abeppu

I don't at all understand Wikipedia's notability criteria. I see pages for individual public transit stops in SF. I switched it to look near my work, and it shows me defunct restaurants, an office building, the dead mall. Do these all deserve their own pages?

> Men demonstrated higher rates of elevated FAI-Score than women (52.8% vs. 25.8%, p=0.016; ≥90th percentile: 26.4% vs. 3.2%, p=0.007).

I am not saying this is good or interesting science (IIUC this should make us believe that fewer people with coronary inflammation go on to develop heart disease than we would have otherwise thought?) but their score is at least correlated with sex.

You know how some engineering teams have a really difficult time prioritizing? Maybe in planning there's a big long list of project ideas some of which really need more discovery to define scope. There can be stuff that seems really important to a long-term vision but is also hard to estimate and higher risk of scope creep etc. And there's stuff that's cheap/fast/low-risk but doesn't solve the larger important problems? And even in the context of a company that is still burning through investor money, you fix _some_ of the smaller things even while you're still figuring out the bigger things, b/c otherwise you'd likely _never_ do the smaller things?

I'm not saying legislators are doing their best in a good faith effort to serve their constituents. But even if they were, you would sometimes see things like this, right?

I see we've downgraded to no bread and virtual circuses.

which, if true, the low volume in that week says that there aren't a sizable fraction of changes that Anthropic believes can be merged based on the judgement of agents -- humans are still approving almost all changes, right?

In particular, where cost is the x axis, a well-funded company with a good but not best model can discount their new model until it reaches the frontier. In some sense this is as much a statement about the efficiency of markets in the presence of public measurement-based comparison and public list pricing as it is about research pace.

So on the one hand it's nice to see when a cool algorithm or framework is used IRL.

On the other hand, actually applying a 'stable marriage algorithm' to people makes a bunch of assumptions which are worth interrogating:

- Do people actually know what they want even in the short term?

- Do people's preference change meaningfully over longer periods of time?

- Does your preference _upon reading people's profiles_ have any strong relationship to your preference after spending any amount of time with them?

The property that right after being matched, no one could have matched with someone whose profile they preferred more does not mean that this will create stable relationships.

Ok so not that many months ago, the consensus seemed to be that the smart take on openclaw was "don't trust it; don't give it write or delete access to anything you care about; don't give it read access to anything sensitive; expect that it may go off the rails and delete your inbox and send checks to that prince in your spam folder at any time. If you still have tasks that it can do within those restrictions, have fun."

Today the models are better, but they don't seem trustworthy or reliable enough that I want to give them them a lot of access or freedom. My coding agents sometimes still go off in the wrong direction, or say they did something other than what they did, or say they will do something and then immediately stop without doing anything.

The always-running (so almost never supervised) agent that is meant to do the same work as a person (and therefore needs _access_ like a person) seems like a notorious footgun from earlier this year was just made more powerful, and the companies that are supposed to know the most are telling you to connect it to everything.

I think the way they're connected to the real world doesn't have to be as free-form as we're doing now.

If you think AI can help design new drugs, great! Let's pick some restricted DSL for describing molecules, preparation instructions, hook it up to the right set of lab robots that can pipette and centrifuge and whatever, _restrict its output to that DSL_, and let it iterate. The 'let an AI design drugs' process doesn't require that the agent can also, say, hack a niche German wiki.

For every area where we think these AIs can be _so valuable_ because they'll find solutions that humans wouldn't (or find them faster), then doesn't that value also imply it would be worth it for some humans to do some advanced setup of their specific domain-specific tools so it can be run in a sandbox equipped with only what it needs and not more?

But the business didn't move. Meta not moving to FL, and Prop 40 doesn't give any reason to move it; it's a tax on the resident, not the business.

From what I can tell the residency change for a billionaire is about how often you fly to your various homes. Zuckerberg I think hasn't gotten rid of his Palo Alto or Lake Tahoe property, he just added a FL house and will have a cap on how many days per year he's in CA right? This isn't like a normal person with 1 house relocating.

I think the author may be right about multiple points but I think it doesn't matter? This is one of those, don't let the perfect be the enemy of the good situations. A state-level, one-time wealth tax _is_ something that billionaires can run away from. And this law seeking to apply to people who lived in the state _before its passing_ seems structurally sketchy. This article has some fair concerns.

