HNHacker News
TopNewBestAskShowJobs

jimberlage

418 karma · joined September 3, 2019

jamesberlage@gmail.com
submissionscomments
jimberlage··on Ask HN: How to onboard yourself to a new product/industry in a new job?
This is fantastic advice! One additional thing I do is

Identify people they don’t work with, that you would naturally assume they would. Typically this means:

Your understanding of how processes flow in this industry has some gaps, or some thing you thought was crucial is a solved problem that people don’t think about; or

There is tension and some history here, tread lightly while you figure out the office politics; or

You’ve identified a legitimate opportunity! Be careful here though, you need to understand if some higher ups have misaligned incentives to keep the status quo before you suggest changes.

jimberlage··on Hot takes on Devin, the AI software engineer
I like this analogy, especially because lots of homeless people sit on the beach without things to do, and we don’t really envy that they have the time to write the next great American novel.
jimberlage··on Hot takes on Devin, the AI software engineer
It feels like this might be sarcastic? But if not

Based on how the US currently rolls out universal programs, UBI probably looks a lot more like the current Section 8 housing program. They’re gonna guarantee something minimally livable, not beachfront property.

jimberlage··on Brain waves appear to wash out waste during sleep in mice
At first I was like, “oh, that’s a reasonable perspective,” and then I thought about it more and it kinda isn’t?

When I buy a keyboard, the manufacturer doesn’t run a shipping company, so they use $x from the purchase price to subcontract shipping. (They tell me it’s separate at billing, but they don’t make me ring up DSL myself either.)

When I hire an electrician, they buy materials from Home Depot, so they use $x from the purchase price for materials. (They sometimes break down the bill into materials and labor, but they don’t make me drive to Home Depot myself and buy every part.)

When the public hires academics to do research, they have administrative overhead and have to hire a publisher, so they use $x from the grant for administrative overhead and…

Abdicate responsibility for the $y needed for publishing and pretend the public didn’t intend for part of the grant money to go to that??

jimberlage··on C++ Safety, in Context
Rust has an industry and a hobbyist ecosystem, whereas I’m not sure if Ada ever had the hobbyists on board.
jimberlage··on Perf Is Not Enough
Yeah, that was the most interesting point for me as well. If you view it through a management lens, it makes sense; "we hire top CS people, let them do the top CS problems, JDBC drivers aren't a core competency so we can outsource that."

The problem is that people can do a bad enough job at the non-core stuff that it doesn't matter that you're good at your core competency. Outsourcing is not a free ride.

jimberlage··on Ask HN: Why is the crypto booming again?
Did the Dutch Tulip craze also go through a graph like bitcoin? Where it dropped as everyone realized the craze and then went back to its highs? Bitcoin behaves more like a currency and less like tulips.

I think the Fed not providing an easy API for a non-dollar token backed by gold is the real thing that keeps crypto around.

jimberlage··on Ask HN: Why is the crypto booming again?
Oddly enough (as someone who feels lukewarm about crypto) isn’t this a really strong case for it sticking around long-term?

Like, “crypto will go away as soon as humanity stops feeling fear or being greedy”?

TBH, if anyone really wants to take the air out of crypto, they need to make an easily accessible, divisible, provable deflationary asset, that doesn’t require a lot of work to find or store (so not real estate, gold, guns, rare LEGO sets, etc.)

jimberlage··on Bluesky is rolling out Hashtags
This was an important differentiator for Twitter; even if you use a thread reader, the author can’t assume that’s the format their text will be read in. So they have to make the best of both worlds, distill the content into its essence and build long form content where each paragraph stands on its own. Bluesky has an opportunity to have a great first-party thread reader experience.

Longform content where the author is incentivized to see how it reads in chunks is an underrated thing.

jimberlage··on Let's Ban SMS 2FA
Out of curiosity - do you think that difficulty of installation is what keeps people from installing a vault door on their home?

I imagine even if it were easier, you would see low adoption, because “I’ve never been robbed but I have locked myself out, does this mean I can’t call a locksmith anymore?” Would be top of people’s minds.

jimberlage··on Let's Ban SMS 2FA
Yes, and sometimes even locked houses are broken into.
jimberlage··on Let's Ban SMS 2FA
So the case against SMS 2FA boils down to

“There are two factors, but the second factor is something a determined actor can get around by SIM swapping.”

But there are still two factors and SMS 2FA handles disaster recovery much better than the listed alternatives for most people.

