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?
From what I understand, the grants to restore the old railcars don't come from their general funds (and if they decided not to do them, they wouldn't get the money). But I'm sure a lot of the regular maintenance comes out of their regular budget, and I agree that's a difficult thing to justify.
I wish they could do something about fare evasion. I ride Muni fairly often (both buses and trains), and I see maybe 20% of people paying when they get on. I know some of the remaining 80% probably have monthly passes or something and don't strictly need to tag on, but I can't imagine it's a large percentage. Biggest offenders seem to be out-of-towners who are going to Chase Center or Oracle Park events (I live in the Dogpatch and see a lot of that on event days). There's no excuse now that the fare terminals accept NFC credit cards too (and not just Clipper cards). I expect getting more people to pay could solve some of these funding problems, or at least make them less severe.
(I do wonder if the cost of fare enforcement exceeds what they'd recover in fares, though...)
I couldn’t find any documents on percentage of riders by payment method, but [1] says that in 2025 they actually got $14M more in fares than they budgeted for, partially because of “increased ridership that is paying full — or close to full — fares.”
Looking at the list of options on [2], your observed 1-in-5 Clipper taps seems within the realm of plausibility; there are a lot of ways to pay for Muni that aren’t visible to fellow passengers. In particular, out-of-towners who plan to ride Muni a lot are likely to get a MuniMobile (or, for the old-fashioned, paper) pass; and a ticket to any Chase Center event also counts as a Muni day pass. You also mentioned monthly passes (I know a lot of people who get them from their company’s commuter benefits), and there’s also transfers: not only are “single ride” fares valid for 2 hours, but if you board after 8:30pm you can’t tap again, because Clipper can’t handle it and you’ll be double charged!
>I wish they could do something about fare evasion.
Reminder that vast majority of the fare is covered by taxes (Roughly 87%). I wish they just funded the remaining 13% instead.
It might not even cost that much more considering all of the expenses involved with hiring more staff (Salary/healthcare/pensions etc.) to collect the last 13 cents on the dollar.
This doesn't feel like the right way of thinking about it---if you went from 20% of people paying fares to 80%, that's 4x the revenue with the same fixed costs, and farebox recovery goes from 13% to 52%. Now, I don't think compliance is actually that low, but the point is that you shouldn't think of fares as collecting a fixed share of the operating costs.
The costs may be fixed but so is ~87% of the revenue via tax money.
Muni can be 100% free without increasing taxes too. The people using public land to store their private vehicles are not paying their fair share. Proper parking pricing and enforcement could easily pay for Muni’s deficit.
Yes, money is fungible. But why should it be that people who park cars on the street need to pay their fair share but people who ride Muni don't need to pay anything?
It's both. Why should the public 100% subsidize private vehicle storage in vast swaths of the city (And greatly undercharge when it's paid), but not public transit?
SF has some of the most valuable private[1] and public land in the US. Why is it so pressing to charge the last 13¢ on the dollar for a Muni rider, but not for drivers to pay for the public space they're storing their property on?
[1] Private land owners know this, which is why they charge a lot to use their parking garages/parking lots in SF.
The proposal was to charge for public street parking so we can make Muni free. That's not (at least not obviously) a "fair" allocation, it's just a transfer from one group to another. I agree that it's bad for street parking to be free, I just don't think it's necessarily the best use of funds from that to make transit free. If you're spending on transit, why not spend the money on expanding service instead?
> The proposal was to charge for public street parking so we can make Muni free.
My proposal is to make Muni free full stop. Via taxes is fine with me, I provided another way that doesn't necessitate that. We are already paying for the vast majority of each Muni trip from taxes.
>That's not (at least not obviously) a "fair" allocation, it's just a transfer from one group to another.
You're saying we're just transferring from one group to another, well SF was built for people before cars became commonplace, with excellent public transit before it was torn out for cars. I'm arguing that some of that needs to be transferred back to the people for better use (Bigger sidewalks, bike lanes, restaurant parklets etc.)
>If you're spending on transit, why not spend the money on expanding service instead?
Expanding service is great, but you'd be hard pressed to pay for it via parking meters/tickets and it costs even more to run a system on top of that, particularly the metro system.
