The challenge on these types of things is making sure you have somebody modeling the impact then being accountable to the numbers they produce. As a massive Deming guy, any model that has zero variation built into it (whiskers) becomes immediately suspect as a bit of wishful thinking.
Emmanuel Saez, et al, out of Berkeley modeled we would see 10% leave. He is one of the world's most respected economists. Here's a quote from the paper used to model this:
"Won’t billionaires leave California to avoid the tax?
Billionaires who are California residents, as defined by law, as of January 1, 2026 will have to pay the one-time tax in full; leaving after January 1, 2026 will not allow billionaires to avoid the tax. Note that the law for determining residence is not a simple binary under California law; it is a standard that looks to both subjective and objective factors. Therefore, the taxpayer’s residency is already largely set and moving does not allow billionaires to avoid the tax."
So, the premise is essentially "we'll catch them by surprise, so they won't leave." The whole model is built around this element of surprise. From a modeling standpoint, that feels risky. And more than that: zero variance. Exactly 10% is the impact.
We already have data suggesting reality is a bit messier. If you look at just recent high-profile departures, six billionaires publicly left the state: Larry Page ($269B), Sergey Brin ($245B), Peter Thiel ($28B), Don Hankey ($8.2B), Steven Spielberg ($7.1B), and David Sacks ($2B). Plus, the paper is based off of Forbes, and there is a lot of older, quieter money that Forbes isn't going to capture.
The problem is once somebody leaves, they take away some amount of their "normal" ongoing tax revenue. Then, because the precedent is set, you risk making anyone sensitive to this,like budding billionaires, htink twice about staying. I still think the state gets a nice one-time payment, but it might not be as much as projected, and it could create a hole in the broader revenue system. If you run a Time Value series, the ROI is marginal or may even turn out to be negative.
The paper then seems to assume that after this surprise tax, the next crop of billionaires will just shrug it off and continue to be created in California at the same rate. This is a bold statement, unbacked by data, with no potential whiskers on the bar. As an engineer, no process has zero variation, but that's what their paper implies. By the way, I think European data clearly shows that without surprise, it obviously cannot work. And so surprise is great, but it has a tendency only to work one time.
The final twist is we'll likely see a massive ad campaign by these individuals to market against the bill. The biggest issue is now if the proposal fails at the ballot, the state might have damaged trust in its revenue system, didn't get the one-time boost, and an already big problem potentially gets worse.
I'm not saying that you can model everything. However, I think some modelling up front, with an error bar, would really be a good place to start. If the 10% leaving is wrong, we need to reset the model. The latest version of the paper, updated in Feb 25, doesn't do this.
By the way, I've been using a Whisper model, specifically WhisperX, to do all my work, and for whatever reason I just simply was not familiar with the Handy app. I've now downloaded and used it, and what a great suggestion. Thank you for putting it here, along with the direct link to the leaderboard.
I can tell that this is now definitely going to be my go-to model and app on all my clients.
The benches inside of IBM were phenomenally deep. They did not need to go to the outside world.However the research arm was divided into a bunch of buckets in each division had a tendency to be their own siloed operations. The various heads of research were really really good and there was some work to try to pull some of the stuff together but never a single driving overreaching strategy that I saw as an insider.
As an industry insider, I have never heard the one percent. I'm not saying that it's not possible, but I don't recall ever hearing of us testing for it. So what is the source of the 1%?
As somebody that actually had development experiences, both PC companies and in the hard drive industry, I posted an upper level post on this. I won't repeat it here, but an awful lot of this sounds apocryphal but some truth behind it.
I won't repeat it here, but I posted What I saw as an insider. I think that not all of the facts were quite right. However, some of the overtones definitely are.
I arrived at IBM just a little before Gerstner got there. He was an absolutely amazing individual in terms of how he spoke incredibly bluntly to everyone. At IBM, it was all about legacy internal political structures, where you would have enormous presentations held by various executives trying to explain why their business was going to be best and why you needed to invest in it. The ritual of presenting at IBM’s headquarters in Armonk was a symbol of the culture.
