difficult to get a man to see a truth when his job depends on not seeing it).

"coding" is the skill that realizes other skills: it's a delivery medium.

The skill to learn is to be able to see and address the problems that will need solving in the medium term.

Even before AI, software changes a lot (desktop, enterprise, consumer web, IoT, devices, robotics...). So for the "seeing" part, it's like planning to cross the continent not knowing what transport or maps will be available: you can mainly guess they'll come out of the terrain and available materials and expertise.

For the "address" part, it's helpful to have seen similar problems, but probably more helpful to have seen a variety of problems and developed some high-level approaches, typically in academic study. Depending on your domain, that might involve math, or engineering, or biology, or.... All coding is encoding, which is fitting some domain to data structures and algorithms.

For the "delivery" part, it's especially important that your code restrict itself to what's needed by the actual end use. That might not be cool or interesting, and it often requires getting more direct appreciation of the end-use than the often-weak problem statement. It's a trap to just enjoy the beauty of it.

Someone like Noam Shazeer demonstrates the value of good coding: lots of people in his environment had lots of good ideas, but he made it happen (and it looks like he enjoyed it).

If they could eat from higher branches than others, they could avoid competition for resources.

On the list of things that make this possible, AI comes after the anonymity of web sites and of companies (per another comment, Prompt Digital Inc (DBA Qontour) - which is who exactly?), and the fact that the infringer has complete control over their reach.

The asymmetry between stealing and getting caught or stopped was baked in long before AI, but this will become much more prevalent because the cost of infringing has been reduced by orders of magnitude.

Relatedly, legal copying seems just as problematic: I see both software and media being munged and parroted as soon as it appears, which means innovators do not get the benefit of their innovation. I personally have halted any projects where I can't completely control access to the product, which is a huge damper on innovation.

Every home, school and office should have a view of at least three trees, be in a neighbourhood with 30% tree cover, and be within 300 metres of a park

Something like this has always been a driver of my home search. I've been lucky/glad to find something, but it's always been a stretch.

But as a more realistic policy I would start with homes and with ensuring schools have a play area with plants larger than the kids and too dense to see through.

I think the secondary effects on planning are just as important: urban planning goes awry when it fails to consider livability and beauty. Livability includes walkability, non-car transportation, and parking for car-dependent transportation -- things you need to do most days. Beauty has obvious direct effects, but indirectly it also encourages people to take care of the place, and to avoid being seen e.g., as a litterer.

As policy encouraging changes, it would be good to start with environmental enhancement sections in each permit, and (if possible) to publish them before they are finalized for people to contribute proposals. This expertise is rare in developers so advice would be welcome, and providing the opportunity would probably encourage people to contribute, with ideas and perhaps funding. (There should be no provision for blocking permits, to avoid adverse NIMBY incentives.)

Some effect is real, but it's likely overstated by poor metrics.

"Currency" in all fields relates to the recency and frequency with which you dealt with a particular issue. Whether flying on auto-pilot or coding with AI, automated reduces some currency. But is that a reduction in capability?

Measuring concrete tasks makes currency the operative skill; that's why it works to cram for standardized and mid-level tests.

(Indeed, the 2010's interviewing "wisdom" about people being quick to answer simple questions veered into measuring currency, not skill.)

I think this effect is strongest in time-impacted professionals. Doctors doing dozens of endoscopies a week and developers churning out code will use what tool leverage they can, and forget as much as possible to focus on what they need to. I suspect the effect is weaker in personal or research projects.

People riding bikes won't be able to run long distances - because they won't have to, and will be able to outdo any runners. That's only a problem if the supply of bikes is someone constrained. So the risk is not skill loss, but losing control of the means of production.

Running and serving (Hetzner Arm): http: https://swiftos.tech/

Announcement: https://forums.swift.org/t/swiftos-a-from-scratch-kernel-and...

Separate address spaces, kernel TCP/IP, SMP, filesystem. Arm-64 only. Working on drivers as restartable userland processes. POSIX-inspired but not Linux ABI-compatible.

A testament to what can be done by reducing requirements to the essentials.