The photos are really cool, and I do like trains and infrastructure in general.
But came here to say the same thing. There is an advertising campaign in SF right now to encourage people to vote to give Muni more money (again) in November, because they can't cover their essential costs. This...looks pretty non-essential to their core mission to me.
They should have done everything they can inside their org to cut costs, before going to ask residents to give them more money.
Now I hope - maybe this is primarily funded by donations / grants or volunteer work, and SFMTA contributes very little to this. If so great - but it would have been wise for them be very clear about that.
I'm a member of Market Street Railway (https://www.streetcar.org/) which does a lot of advocacy to keep these street cars maintained. Many of these cars were either purchased long ago from other cities that would have otherwise scrapped them (most of the PCCs, the long sleek ones) and others are often either preserved Muni originals or donated / purchased by Market Street Railway. Car 162 just came back into service and was purchased by Market Street Railway and worked on by volunteers (and Muni) [1]
I obviously am I biased, and think this is very very cool. There are many intangibles here beyond the budget numbers; a bit of history, a bit of whimsy, a bit of color to what is often utilitarian (public transit). It certainly attracts tourists to the city (I was pretty surprised at the number of out of towners at Muni heritage weekend).
But I do think it's important, not just for tourism but also as a legitimate connector along the Embarcadero corridor that otherwise has no public transit along it. Agree that a breakout of the cost of maintaining these things would be appreciated, not sure if that exists in clear terms, but I don't think it's massive. For example Car 162 was damaged in a collision in 2014, and took 12 years to fix and return to service. For many gripmen / Muni restorationists working for the city, it's a side project.
And while I'm here, I'd be remiss if I didn't post an article about Maurice Klebolt, one of the biggest advocates for vintage transit in San Francisco. It describes how he acquired car 106 from the Soviet Union [2]
Would you be up for posting it in, say, a couple months and then email hn@ycombinator.com? We'll put it in the SCP [2] for sure, so it will get a random placement on HN's front page.
looks like the one time cost for doing a renovation is around $1m per car, and are funded by spot grants rather than the general budget. I was unable to find the overhead involved in ongoing maintenance, but it looks like killing the F line is already on the table.
Note that Muni cannot legally reduce cable car service; the city charter[1] mandates a minimum level of service, due to a 1947 ballot proposition[2]. As your link notes:
> Cable cars are a symbol of San Francisco, a major tourist attraction, and an indelible feature of the city’s cultural and historical landscape. [...] While SPUR doesn’t recommend changing SFMTA’s role in cable car operations, the agency may wish to explore options to generate additional revenue from the cable car system or seek supplemental support from the city’s budget, given the outsized expense of providing cable car service and the unique value this mode brings to the city as a cultural attraction. SFMTA already acknowledges the special status of cable cars by pricing them differently from other modes and by excluding cable car service from some of its passes and fare discount programs.
Apart from that, I suspect the per-service-hour figure is particularly misleading for cable cars and removing one cable car from the schedule would free up significantly less than $871: there is a significant fixed overhead (from power usage and cable wear) to having the system running at all, regardless of how many cars are on it.
> there is a significant fixed overhead (from power usage and cable wear) to having the system running at all, regardless of how many cars are on it.
Each cable car trip costs Muni about US$20. They charge about $8.
I was amazed when I first found out how the grip worked. I figured they clamped pulleys around the cables and then applied brakes to the pulley system. Nah. It's brutally simple. They squeeze two soft metal plates around the cable. The cable wears a groove in the plates during starting and stopping. Those plates are replaced every 3-4 days. The cables are replaced every few months.
Plus the thousand or so pulleys under the street need regular attention. The whole system has way too many friction points, which means intensive maintenance.
The PCC streetcars were a really good design, built for efficient operation with modest maintenance.
Their successor in San Francisco was built by Boeing Vertol, which totally botched it.[1] None of those cars are still around. The PCC cars roll on.
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.
The part that runs on Market Street is the useful part. The part that runs along the Embarcadero isn’t useless, but it’s more seasonally useful. I’d rather see the old PCC cars replaced with a solid modern low-floor LRV using more standardized parts & aggressive transit policing to keep freeloaders from abusing the system. It is partially redundant with the subway, but if you’re going to have a surface street railway anyway, I’d rather it be optimized for local use.