This argument strikes me as kind of like - “a determined actor can get around a deadbolt pretty easily, so the standard for homes should be a vault door and fines if you leave a hide-a-key rock anywhere near your house.”

jimberlage··on UI = f(statesⁿ)
The right way to handle this is - store both the user input and the validated value (potentially null) and an error state (potentially null.) Then, when the user takes action to submit the form, validate and set the error state if needed. That way you’re not constantly setting the error state on typing, but when some concrete action is taken you can signal what went wrong.

Admittedly, anecdotal, but most of the guys in my career who have been really into UI = f(state) only consider the validated state, and wish the stuff that’s purely user input and never sent to your backend would just go away.

And when you consider that state is both what you want to send to the server, and a bunch of intermediate state that only ever means something to the client, the UI = f(state) idea (while true!) doesn’t seem all that helpful. Technically every app is a function of state, but if your definition of state is that broad what help is the idea?

jimberlage··on UI = f(statesⁿ)
TBH - I find the opposite is true. Far more complaints have been of the form “this isn’t an illegal state, the dev mistakenly thought so because they’re not a domain expert - why won’t they let me do this?”

And some illegal states are useful. Letting a form exist with a field the user can’t fill out, along with disabled logic and a helper message, is often the best way to onboard users to your tool. Lots of proponents of making illegal states unrepresentable take those fields away so they don’t have to muck with validation logic, which takes away your best way to explain how to use your tool to a user.

jimberlage··on Commercial property deals in the US are starting to pick up at deep discounts
I live in a place like this with multiple kids in my 30s, and it’s fantastic.
jimberlage··on Is the "modern data stack" still a useful idea?
I don’t have great links handy, but read up on observability.

You’ll want to read up on what people do on error, especially in the case that contracts change (search “api versioning”.)

jimberlage··on Cousins are disappearing. Is this reshaping the experience of childhood?
This is absolutely correct. Let’s assume a hypothetical 30 year old who bought a house at the absolute peak of 30 year fixed interest rates, 16.64%. The median US home price was $70,400. Assume 20% down.

Let’s also assume a hypothetical 30 year old buying a house in 2023. 30 year fixed rate is 7.00%. The median US home price is $417,700. Assume 20% down.

Looking up inflation, a 1981 dollar is 3.37 2023 dollars.

The 1981 homebuyer pays 57% of the 2023 homebuyer’s down payment, and only pays 19% more each month. And that’s cherry-picking the absolute worst time to buy for that generation.

Down payment 1981 (inflation adj.): $47,440.60

Monthly payment 1981 (inflation adj.): $2,650.50

Down payment 2023: $83,540.00

Monthly payment 2023: $2,223.17

jimberlage··on Why it's impossible to agree on what's allowed
A reasonableness test (like what a court system uses) makes this quiz really easy to answer. Basically everything is either allowed or a margin case except for the person driving their Civic in the park.

TBH, I think if you have this quiz to a non-engineer, they would come away wondering what the point is.

Perhaps the answer is that content moderation should just let a judge make a decision and maybe appeal?

jimberlage··on The web is mostly links and forms
Yep! Picture this from the engineering or product manager's point of view (especially in the case where it really is simple, and the app is links and forms with a consistent look and feel.)

Imagine a split for a Rails app (like what was common in 2000-2010) where you have

- Backend engineers managing most of the logic in ERB templates

- One designer-turned-frontend guy who is deep on HTML/CSS/JS, spends most of his day writing CSS so the backend engineers don't have to

CSS, even for people deep on it, is not very fun. It's especially not fun if your backend devs control the codebase where you can actually define the HTML tags that make up a component. So a really common situation was

- Designer-turned-frontend guy burns out, or wants to transition into pure development

- You have a hellish search for their replacement, since the pitch is "seeking a person who is interested in a low terminal IC position, is interested in the least fun part of web tech, and whose hand will be slapped if they dare to touch the backend engineers' code." In practice once you lose that one guy, you have months of a gap before you find your CSS expert again.

- In the meantime, you have a workforce that doesn't get why look & feel is important and stakeholders who can't understand why no one can put polish on any of their new GUI features, despite building a GUI.

SPAs made this much, much easier. Now, frontend devs

- Have full control over the structure of a component for styling

- Don't have to get the backend devs' permission to change things they know they need to change to style things correctly

- Can make reusable components with the same styles, allowing design systems to take off in ways they didn't before

- If the person wants to branch out and get more programming experience, some of the logic is on the frontend too. The job isn't "all styling, all the time, leave the logic to the other teams" and there are areas to show skill. Retention is 10 times easier and there are fewer key people who can never take a vacation.

- And critically - you have a team for whom look and feel is not an afterthought, with the power & ability to make code changes. No more explaining to stakeholders why buttons look totally different page-by-page and why the team can't be convinced that consistent branding is important for businesses.

jimberlage··on The web is mostly links and forms
I agree, with one caveat - the web is links, forms, and styled components.

