HNHacker News
TopNewBestAskShowJobs

milesvp

3,132 karma · joined July 2, 2010

submissionscomments
milesvp··on How to make sense of any mess
"Assuming we just keep doing what we're doing, where will we be in 12 months?"

I find this one interesting, a business youtuber I follow said he finally realized that his teams all got ~5% better every year if he just left them alone and changed nothing. He said he used to have all these ideas he wanted to implement, but that if they didn’t have a lot of potential upside, they weren’t worth the short term drops associated with reimplementing, retraining and the teams having to relearn and explore their new problem space.

milesvp··on US cities pay too much for buses
I've thought about this a lot, and wonder if the last mile problem could be lessened with an uber style pickup you suggest. I have a civil engineer relative who follows this stuff better than I do, and he says all the pilot programs he's seen (in the US) tend to be wildly unprofitable.

That said, I think that some program like this is essential to bootstrapping a really good transit system. The last mile problem really does stop a lot of would be commuters and is a huge, largely hidden cost, in regional transit planning. You could have fewer, more reliable trunks, that can run less reliably after core commuting hours, all because you have ways of alleviating the pain associated with difficulty getting to out of the way places. This allows people to make life decisions that they might not otherwise be able to make. And once you have a solid core, you can continue to grow it, by continuing to encourage long term ridership. Couple this with increasingly aggressive zoning changes to allow for density, and I think you could really grow out a transit system in 10-20 years.

But this is a fantasy of mine. It would likely be wildly unpopular to run an unprofitable program long enough to make all of this possible, and would probably only work in regions that have the potential for good transit anyways. You'd also need a large cohort of YIMBYs, that while currently growing in many regions, aren't guaranteed to still vote that way in a decade when they have more to lose.

milesvp··on California issues fine over lawyer's ChatGPT fabrications
I know a lawyer who almost took a job in state government where one of the primary duties was to make sure that the punctuation in the bills going through the state legislature were correct and accurate. For her, part of the appeal of the job, was that it would allow her to subtly alter the meaning of a bill being presented. Apparently it is a non trivial skill to be able to determine how judges are likely to rule on cases due to say, the presence, or the absence of an oxford comma.

There was an entire team dedicated to this work, and the hours were insane when the legislature was in session. She ended up not taking the job because of the downsides associated with moving to the capital, so I don't know more about the job. I'd be curious how much AI has changed what that team does now. Certainly, they still would want to meticulously look at every character, but it is certainly possible that AI has gotten better at analyzing the "average" ruling, which might make the job a little easier. What I know about law though, is that it's often defined by the non average ruling, that there's sort of a fractal nature to it, and it's the unusual cases that often forever shape future interpretations of a given law. Unusual scenarios are something that LLMs generally struggle with, and add to that the need to creatively come up with scenarios that might further distort the bill, and I'd expect LLMs to be patently bad at creating laws. So while, I have no doubt that legislators (and lobbyists) are using AI to draft bills, I am positive that there is still a lot of work that goes into refining bills, and we're probably not seeing straight vibe drafting.

milesvp··on I regret building this $3000 Pi AI cluster
Seriously. Imagine the fractal nature of software problems (you do need firmware right?), and add that on top of a layer that has poorly documented errata and non standard peripheral variants. I'm currently making a board to solder onto an existing production run because we finally found the 1 sentence in a 1700page data sheet that explained why many of our MCUs are overheating. Had the datasheet simply labeled VREFP as VDDA, it probably would have triggered a review of the power schematic. Oh, and the CPU overheating was masked by component variation, and CPU binning, which means the overheating wasn't obvious on any of the early prototype boards...