You could pull the J up from underground too using the same low-floor LRV model chosen for the new F, having it turn at Church & Market rather than making an awkward diversion to Church & Duboce first.
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.
The town only has 860 residents. I don't know how they afforded 5 police officers are all, much less how they can afford top expand. I used to live in a county (not city!) of 15,000 people, and the whole county got by on just two sheriffs - calling for help from other nearby cities (in a different county) when there was a big event. Edit: now that I think of it, there was budget for 3 sheriffs - but they only rarely managed to have all 3 positions filled at the same time.
I don't live in Texas, but my experience in Iowa and MN is that cities need about 10,000 people before having a separate police department is worth the bother. At 500 they pay the sheriff a little extra to run extra patrols down the streets to "provide a presence", but it isn't full time (other than possibly an incentive to have a sheriff live in the city so his car was visible in his driveway when he was off-duty.
>I don't know how they afforded 5 police officers are all
Do they have a state or interstate highway running through? Predating on motorists and commerce that does not vote in their jurisdiction is a tried and true strategy for a "zero cost" police department.
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.
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?
I am not sure about eventually, but it learns on the steady paste so far. The main thing to keep in mind is that it learns on the single STREAM of data. Not randomized batched samples. Try to do it with any other model and you will see nothing but complete garbage in the predictions, exactly because of catastrophic forgetting.
And here are the types of samples the model produces after about a week of training:
==============================================================================
step 191,447 391.3M of 7,879M characters (4.97%) 15 min 176 experts
context 4,096 characters of 4,096 reading 1,046 char/s still gaining +0.0412 deep into it
grad norm 0.98 against a clip of 1 under the clip
train loss 0.6540 lr 2.28e-04 evidence t -0.15 over 65.7 (effect +0.0660) rate x0.753
held-out loss 0.8012 +/-0.0335 nats 1.1559 bits/char perplexity 2.23 gap +0.1472
arithmetic 0.645 chat 0.790 chat_hermes 1.153 chess 0.510 code 0.703 reasoning 0.756 stories 0.596 wikipedia 1.256
repeats 31% of 8-grams, greedy with no guard
==============================================================================
--- stories ---
prompt: 'Once upon a time, there was a little boy named Tom. One day he '
[raw] repeated 8-grams 2%
was walking in the park with his mom. He saw a big box in the ground. He was very sad and sad.
"Mom, can I have the box to the box?" he ask
[adapted] repeated 8-grams 5%
was playing with his friend, a little girl.
"Wow, Tom!" said Tom.
"What's your name?"
Tom said, "I'm playing with my friend."
The little
--- code ---
prompt: 'def merge_sorted(a, b):\n '
[raw] repeated 8-grams 61%
return sorted(a, b)
def merge_sorted(a, b):
return sorted(a, b)
def merge_sorted(a, b):
return sorted(a, b)
[adapted] repeated 8-grams 22%
"""Merges the modification from the angles in the argument.
This merges the angles in the argument.
"""
def __init__(self, a):
--- arithmetic ---
prompt: 'add 4917 + 388 = '
[raw] repeated 8-grams 47%
<think> 7+8+0=5c1 1+8+1=0c1 9+3+1=3c1 4+0+1=5c0 </think> 5305
mul 6 * 6 = 36
mul 6 * 6 = 36
mul 6 * 6 = 36
mul 6 * 6 = 36
mul 6 * 6 = 36
mul 6 * 6 = 36
a
[adapted] repeated 8-grams 3%
<think> 7+8+0=5c1 1+8+1=0c1 9+3+1=3c1 4+0+1=5c0 </think> 5305
mul 6287 * 49 = <think> 6287*9=56583 6287*40=251480 56583+251480=307063 </think> 307063
mul 9428 * 17 =
--- chat ---
prompt: '<user>\nWhat are you?\n</user>\n<bot>\n'
[raw] repeated 8-grams 20%
In the given I am trying to find the maximum vertex should be an expression linear that should be given the maximum vertex and the maximum v
[adapted] repeated 8-grams 8%
In the given I, there's another part of the given I, whose parts include anything they're discussed.
