20,307 karma · joined April 23, 2011
I'm a programmer and founder with experience ranging from web development to embedded systems and network programming.
Working as a principal engineer for a unicorn supply chain startup.
https://github.com/sarchertech/
That can’t be true. A group’s probability cannot exceed the overall probability by more than the reciprocal of its population share. A 35× ratio therefore requires the group to represent at most 2.86% of the relevant population. For comparison, NCES reported that 9.3% of students in grades 1–5 attended private schools in 2016.
Taking the end result of a 22 year old dev process and telling Claude to rewrite it, is in no way equivalent to doing 22 years worth of the work put in by the developers working on it.
That’s like saying replicating a paper that was the result of 22 years of research is the equivalent of 22 years of the author’s labor.
No Man’s Sky isn’t the equivalent of 10,000x the devs who made it. Because the procedurally generated content isn’t equivalent to handcrafted content. I’m also not aware of any procedural generation techniques in No Man’s Sky that weren’t possible 30 years ago.
As far as Tamriel rebuilt goes, if you’re talking just about just reimplementing Daggerfall in the Morrowind engine, you might be able to do that in a month of vibecoding.
But that’s not hundreds of dev years worth of work.
> In the pre-LLM era, the economic value of "understanding" and "shared knowledge" about the programs themselves was only that it made maintenance and changes cheaper.
This only holds if there is a bright line separation between the problem, the solution, and the implementation of the solution. And that is not the case.
Understanding the implementation frequently improves your understanding of the solution, which improves your understanding of the problem, which leads to better solutions. Sometimes it leads to realizing that there is a completely different problem you should be solving.
So what I’m refuting is that there is no value to understanding the program.
But beyond that black box products that users know how to use but don’t understand is also a vast oversimplification.
Frequently software is useful specifically because it forces users to adopt a process not just because it allows them to do something they were previously doing faster. In those cases the process is usually informed by the implementation, so the user needs to understand parts of the implementation even if it’s indirectly.
They come up with the features and the implementation is an unimportant detail in realizing their designs.
In reality implementation details frequently drive design and in most cases they are at least highly intertwined.
I can’t count how many products and features I’ve built that ended working completely differently than the initial design, or how many times I was able to simplify a workflow after spending a days thinking about the implementation details.
You can kinda do something similar by spending days or weeks going back and forth with an LLM, but at least in my experience that doesn’t really save much time and I don’t understand the problem quite as well by the end of it.
But paraphrasing
> I can produce in a month what took years for teams of hundreds of people
Is completely insane.
If you can really do that Anthropic would like a word with you.
2. You can go here and look up any payments (cash or otherwise) from pharmaceutical companies to your doctor. All payments over $13.82 are reported.
https://openpaymentsdata.cms.gov/
3. Even if the dentist’s drug rep was interested in committing a felony. Fluoride toothpaste doesn’t have enough profit margin to send dentists on expenses paid international vacations. We’re definitely not talking about enough money to risk jail time.
4. Most physician groups and hospitals prohibit taking money from pharmaceutical companies or at least low limits.
5. My wife is a physician and most of our friends are physicians. No one that either of us knows has ever gone on drug company sponsored conference trip.
6. I always check my physicians and dentists on the open payments site. I’ve never seen any of them taking more than a few bucks for meals.
1. Candy Crush is one of the highest grossing games of all time, but if you look at the other top 20 highest grossing games, the vast majority are nothing like it.
2. Janky crush wouldn’t have been the success it was without massive attention to detail and polish.
I one shotted a math game for my 4 year old complete with lots of juice “screen shake, explosion animations etc…”.
It worked great, but if I tried to change even the slightest thing the result was worse than the initial output. No matter how hard I tried to nail things down with tests and specs, each change looked or felt worse.
There are about 5 states that are red enough to even conceivably split off. But those states also have such low populations that they have disproportionate power that they would lose in a split. And none of them have any kind of state identity that would make up for that.
The only states with enough state identity to even consider splitting off, Texas, isn’t red enough to do it.
Very blue states aren’t all that blue (the presidential vote margins in 2024 were much smaller than they were for the reddest state) and they are far too integrated with nearby red states.
Comparing those 2 datasets is meaningless. It’s just an exercise in how good the researchers are at guessing when trying to control for selection bias.
There were also significant changes in the test and how the test was administered (from computer or paper to only tablets).
Also if you look at the results some of them are too large to be readable. A 50% increase in people operating at or below level 1 (40% if you exclude non-native born). That would have to mean a drastic drop in 2017 such that something like 75% of kids graduating high school are doing so at or below level 1.