And then, there's the sourcing problem. Components that looked like they were in big supply when the hardware was specced, can end up being in short supply, or worse end of lifed while you're trying to get all the firmware working.

milesvp··on Raspberry Pi Synthesizers – How the Pi is transforming synths
For more context, dsp chips historically were how you could get better better performance than was typical in the embedded space. At a time when you had sub 100Mhz typically in a chip that would be used in musical equipment dsp chips were super useful because they had many op codes that could be done in one cycle that might take many multiples of that in a general purpose chip. One prime example is add and accumulate. This meant that effectively, if you had an 80Mhz dsp doing calculations for you, you effectively had a 160-240mhz equivalent in your hotpath. This bump up in processing is necessary to meet hard realtime deadlines in audio processing at 48khz stereo processing.

Now, processors are so fast and cheap, that the need for dsp is way down. I personally ported the firmware for a popular pedal where the chip used was no longer available at quantity, and while it was tricky to get the same performance on a dual core 240mhz processor with floating point coprocessor, as the 96Mhz processor it was replacing, I just had so much compute, that I was able to port all but the gnarliest simultaneous effects on a chip that cost $6 cheaper but also came with enough ram to not need that $3 component either. And I can tell you, saving $9 dollars off a BOM means the product can retail at least $50 cheaper in an industry where volume matters a lot (no pun intended).

We have passed the inflection point where dsp chips are generally not worth the cost in most audio equipment. I’m not sure I’d use a raspberry pi, their reliability is not what I’d want in my products, but a similar single board computer? maybe. I’m not sure where linux realtime kernals are at the moment, ignoring the added complexity of a full linux running to process audio, audio requires hard deadlines. You will hear pops and other artifacts if you miss these deadlines, and buffering audio is also not something you can do for long, artists can feel delays in the ms. But, a single board computer running at ghz speed, gives an awful lot of compute so long as the OS can gaurantee your audio thread won’t get starved.

milesvp··on Microsoft is officially sending employees back to the office
It’s a common thing here on HN to believe that remote is superior for productivity, and I’m always reminded of Richard Hammond’s observations about open door vs closed door coworkers. He noticed over time that the closed door workers were more productive. He also noticed that the closed door workers were less impactful in their fields years later. His were observations in R&D settings, but I suspect they can be extrapolated. People who are interrupted get less done. This seems largely indisputable, but what is the other takeaway? People who don’t interract with peers don’t course correct enough, seems to be solid advice based on what we know about the OODA loop. People who don’t interact with coworkers don’t get enough time saving advice? I know I’ve saved lots of effort by having coworkers who know things I didn’t about related problems.

What complicated things, is return to work will cause all the best to rethink their employment. I’ve seen HBR surveys that suggest the top talent is ending up places that allow them to stay remote. I think this leaves businesses in a tight place. I have every reason to believe that companies with lots of employee interactions have better acceleration/trajetory than fully remote, but it’s a big hit to lose top talent. And remote may have so much velocity from gaining this talent that they don’t care about the acceleration tradeoff.

Further, concentration of talent in a region also cannot be discounted. Certain things can’t happen without the exchange of ideas (partly why I think cities/counties should ban non competes). I don’t know how much a given company can control this concentration of talent, but I know that Seattle wouldn’t be what it is without Boeing, and then Microsoft attracting very smart people.

milesvp··on U.S. added 911k fewer jobs in year through March than reported earlier
The interest rates were wild in 1982. My dad said that no one would lend for a mortgage, and we ended up doing seller financing to buy my childhood home.

Also, not enough people talk about housing prices in high interest rate environments. Mortgage payments are the thing that tether housing prices, and act as a lever. We see it today with just an extra 4% with low housing volume and people reluctant to sell, because then they’d need a new mortgage at the new rate. People who would like to downsize don’t since their mortgage payments often ends up similar.

