Notices by Dan Luu (danluu@mastodon.social), page 4

I think a lot of this comes from people having no idea what a poor person might spend on something. If you eat rice and beans from 50lb bags, that's really cheap. If you want to beat this cost with "purchasing only a rich person can do", you basically have to buy by the pallet. How many middle class or rich people do this?

I do know one person who does this (a formerly poor person, of course) but, in practice, rich people aren't spending less on food because they can buy in bulk or whatever.

Housing is another common example with the appreciation of a home vs. the cost of rent. But almost every time I've had someone brag about how much their home has appreciated, when I've compared how my investments have done (boring mostly world market cap weighted stuff, using some instruments to get leverage), my investments have outperformed, so this is really about having investments vs. not, which is super different from Vimes' boots as stated as well as how people generally reference it.

Yeah, having enough money that your investments throw off significant income is pretty great. So is winning the birth lottery and having a trust fund you can live off of or winning an actual lottery.

But that's a different thing than having to buy lower quality items that cost more in the long run.

If you look at the actual thing, the price of homes purchases by poor people or the cost of rent paid, it's generally lower and not higher.

The idea that middle class folks are eating more cheaply than someone who buys 50lb bags of beans, living more cheaply than someone who, e.g., puts an entire family into a studio in a cheap neighborhood, drives more cheaply than someone who takes the bus, etc., is obviously absurd, so it seems like people just have no idea how a frugal poor person lives

A lot of people are replying with "that stuff isn't as nice as [rich person stuff]". Yeah. I know. But that has nothing to do with Vimes' boots

One person I polled replied with something like "[laugh] have any of these people been to a place where you can buy cheap bulk food [that isn't one one of the expensive membership places like Costco]? Look at the SES of the shoppers!"

Well, there's that one guy I know who owns two planes who buys rice by the pallet, but that's really unusual. I suspect no one who's saying that poor people pay more for food than rich people know someone who does this.

It's amazing how often, when I look into why something turned to junk (consumer products, tools, etc.), it turns out that it's because a PE firm or a PE-like software company acquired the thing and then made an extremely short-term optimized move that wiped out most of the value and potential revenue of the thing, e.g., I was looking into why a formerly active automotive forum is now a ghost town and it turned out they got bought by VerticalScope, which has apparently killed ~1k forums?

If you actually look at how forums scale up and maintain semi-decent discourse, having extremely highly weighted downvotes/flagging is one of the most effective tools (if you don't want all controversial topics killed, you can do what HN does and have mods manually rescue flagged items at their discretion; of course you can do the opposite and have mods try to chase down every bad comment and discussion, but that's harder to scale).

Discourse makes exactly the wrong choice in quite a few ways.

I wish I saved the link to this message, because there was a brilliant summary of the problem when someone asked a jerk why they wrote such a horrible reply

The jerk replied that they needed to let this other guy know they were an idiot an they couldn't downvote them, so they "had to" write a message explaining why the other person was wrong

It's a common fantasy that no downvotes or forced reply with downvote will make people amicable, but no one who observes people can believe this will work

if you watch someone who's road raging, they often try to get the attention of their target and they keep getting angrier and angrier and angrier when their target doesn't notice they're being given the finger or whatever.

If you ever watched a forum migrate to discourse, you can see this exact thing play out. I read the blizzard forums around when I played Overwatch for https://danluu.com/overwatch-gender/ and it was exactly as you'd expect. The number of mean/toxic/jerk-y messages went way up.

Its backend performance also appears to be quite bad, e.g., they banned Bing from crawling them because they couldn't handle the 0.5 QPS of load that Bing was sending their way: https://twitter.com/danluu/status/981992814824378369.

But, computer performance continues to improve and discourse will probably be usable on a cheap phone within a decade and maybe servers will also become fast enough that they'll be able to handle 0.5 QPS of load.

The bigger issue is that the design is strongly anti-user in a lot of ways, e.g.,

One of the major design decisions was to have no downvotes, which was justified as a way to created "civilized discourse", the idea being that people should discuss their disagreements, resulting in "civilized discourse".