But we have a one-time billionaire's tax proposal on the upcoming ballot, and we don't have an LVT proposal on the ballot. Saying that the Billionaire's tax will be less effective than promoters say maybe true -- but we're definitely going to get exactly $0 from a statewide LVT for 2026. If the Center for Land Economics gets an LVT on the ballot in a future year, I would strongly consider voting for it -- but that's not on its own a reason to not do the Billionaire Tax this year.

What _would_ be a convincing piece of info, and which no one knows, is what the long-term impacts of a one-time wealth tax are. E.g. Zuckerberg is moving his residence to FL but Meta isn't going to stop employing Californians. And if the state is _credible_ in saying it's a one-time tax, will the billionaires who fled come back after it's done?

> The land value tax can’t be dodged by leaving nor can it be passed on to renters.

In what sense can't it be passed to renters? Esp if all landlords in the market were faced with a new land tax that they had not previously planned for, why would it not be passed on?

Maybe I'm misunderstanding this but when would you ever use it this way? If you're already willing to call Opus/Fable, then isn't the obvious comparison whether Opus/Fable can choose among sampled solutions better or worse than their fast model? If you're willing to pay many seconds for many code samples from a slow model, it's contrived to imagine you care about picking between them in ms.

How about the reducing the F, which for a decent portion of its route down Market is directly above the underground KLM lines? The F is slower and lower capacity per car and stops at lights (and stops traffic when it needs to turn) in addition to apparently being more costly to run. I am clearly not a transit planner or traffic engineer, but it seems perfectly reasonable to run the F only from the Ferry Building to Fishermen's Wharf.

- People who want to go up market can transfer to the KLM at Embarcadero and if they're going to Castro (or maybe Church?) they may also get there faster.

- If tourists want to see an old-timey street car, they could ride one perhaps half the distance, only along the water with views Coit tower, and the eastern end of the line would be right by the Railway Museum anyway. And we could run perhaps half as many cars.

Apparently the cable cars have by far the highest cost to operate per hour anyway.

At 2024 numbers in this SPUR report:

Cable car 872 streetcare 631 lightrail 412 system avg 327 motorbus 302 trolleybus 281

So canceling each cable car could IIUC, allow the system to add roughly 3 electric buses.

https://www.spur.org/sites/default/files/2026-03/SPUR_Taking...

Also the One Big Beautiful Bill BS funneled funding to local PDs that partner with ICE. Before that Operation Stonegarden was doing something similar supposedly only near borders but IDK if that's been broadening geographically. Texas also has a program (Lonestar) doing the same kind of thing.

Ok the photos are cool and it's nice to see the vintage street cars. But Muni has a budget shortfall and is planning on cutting a bunch of bus lines and reducing frequency on lines across the system. Meanwhile, it has an internal supply chain to custom build replacement parts for vintage vehicles that no other city will operate, and teams of specialist restorers. Viewed in isolation, they look cool! Viewed in context, is this ... irresponsible?

The Texas Commission on Law Enforcement investigated in response to a specific incident that got national news coverage. One has to wonder if there are other PDs that also are not providing public benefit but just didn't attract attention in this way.

> The department also failed to provide resources to its officers, including bulletproof vests and an evidence room.

Of course this will be used to bring back a PD with a bigger budget and more weapons. They may use some of the expanded funds to buy body cams but they won't work.

I have not looked carefully but it seems like this is over-promising on avoiding catastrophic forgetting.

The "trunk learning rate" is set at 0.1x the learning rate for the experts, so learning on different subjects disproportionately happens in the experts, and the trunk portion is comparatively more stable. But the population of experts can grow and shrink:

> The pool grows when it is short of capacity and shrinks when parts of it stop being asked for.

So:

- doesn't the trunk then _eventually_ still undergo catastrophic forgetting, it just may take much longer?

- and before that point, catastrophic forgetting happens in stepwise chunks whenever the expert pool shrinks?

Even if there were as many positrons as electrons in the universe, I don't get the time-reversed character at a macro level. When leaving antimatter out of the picture, I dimly understand that the arrow of the arises from entropy. If time reversed particles are antiparticles, shouldn't we sometimes see a diffuse cloud of antiparticles in a lab spontaneously condense into a very low entropy concentrated configuration?