It’s the styled components bit that makes everyone reach for react over ERB/HTMX/whatever. Web components wants to solve this but the API is just not there yet for smaller teams.

I think people who want simplicity are on a great path, but without addressing how bad the intersection of CSS and HTML frameworks is compared to React & friends, it’s an uphill battle.

jimberlage··on Empirical failures of the claim that autistic people lack a theory of mind
A part of all psychology diagnoses is that it must both match a definition and negatively impact someone's life. So if you never had to care about what neurotypical people think, then it wouldn't rise to the level of a problem and there's no diagnosis.

You see people care about this stuff because they go to their therapist and say, "boy there are a lot of neurotypical people I have to predict, and being slightly better at reading autistic people doesn't make up for it."

jimberlage··on Empirical failures of the claim that autistic people lack a theory of mind
It's also worth noting - the concept is bottom-up, practitioner-driven, while the experiments that get at theory of mind are top-down, academic-driven.

Theory of mind would exist even if it stopped showing up in academic papers, only because it makes sense to have a shorthand for "my patient keeps complaining about uncomfortable social situations, and after I dig in, it turns out they were missing important social cues". That was known by therapists long before we started trying to operationalize theory-of-mind issues for experiments.

It's really, really tough to design an experiment for this, because it's not super easy to design a multiple-choice questionnaire that will predict if you'll have difficulty understanding subtle social cues. It's a thing that a face-to-face interview would uncover, but not necessarily someone's written output.

You can also run experiments where an investigator interviews the person and records when they miss a social cue, but even then lots of high-functioning autistic people would be able to pass that even if they genuinely have trouble interpreting social cues on a regular basis.

jimberlage··on Empirical failures of the claim that autistic people lack a theory of mind
I've heard that theory of mind issue described as "cannot reliably make predictions based on what other people might think or do.", which is more nuanced than the other definition of "doesn't understand that other people have minds of their own."

Using the first definition, about prediction, makes a lot of sense and is a useful psychological concept.

The first definition would predict theory of mind issues in unrelated disorders, which I think you do see:

- Autism: the mind's ability to process social cues is not there or severely diminished. You'd expect these individuals to be bad predictors of others because they don't get or can't use sensory information that others do.

- Schizotypal personality disorder: they experience intrusive thoughts. They might be bad predictors of others because they assume that everyone else gets intrusive, paranoid thoughts on a regular basis.

jimberlage··on Empirical failures of the claim that autistic people lack a theory of mind
This was my takeaway as well - theory of mind can sometimes be described as 1. "literally does not understand that other people have their own thoughts" or 2. "cannot reliably predict what other minds would think". Most high-functioning autistic people I've met clearly don't meet the first definition, but definitely meet the second.

The second definition makes more sense to me since it also comes up in non-autism diagnoses, like personality disorders. They have trouble reliably predicting what other minds would think, but for different reasons.

jimberlage··on Microsoft seeks Rust developers to rewrite core C# code
MS has lots of battle scars from the security headaches they went through in the 2000s/2010s.
jimberlage··on Never Retire
Alternatively, retire and find a fulfilling volunteer opportunity you're passionate about. Most volunteer places need people during the workweek, and the best fit is college students or retirees.

If you need to be super entreprenurial, start the volunteer organization you'd like to see in the world.

jimberlage··on Tear up unused parking lots, plant trees
I’ve lived in areas with this parking problem in multiple cities for over a decade now.

I can confidently say, I don’t care about this problem at all. Parking further up the street from my house is a small, small, small price to pay for the benefits of being walking distance from interesting things.

jimberlage··on Safetyism killed playtime
Are you lucky, or are necks less likely to break than arms due to evolutionary pressure?
jimberlage··on Hell is other people: performance management at Big Tech
This is really good - lots of studies from I/O psychologists measuring at companies in this analysis. This makes a strong case for anonymous peer reviews in addition to the manager’s review (peer reviews are the most correlated with performance but the manager review is correlated as well.)

If I read this correctly, this correlation is despite all the drawbacks people bring up in this thread, since they would be baked into these datasets too. Seems like you would want multiple anonymous peer reviews to try and get something predictive.

jimberlage··on Hell is other people: performance management at Big Tech
I’m trying to find the study you mention, but a huge caveat IIRC is that these were not people who knew each other, but a random sample of experimental participants (likely college students.)

I’d expect this effect to be less pronounced among a group of peers who sees how the others work on a regular basis. Still there, but a lot less pronounced. Behavioral psychologists don’t tend to have access to a group of peers like that, though.

← PreviousPage 4 of 6Next →