milesvp··on The buyer-pull and seller-push theories of sales
It doesn’t even have to be as nefarious as trying to maximally charge. Very often what seems like a simple product offering can actually be quite nuanced with services or goods bundled or unbundled. I’ve listened to many sales calls where the buyer was asking for the wrong thing, and giving a quote for that thing doesn’t really do anyone any good. In my experience the best sales people are really trying their best to tailor their offerings to the needs of the customer even going so far as pointing them at cheaper products or even another company if it fits their needs better.
milesvp··on How to make things slower so they go faster
What I find fascinating, is how much this concept scales to places it seems like it shouldn't. I had taken the idea to heart early in my life for anything that require dexterity. But it wasn't until mid career that I saw it work at an organizational level. At one point the team I was on stopped promising so much. We essentially decided to slow down. I don't quite remember what lead us to this mindset, though I know our weekly retrospectives were part of it (we had some really good retros, like I cry at the thought that I will likely never have that level of mutual trust in a team again). And, what was sort of unexpected, was that our velocity basically went up. We knew we wanted to make sure we focused on higher value items, and push back on low quality requests, but the amount of requests we could accommodate also went up along with the average value. I still don't fully understand the theory behind it, certainly we were using a lot of cycles on low value things, but just promising fewer deliverables allowed us to deliver more. I know that brains are bad at time slicing, but this seems to also expand to the organizational level too...
milesvp··on The air is hissing out of the overinflated AI balloon
Even better when the customer just plain orders the wrong thing. I have a vivid memory of my pregnant wife and I at breakfast one morning, and she ordered apple juice. When the waiter brought apple juice she said she’d ordered orange, and when the waitress looked confused, I had to remind my wife she’d ordered apple. Total brain fart on her part (pregnancy brain seems to be a very real thing). Waitress didn’t care, especially after we laughed it off, and brought orange. But I know I’ve undoubtedly done the same thing, so I’m patient with mixups.
milesvp··on Making games in Go: 3 months without LLMs vs. 3 days with LLMs
> we didn’t see a proportional explosion of good game, just more attempts.

I'm not sure the 2 of you are disagreeing. We definitely saw an explosion of indie games. In 2010, there were less than 10 indie games released on steam per month. By 2022, there were ~500/mo, and today there's ~750/mo (I expect that the 250/mo jump around 2022 can likely be attributed to LLMs).

What's hard to say is if this increase significantly increased the number of good games. Mostly because "good" is highly subjective, but also, I think something else happens. I've been playing games for the better part of 40 years, and what I noticed, is that in that time, the number of must play games each year has largely gone unchanged, despite the industry being orders of magnitude larger than it was 40 years ago. But that is also tricky, because 2 things happen every year, our standards get higher, and our preferences get more refined.

https://steamdb.info/stats/releases/?tagid=492

milesvp··on 7 years later, Valve's Proton has been a game-changer for Linux
I haven’t needed windows for a game in over 3 years. between proton and lutris I have all the games I’m interested in covered. There are multiplayer games I can’t play, but I have little trouble with blizzard games. I’ve had a few games I’ve had to come back to when proton fixed things, maybe 2% of the games I’ve tried. AMD graphics cards have done me well.
milesvp··on What medieval people got right about learning (2019)
I would argue that learning in groups is potentially exponentially more effective. There is a lot of individual interactions that go on when people are all learning something at the same time. Tidbits that they each share with one another. And emotion that social interactions evoke is a powerful motivator for mental rewiring.

But I don’t want to dismiss your insights. I’m curious what the difference is that I’ve experienced. Certainly just putting random people together isn’t nearly as beneficial as grouping by ability, or motivation to learn the topic is useful. Maybe that is a necessary requirement for effective group learning?

milesvp··on Did California's fast food minimum wage reduce employment?
I’ve seen some interesting research suggesting that higher minimum wages lead to lower turnover, which can lead to some very real cost savings. I had an interesting epiphany while watching a business lecture about calculating costs associated with hiring, that there are very real points in the minimum wage curve (which should be laffer shaped) where raising the minimum wage has the potential to both increase labor participation and decrease total labor costs.