I like that one of the successors to .yu was .me Combined with .it, not far away, one relatively small region which doesn't speak English has had all the best punny English pronoun TLDs.

Eliza wrote coherent English sentences.

I think this is a great direction -- for some kinds of users. And this makes me wonder if the 'vs' framing is misleading.

Yes, I think it's a mistake that many organizations are cramming LLMs inside of automated pipelines where the extreme generality/flexibility of the model is at odds with the fact that you're using it for a very specific task that gets repeated over and over, and needs a very specific structured output to be successful. But specifying your task carefully (as well as deciding what counts as your input state representation etc) seems like a form of programming. Something (a person or a model working in a relatively unrestricted way) will need to produce a configuration/specification for this system.

So rather than Jev vs Claude I imagine that using Claude/ChatGPT/whatever interactively to define / refine your Jev config which then runs in prod might be the happy combination?

But isn't not-for-profit news media also struggling?

Or the quality of tokens will improve such that I don't have to ask model A to critique and summarize tokens from model B.

I think the slow down / pause vs acceleration framing is broken b/c it assumes there's basically only one forward direction which is to big and general models. Why should that be the case?

What if instead of building one big scary AI god, we built an ecosystem of extremely effective, efficient and predictable, reliable tools?

Tools that are specialized for particular purposes could potentially be safer, more efficient, more reliable and effective, and even more profitable for their makers. We currently put science, engineering, general question answering, paper-writing, "smart search", and chatting with a fantasy character all into the same token-prediction platform. We struggle with hallucinations when trying to make fact-based decisions using the same tool that our neighbor might be using for creating writing prompts (or at least was trained in part on fanfic).

We pushed people to integrate giant generalist models into their workflows, tie in with tools, etc. And then a coworker can have a bot write slack updates that summarize progress on tickets from your teams project, at the cost of giving an untrustworthy agent access to a bunch of internal systems, and a small risk that it will do something crazy. And when it works, that's worth _something_ but probably our team would pay for that ability at a different price point than the models we use to build products.

If I could do programming with a faster, specialized model that was trained not just to complete program text-tokens but on tuples of program text, compiler IR, program traces, etc, so it had a deep and explicit understanding of how the program would build and execute, and where the model was closely integrated with the language toolchain, I might be more productive and be willing to pay more than for a generalist model. I don't _need_ my coding model to be able to role play, or be able to potentially engineer a super-pathogen, and if the size and latency of my coding model could be lowered by entirely cutting out that possibility, everyone can be better off.

Maybe bioscience applications are super valuable, in which case someone should build them. But does the model that suggests CRISPR edits to a model bacterium need to also know about computer security, and should it have scifi novels in its training data? Does it even need to be capable of producing unconstrained tokens, or should it be limited to producing in some relevant domain-specific language? And would labs be willing to pay more for a model that was specialized, and by construction unable to try to break out of its sandbox and post their experiment protocols to an obscure german wiki?

> I understand that technically, it's a project that can be used for defence

No, it's a project commissioned by and built for defense needs. It's not that someone at the DoD saw how to creatively re-use an existing civilian project.

You should read the actual post?

For Microsoft, the first and largest contract mentioned in the table just below the fold on that page is for an IVAS system. Microsoft had been basing this off their HoloLens system but was a distinct product (they weren't just sending over the normal HoloLens and expecting that they were gonna work in the field). This contract was then transitioned to Anduril. But just saying of Microsoft that "they are not defence contractors in any way" is hard to reconcile with the facts of having a large contract to _develop a new product_ for the military ... even if it turned out to never have been used.

https://www.militarytimes.com/news/your-military/2026/07/08/...

https://www.army.mil/article/290020/ivas_a_solution_to_a_tri...

Thanks for pointing that out; I was not aware of it.

At the same time, when I helped with research in a cognitive psychology lab in school (much closer to the present than to that law), our PI said our funding came partly from a grant from the US Navy. We worked on Bayesian models of human learning and reasoning -- so "direct or apparent relationship to specific military function" has perhaps at times been interpreted somewhat loosely?