While I don't remember every single speech word for word, I remember one of his earliest all-hands meetings. He said, "If I have one more exec come to me with a bunch of slides that can't simply describe what's happening in their business with the key strategic attributes, I'm going to start firing people." He continued, "If you're running a massive section of the business, I should be able to come to you and we should just be able to have a good, solid conversation about the levers. You need to demonstrate to me that you know exactly what's going on, and you can't have your staffer answer my questions."
He also stated that he was completely appalled because, at every other company he had been at, they woke up every day paranoid about the competition. He told everyone we had completely lost our way. if you weren't waking up every single day thinking about the scorecard, you probably weren't right for the new environment.
If you were internal and under any scrutiny at all during these years, you would understand that Lou, without Jerry York, was only half of the story. Jerry was an amazing financial mind who could sniff through spreadsheets and smell bullshit. People would come out of executive reviews hanging on for dear life. Some of these executives, after they had been exposed to York and actually performed well, would leave IBM and succeed elsewhere. However, to be clear, he was an A-hole.
They achieved an enormous amount of cultural change, which was excellent. The problem was that while they had a great cultural leader and a great financial leader, IBM which was built to the brim with some of the most incredibly technical people you had ever seen, was completely lacking a clear technical lead. Inasmuch as these two guys created an amazing turnaround story, they never stopped for a moment to ask how to find the technical leadership to truly transform the place. While they did a great job, you cannot support a stool with only two legs. If IBM had found that third leg, we would still be counting it among the Mag 8 of today.
It's almost comical isn't it, but it actually turns out that this is a big foundation behind behavioral economics. In essence you can get trapped in an upper level heuristic and never stop for a moment and thinks things through.
Another one of my favorite examples is that there is some research out of Harvard that basically suggested that if people would take and spend 15 minutes a day reviewing what they had done and what was important, they increased their productivity 22%. Now you would think that this is so obvious and so dramatic you would have variety of Fortune 500 companies saying "oh my goodness we want all of our workers to be 22% more productive" and so they would simply send out a memo or an email or some sort of process to force people to do some reflecting.
I would also suggest that Microsoft had a unique advantage based out of the idea that people should have their own enclosed workspace to do coding. This was deeply entrenched when Bill was running the company day-to-day. And I'm sure as somebody that was a coding phenomenon, it simply made sense to him. But academically, it also makes sense.
Microsoft has reversed this policy, but as far as I can tell, it doesn't have anything to do with the research. It has to do with statements about working together efficiently. or AI productivity. If there's real research then it's great.
My problem is it just doesn't appear to be any real research behind it. Yet I'm sure many managers at Microsoft thinks that it's very efficient. Of course, if you do know anybody at Microsoft that codes, they have their own opinion, and rather than me repeating hearsay, it would be fantastic to have somebody anonymously post what's really going on here. I'll betcha a nickel that 90% of them are not reporting that they feel a lot more effective.
But thank you for actually some very insightful comments and actually digging into the details. And I do agree with your contention that there is some sort of circular system issue going on here (ala Jay Forrester out of MIT).
It is pretty interesting. While you reported everything perfectly, I'll just paste in the detailed section at the bottom as it does add a little more detail and really does give us something to think about. FDR in 1944 suggested that there should be a second bill of rights. In many ways I am attracted to his framework. In his second bill of rights, the very first one was "The right to a useful and remunerative job in the industries or shops or farms or mines of the nation."
It strikes me that having gainful employment in which you feel like you are contributing in some method to a society is incredibly foundational to good mental health. I think FDR recognized this and I don't think he was thinking about communism. I think he was indicating that we need to find worth for individuals. Of course, with World War II and his health issues, the somehow seemed to go by the side.