Also level 3 has no direct grade equivalence. If it did it would probably be around 9th grade not 6th.
It was a review or meta analysis. There was no criteria for acceptance. It’s just 2 random doctors cherry picking studies. The authors refer to it as an editorial and not a meta analysis for that reason. The language even sounds like a newspaper editorial. They make far too many unsupported assertions for an academic paper.
It’s also 20 years out of date. And we’ve had 2 decades of additional research.
>It is possible…
Of course it’s possible there’s a hidden confounder. That’s possible with literally every single observational study.
A sufficient RCT in this case is essentially impossible and likely unethical.
But you failed to read the conclusion that
> Our results indicate a 10% reduction in risk of colorectal cancer for each 10 g/day intake of total dietary fibre and cereal fibre and a about a 20% reduction for each three servings (90 g/day) of whole grain daily, and further reductions with higher intake. These findings thus have important public health implications. Our results provide further support for public health recommendations to increase the intake of dietary fibre in the prevention of colorectal cancer.
Despite the fact that it may be wrong.
Read the other 4 very large meta studies.
Read the statements from essentially every relevant heath authority.
Read the theoretical science.
Of course it’s possible that there are hidden confounders. But the preponderance of the evidence is that high fiber == less colon cancer.
Specifically:
1. We don’t know for sure what the relationship between fiber and CRC is, and we likely can’t.
2. The majority of studies that correct for lifestyle and diet still show a correlation.
3. There are plausible explanatory mechanisms.
4. The most likely explanation is that there is some protective effect.
There is broad consensus among Doctors, health agencies, and researchers that the above are true.
Here is a list of 10 widely respected health organizations who state that a high fiber diet likely reduces the risk of colon cancer
World Cancer Research Fund
American Institute for Cancer Research
American Cancer Society
Cancer Research UK
National Health Service
Bowel Cancer UK
British Dietetic Association
American Gastroenterological Association
Dana-Farber Cancer Institute
Cleveland Clinic
Here is a list of 5 very large meta reviews that show a correlation between low fiber diet and colon cancer even after adjusting for diet, lifestyle etc… All published after your editorial in top tier journals.
https://www.bmj.com/content/343/bmj.d6617
https://jamanetwork.com/journals/jamanetworkopen/fullarticle...
https://pubmed.ncbi.nlm.nih.gov/30638909/
https://pubmed.ncbi.nlm.nih.gov/31495339/
https://pubmed.ncbi.nlm.nih.gov/34854791/
>fiber in the 80s
The mean fiber intake in 1989 was ~14g per day according to the USDA https://www.ers.usda.gov/sites/default/files/_laserfiche/pub...
I definitely remember the 80s. We eat healthier than my family did in the 80s. But my parents were working class and we’re upper-middle class.
Sample bias and confirmation bias are the reason that you can’t rely on memory and personal experience. If you look at the data we know that people consume far more of their calories from ultra processed foods than they used to. And we know they eat out far more often. We also know that the average restraint meal is lower in fiber than the average meal at home.
The USDA survey does show a minor increase in total fiber consumed, but there are large differences in sample methodology and as I said previously resistant starches are now labeled as fiber. It is common for low carb breads, tortillas and similar to contain 20-30g of “fiber” in the form of resistant starches.
>no way empirically worse for you than similar foods eaten at home. Like I get far more veggies and fibre from a Wendy's burger than when I make the burger at home. This is one of those lazy correlation hand-wavy things people use to invent data where there is none.
Here is a study from the USDA showing that the average at home meal has 30% more fiber per calorie than the average fast food meal. And the previous survey I posted shows a similar difference in the 80s.
https://www.ers.usda.gov/amber-waves/2013/february/americans...
If you eat hamburgers and French fries for every meal, then yes replacing that with fast food isn’t any different.
But according to the USDA that’s not what people are eating for every home cooked meal.
Away from home meals are higher in fat, higher in sodium, have less fiber, fewer vegetables, are more calorie dense, and are served in larger portions.
https://www.ers.usda.gov/sites/default/files/_laserfiche/pub...
You’re the one that said that when studies adjust for these differences they no longer show a correlation.
That is just flat wrong. You used that exact phrase “adjust for”. You can move the goalpost if you want to “I don’t think they are correctly adjusting for all differences” but that’s a very different argument than the one you presented.
The preponderance of the evidence is that fiber has a protective effect against colon cancer and there are multiple plausible explanatory mechanisms.
“I don't know what notion people have about common Western food practices decades ago, but in the 1980s, for instance, fibre was basically non-existent in the common diet”
This is misleading.