I predicted this wouldn't work because people sometimes have a very strong desire to express displeasure, and if they can't do it by downvoting, they'll do it by writing nasty comments. You don't even have to have used the internet to guess this, e.g.,

As a follow-up to https://mastodon.social/@danluu/109798007902048311, I wonder why there isn't a serious, well-funded, attempt to create a modern forum

If you look at Wikipedia's list of forum software, it's all ancient except discourse, and discourse seems unlikely to ever be something great for users

Its performance is famously terrible. People often point out how unusable it is unless you have a fast phone and the founder's response to this has been to rant about how Qualcomm sucks and need to make faster processors

I wonder why nearly every company is, as Patrick Mckenzie puts it, "fundamentally unserious" about hiring.

I've done a lot of hiring for multiple companies and, even when the tech job market was the hottest it's ever been and tons of founders, managers, etc., were complaining that it was impossible to hire, I found it very easy to hire because every other company out there is doing such a bad job you'll do an outstanding job if you do anything remotely reasonable.

Saw a headline about how an MIT student used Playground AI with the prompt "Was trying to get a linkedin profile photo with AI editing & this is what it gave me" and it made her white. Tried this with my profile photo and the result reproduced and I became whiter.

Using a summer photo where I've had more sun exposure, it often turns me South Asian or African. Occasionally (maybe 1 time in 10), I'll become East Asian, but I never stayed Vietnamese or turned into any sort of Southeast Asian.

As much as I respect Keller and think RISC-V is neat, I'll take the other side of this bet and say that RISC-V will not have "take[n] over all the data centers" "in the next 5 to 10 years".

Like https://twitter.com/danluu/status/1554559905167597568, this feels like a fairly safe bet just based on inertia.

A lot companies are running servers that are almost a decade old (e.g., Twitter had plenty of ~8 year old machines) and even companies that are quick to replace machines often still run 4+ year old hardware.

Interesting to see the rise of ChatGPT-enabled Github spam.

The screenshots below are from one account, but if you search a bit it's very easy to find other examples.

Github doesn't appear to be on top of this at all, e.g., the account mentioned in https://github.com/swarna1101/VeChain-Thor/issues/1 has been spamming repos since last year and hasn't been banned (I reported a bad account that was generating abusive spam once and got some autoresponse and then nothing happened and haven't bothered reporting since then).

It's generally considered absurd for startups to compete with public companies on pay, but paying roughly twice as much as our startup competitors (all of whom ended up going under very quickly, with the exception of Transmeta, which managed to drag things out for a while due to the massive amount of money raised and won via lawsuit) and getting 30x the productivity seems obviously worth it?

The range in productivity is so much larger than the range in pay that this seems like an easy call?

I often wonder why companies don't pay more in order to hire more productive employees.

The local Costco has checkout folks who can scan and bag items 2x-5x faster than I get at local grocery stores even when I only have a relatively small number of items, and they do a better job at packing items in to not damage them, are friendlier, etc.

Because almost no one tries to pay for better employees, the market isn't competitive and these people don't cost 2x let alone 5x.

Of course the dynamic range in productivity of engineers is much higher but, similarly, the market isn't that competitive and you don't have to pay that much more.

When I worked at a chip startup, we paid ~20% better than public companies (if you wrote off our options). Our 40 person team taped out a CPU in 1 year. Our most famous startup competitor, Transmeta, had ~300 people and took 4 years to tape out, a 30x difference in productivity and why we survived and they lost $1B and went under.

https://mastodon.social/@HalvarFlake/110022378590690514 reminds me that when I was an intern at Micron, I found that you could easily defeat the security on their security products, which were used for things like set top boxes, by applying heat, which would cause bit flips.

They loaded privilege level from flash onto a tiny RAM so that reads weren't super slow, so you could just repeatedly attack the RAM and reset the box if the heat messed something up and it crashed and you'd eventually get a privilege escalation.