This is not somebody telling somebody on the street to get a job. It's a question of how do we enable people to get a job? And I believe if there is an opportunity for the government to spend tax dollars, it may be in incentivizing employers to take these individuals and be creative in how they employ them for direct benefits. It's hard for me to imagine that there isn't some economic way of incentivizing business to show entrepreneurship if we incentivize them correctly.
This doesn't mean that you don't figure out how to solve housing. It simply means that we think about things systemically.
"Participants noted substantial disconnection from labor markets, but many were looking for work.
Some of the disconnection may have been related to the lack of job opportunities during the pandemic, although participants did report that their age, disability, lack of transportation, and lack of housing interfered with their ability to work. Only 18% reported income from jobs (8% reported any income from formal employment and 11% from informal employment). Seventy percent reported at least a two-year gap since working 20 hours or more weekly. Of all participants, 44% were looking for employment; among those younger than 62 and without a disability, 55% were."
Okay I think I understand what happened. A couple posts ago you listed to an executive summary for CASPEH. I don't believe you've ever read the complete report, which is around 96 pages.
If you dig into the details, you'll actually find out that all of your assumptions are spoken about in terms of coming out with a reasonable amount of hours worth inside a California based upon the survey data from this research. The detailed report includes the following:
Median monthly household income in the six months before homelessness: $960 (all participants), $950 for non‑leaseholders, $1,400 for leaseholders. State the obvious if the weighted average is 960 and you have two groups, you can run the math to show that the non-lease holders were 98% of the sample.
Why we do want to think about Least Holders in reality is the renters where 98% of the problems exist. This is a clear application of the Pareto Principle, and so we should look at renters as the core of the homeless issue.
Median monthly housing cost: $200 for non‑leaseholders (0 for many), $700 for leaseholders. Of non-leaseholders, 43% were not paying any rent; among those who reported paying anything, the median monthly rent was $450.
In essence, if you look at the details you'll see where you're assuming are a lot of assumptions are actually somewhat addressed by the detailed report. Unfortunately, I'm going to suggest the detailed report is pretty shabby in terms of forcing somebody to dig out a lot of information which they should offer in some sort of a downloadable table for analysis.
Computationally, we can therefore figure out the minimal amount of hours these people must have been working based on the fact that they must have made at least minimum wage in the state of California.
There's not a lot of assumptions in this. It's based upon the detailed survey data and utilizing California minimal wage, which is where the survey was taken. The issue is digging into the details and computationally extracting information and assumptions that is not blinded by our own biases walking into something.
Again, there is excellent work out of University of Washington to suggest that higher housing costs lends itself toward greater rates of homelessness. That's not under debate here. The issue is from the survey data, it's very reasonable to do some basic computation to put some parameters around the data. It's not assumption, it's critical thinking.
Do me a favor. Tell me why do you think it's a stretch (to assume that this is a job's issue). This would appear to me to be an intuitive statement and possibly is simply created because you've already made up your mind. Unfortunately, after we make up our mind to do something, our brains are heavily subject to confirmation bias, which means it's incredibly difficult for people to take in new information or to consider new viewpoints. On the other hand, if you have good rational, logical rationale, then it should be able to be laid out fairly crisply.
However, I think it's intuitively obvious that there is a social contract that people should be expected to work a 40-hour work week. And when we find they can't work a 40-hour work week, and then they are homeless, this would appear to me to be a problem. Feel free to tell me why you would think this would not be a problem.
In your reply to me, your way of dealing with the job issue is to simply take what you initially thought and provide yet one more graph. However, this meaningfully doesn't add anything to the conversation because I already stated that it is clear that there is a correlation between housing and homeless.
As I stated, I'm familiar with Gregg Colburn, who has a methodology which goes well beyond simply doing a Fred graph. In his methodology he basically takes a look at different Geos and the different lodging cost in those geos and then he wraps it back into homelessness. There is no doubt when housing becomes more expensive, people find themselves out on the street.