1. There is a single gram per thousand calorie difference in aggregate between 1980 and today, so if it was non existent then, it’s non-existent now.
2. Percentage of ultra processed food and food away from home has gone up drastically since 1980, so the types of fiber consumed on average have changed drastically as well.
3. “Fiber” is now a much broader category than it used to be. Resistant starches can now be labeled fiber. If a tiny fraction of the population eats the equivalent a single low carb tortilla with 30g of resistant starch per day that’s enough to completely throw off the aggregate fiber stats. There is much less evidence that resistant starch is protective against colon cancer.
4. Stability in the aggregate doesn’t preclude changes in sub populations. Changes in the way we eat leading to a larger sub population with very little fiber intake.
We know that there have been changes in sub populations. For instance one study that found fiber consumption was stable overall between 1999-2017 found that it dropped among White adults, but the aggregate was stable because it rose among Black and Hispanic adults.
As far as I know the only RCT that fits that description is one looking at people who had already developed adenomas and then tested for reoccurrence.
An RCT actually designed to answer the question would need to last decades and just isn’t feasible.
There are a large multitude of studies that show a negative correlation between CRC and fiber intake even when adjusting for diet, exercise, BMI, income, alcohol consumption etc…
> but the current consensus is that fibre, or lack thereof, is not a contributor behind the rise in colon cancer among younger subjects.
There’s not good evidence to say it’s the sole or primary factor, but saying there’s consensus that is not a contributor is way too strong.
Every 10g per day is correlated with a 10% reduction in all cause mortality (up to 30g). Mostly from a reduction in cardiac and cancer deaths.
“Can produce”. Is doing a lot of heavy lifting there. It can, but then they also regularly write truly bafflingly bad code that no decent experienced dev would ever produce.
I’ll give you a few examples from last week. I had claude whip up a Kafka consumer. After a lot of back and forth I was pretty confident it worked and then I decided to dig into the code. I spotted some real head scratchers.
When starting up the consumer it wanted to pass a specific rebalance strategy, but then it realized that the library it was using didn’t support that strategy, so it needed to use the default.
Instead of just deleting the config option (and maybe leaving a comment), it created a function that removed that key from the map and used it like:
foo: “bar”,
baz: drop_strategy_key(
strategy: “blort”,
bit: “bang”
)
Another frontier model reviewer was supposed to prevent that kind of thing but it didn’t catch it either.
I had 10 public functions that were only called in tests throughout the whole project. Essentially the tests were useless because they weren’t actually testing anything.
Clearly current frontier models are useful. And maybe they produce code that is better than the median dev by many metrics. But I don’t think they are anywhere near consistently producing better quality than a say a 70the percentile dev with 10 years experience.
And even when the quality is better than the median dev, I’m finding that the bugs are not the kind of thing that humans or LLMs are good at spotting.
Employers have much more leverage over H1B workers in the first 2-4 years than they do over other workers, and then they still retain more leverage after that.
This lowers wages. If we’re going to have this program, we need to remove any additional leverage employers have over employees from day 1.
Say that’s 12 months. If their I-40 is approved at 30 months, the employer can withdraw it for 180 days and the approval will be revoked.
So we’re talking 36 months until changing employers doesn’t set you back.
That is longer than the median FAANG tenure. That means you have an employee whose floor tenure is longer than the median.
And even then if at anytime the employee is fired (until they get their green card, which can take many many years), they have 60 days to find a new employer or leave the country.
And depending on where you are in the green card process, you might have to start all over again if you change employers.
A subordinating conjunction like if turns an independent clause into a dependent clause, so you don’t use a comma.
> And almost nobody writes sentences like "I love her and she loves me back" with a comma before "and".
In informal writing, people frequently leave out commas for everything but lists.
But it’s also very common to see it used “correctly”. In any kind of writing where someone would be expected to proofread what they wrote, I’d expect it.
The comma prevents readers initially grouping words incorrectly.
“I called John and Mary answered.”
Your brain will start to parse that as “I called John and Mary” before backtracking once you hit the last word.
This is a big part of the reason you see so many successful immigrant owned businesses. The owners (and family) work nonstop.
The number of men who make it to 65 after having made it to 21 is about 20% higher than it was in 1940.
And the average remaining lifespan for men reaching 65 is about 3 years higher than in 1940.
These factors certainly change the calculations, but the size of the baby boomers is a much larger factor.
If you’re in the US, the top 10% might. But the median residential electrician certainly does not.
Sure Italy certainly wanted to colonize Ethiopia, but they were driven out with the help of the British before they could.
There’s broad acceptance that Ethiopia is an uncolonized country.