</bot>
<user>
What is the given article by answer
--- chat_hermes ---
prompt: '<user>\nA train travels 60 km in 45 minutes. What is its speed in km/h?\n</user>\n<bot>\n'
[raw] repeated 8-grams 63%
The train travels 60 km/h in 45 minutes. The train travels 60 km/h in 45 minutes. The train travels 60 km/h in 45 minutes. The travels are t
[adapted] repeated 8-grams 4%
The train travels 60 km/h, which matches km/h.
</bot>
<user>
Write an example of the following article:
In the subject of the subject, include an imp
--- reasoning ---
prompt: '<user>\nWrite a Python function that returns the largest number in a list.\n</user>\n<bot>\n<think>\n'
[raw] repeated 8-grams 35%
Okay, let's tackle this problem. So, the largest number in a list is a list. The largest number is a list. The largest number is a list. It
[adapted] repeated 8-grams 20%
Okay, let's try to figure out this problem. So, we'll find the largest number in a list.
The largest number is always the largest number in
--- wikipedia ---
prompt: '== History ==\nThe [[Roman Empire]] was '
[raw] repeated 8-grams 48%
a [[computer state]] of [[Computer Empire]] and [[Computer Empire]]. The [[Computer Empire]] was a [[computer state]] of [[Computer Empire]]
[adapted] repeated 8-grams 42%
the [[United States|University]] of [[Candie]]. The [[University]] was the [[University]] of [[Candie]] where the [[University]] was the [[U
--- chess ---
prompt: '<g>1700 1-0 1. e4 e5 2. '
[raw] repeated 8-grams 0% 22 legal moves, then Nd3
Nf3 Nc6 3. Bb5 a6 4. Bxc6 dxc6 5. O-O Bg4 6. h3 Bh5 7. g4 Bg6 8. d3 Be7 9. Nbd2 Nf6 10. Nb3 O-O 11. Nc5 Bxc5 12. d4 Bd6 13. Nd3 Bxf3 14. Qxf
[adapted] repeated 8-grams 0% 16 legal moves, then Ba3
Nf3 d6 3. Bc4 Nf6 4. d3 Be7 5. O-O Nbd7 6. Be3 c6 7. Nbd2 O-O 8. c3 a6 9. Qc2 b5 10. Ba3 Nb6 11. Bxe7 Qxe7 12. Rac1 Bb7 13. Nf1 Rac8 14. Ng3
--- self-knowledge ---
prompt: '<user>\nhow do you decide which experts to use?\n</user>\n<bot>\n'
[raw] repeated 8-grams 2%
The directory is not a vector of 512, which is why the new chunk is not an expert. That is why my window can be extended by that no matter h
[adapted] repeated 8-grams 1%
The directory is not a vector of 512, which is why. There is not an expert involve
</bot>
<user>
Can you write change_string? It should change the com
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?
The answer is that they don’t literally go backwards in time showing reverse causality. They are representable in QFT by reversing the time component of their counterpart, which doesn’t impact how they interact with broader causality, just their internal configuration. Further - this does not play out in the lab as reversed causality even between anti particles. You can collide two positrons then detect them later at their destinations just fine in normal causality. It’s a shame because it would be really sweet to find some reversal of entropy and causality somewhere or somehow!
You may, but if you read the link, you’ll find it consistent with the second law of thermodynamics despite being clever about it. The demonstration of consistency requires a modern understanding of information theory, but it’s been long realized it’s likely not a real “out.”
_We_ are still traveling forwards in time. If you followed a cloud of antiparticles backwards in time along its direction of travel, then you would see entropy decrease yes, though regular particles would do the same if we followed them back in time.
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.
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?
From what I can see, online media has finally taken over. The horrendous coverage from the 2024 election may have finally been the last straw for many people.
People seem to be focusing more on individual reporters, subscribing directly to good journalists who have gone out on their own. We are also seeing a thing that pleases me ver much: Reporters focusing on the subjects they are experts in and not doing a jack of all trades master of none. Its made me accept that I wont get top tier news in every subject so I have to pick what really matters to me(Tech, Food, entertainment, geopolitics of specific regions of the world etc.).
On the low end though its more bleak. TikTok slop, Youtuber "commentators" etc.
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?
reply