There is very good research to indicate that when housing costs a lot, versus geos where housing costs a little, homelessness clearly is lower. while this is not causation, the correlation is extremely clear. I think that Gregg Colburn, The University of Washington has done a good job arguing for this correlation and it's difficult to argue against it. What's nice about his research is it's not reliant on self-reported surveys to dig out these trends.
So, if somebody is inside of the house, we definitely want to try to keep them inside of the house. I also agree with your contention that when somebody hits the streets, they actually turn the drugs. And I believe the evidence points toward the ideas of this being a system That doesn't have a reverse gear on the car. If you keep somebody in the house, they won't go homeless. But if you give homeless a house or lodging, it doesn't return them back to the original function.
But one of the really interesting facts to me, which is in the study that you linked, but also in the other studies that I've red covering the same type of survey data, is almost never highlighted.
When you actually dig into the survey data, what you find out is that there is a radical problem with under employment. So let's do that math on the median monthly household income. I do understand it is a medium number, but it will give us a starting point to think about at least 50% of the individuals that are homeless.
Your study reports a median monthly household income of 960 dollars in the six months before homelessness. If that entire amount came from a single worker earning around the California statewide minimum wage at that time (about 14–15 dollars per hour in 2021–2022, ignoring higher local ordinances), that would correspond to roughly:
- 960 dollars ÷ 14 dollars/hour ≈ 69 hours per month, or about 16 hours per week.
- 960 dollars ÷ 15 dollars/hour ≈ 64 hours per month, or about 15 hours per week.
For leaseholders at 1,400 dollars per month, the same rough calculation gives:
- 1,400 dollars ÷ 14 dollars/hour ≈ 100 hours per month ≈ 23 hours per week.
- 1,400 dollars ÷ 15 dollars/hour ≈ 93 hours per month ≈ 21–22 hours per week.
We need to solve the job issue. If thoughtful analysis is done on this, it may actually turn out to be that the lack of lodging is a secondary issue, It may be the root issue is the inability for a sub-segment of our population to a stable 40 hour a week job that is the real Core problem.
Peter Drucker popularized the phrase "Efficiency is doing things right; effectiveness is doing the right things."
Being a credibly efficient at doing the wrong things, turns out to be a massive issue inside of most companies. What's interesting is I do think that AI gives opportunity to be massively more effective because if you have the right LLM, that's trained right, you can explore a variety of scenarios much faster than what you can do by yourself. However, we hear very little about this as a central thrust of how to utilize AI into the work space.
The challenge is that we have a rapidly evolving GLP/GIP/Other landscape being developed. In other words, you take a risk that the government buys the wrong thing. However, I think with a little push, you could have a highly competitive field to lower the federal cost, and the ROI should be easy to plot.
Actually, you don't need to do everybody all at once. Target the biggest (no pun intended) opportunities first.
At the time of the FAA making up this framework, I would venture more than half of commercial pilots had some military aviation background. It is so close to Boyd's model, I would think that a research might find that it was highly inspired if not a direct descendant of Boyd's work.
However, Body was real time combat, and I think the FAA is supposed to be beyond a cockpit crisis, and maybe another framework is Demings PDCA framework, which looks like you could roughly match the pieces.
Throw in confirmation bias https://en.wikipedia.org/wiki/Confirmation_bias and you have a lot of inertia from changing. Not only do they not have the right info, but because they have invested in the ongoing solution, it is difficult to get any change going because humans tend to simply see everything as supporting their current viewpoint.
I forget all the details of my tweaks, but I remember that I had better throughput on my version.
I know the OP talked about wanting it local, but thomasmol/whisper-diarization on replicate is fast and cheap. Here's a hacked front end to parse teh JSON: https://github.com/Sanborn-Young/MP3_2transcript
While I love the Hacker News purity, takes me back to Usenet, it makes me wonder if a little AI could take a repost and auto insert previous postings to allow people to see previous discussions.