I now like to joke that minimum wage laws are subsidies for businesses too dumb to factor in hiring and turnover costs.

milesvp··on Diet, not lack of exercise, drives obesity, a new study finds
I once worked a night shift stocking job just after college. I was in phenomenal shape without hitting the gym. I was at my lowest weight with a ton of lean muscle as a result of moving heavy loads and stocking paint every night. I also did the math at one point, and given the size of the warehouse I was probably walking quickly 8 miles during my shift. It became a chore to eat enough calories every day. Somewhere around 4000 calories/day, you may still be hungry, but you are generally full. Also food sort of becomes boring, and the desire to eat just isn't as strong.

That said, it was 4-6 hours 4 nights a week. That is a lot of time to spend to burn all those calories. It is really not hard to eat an extra 100 calories per day, but it takes a lot of effort to burn an extra 100 calories. It's the asymmetry here you absolutely have to respect. Further, at least for me, there is another asymmetry in terms of satiation vs hunger. It is much easier to be slightly satiated than it is to be slightly hungry. What this means, is that there is a tendency to be driven to eat slightly more than your body needs. This is partly why the GLP-1 drugs seem so effective, is that they seem to flip this asymmetry in the other direction, which means weight loss is the default, instead of weight gain.

milesvp··on Ask HN: What are your thoughts on the impact of AI on programming careers?
>5 people over a span of 6 months

As a very seasoned developer, I have seen lots of projects built by a single person in a short time that “would normally” be built with more people over a longer period of time. What is invariably missing from these projects are the very things that the 5 people would have added. The main thing that is usually missing is robustness. These single dev projects will usually fall over hard when they stray from the happy path (why would anyone enter a value other than a small digit for this input???).

Data integrity is often ignored as well, both internal to the application and how the data flows through the rest of the systems and the org as a whole.

Security is generally an afterthought as well.

And then, there’s sort of the elephant in the room, which is part of why these projects take a long time, is that they need to be defined and refined in the first place. It is not uncommon to find some edge case that requires a lot of effort to decide how best to handle it, and it’s often the product owner who needs to decide tradeoffs for thw product.

Incidently many of these reasons are why offshoring also fails.

So, where does that leave things? AI will undoubtedly allow a team of 9 to do what a team of 10 did 2 years ago (maybe even 3 of 5). AI also will potentially help with each of the things I’ve mentioned as well, I’ve had good luck with AI helping with missing test coverage. I’ve also had a lot of luck with AI helping where I have gaps in my knowledge about things. So I expect the quality of software will get much better too.

I will say, right now there is a big shake up of the industry. I expect short term prospects to be bad in the same way the offshoring frenzy of the early 2000s. What you are going to find, is that long term demand for technical talent will only go up. In the same way that efficiency in anything tends to increase demand, things get cheaper and people find more uses for said thing.

Another thing to think about; I sort of saw the writing on the wall 6 years ago. I continually saw the barrier to entry fall in software. So I changed my niche and moved closer to the hardware, where fewer people seemed to be entering. I will tell you that the places LLMs struggle with the most are the places others haven’t spent much time documenting. In the domain I currently work, it is impossible to use any code auto generation, there are too many places it transposes things or makes up api calls. But the code it generates can be very useful when the documentation is ambiguous and some obscure forums post likely has an answer the LLM can draw from. So it is possible to stay ahead of AI by moving to the fringes.

milesvp··on LLM Inevitabilism
I think this can't be understated. It also destroyed search. I listened to a podcast a few years ago with an early googler who talked about this very precipice in early google days. They did a lot of testing, and a lot of modeling of people's valuation of search. They figured that the average person got something like $50/yr of value out of search (I can't remember the exact number, I hope I'm not off by an order of magnitude). And that was the most they could ever realistically charge. Meanwhile, advertising for just Q4 was like 10 times the value. It meant that they knew that advertising on the platform was inevitable. They also acknowledged that it would lead to the very problem that Brin and Page wrote about in their seminal paper on search.

I see LLMs inevitably leading to the same place. There will undoubtedly be advertising baked into the models. It is too strong a financial incentive. I can only hope that an open source alternative will at least allow for a hobbled version to consume.

edit: I think this was the podcast https://freakonomics.com/podcast/is-google-getting-worse/

milesvp··on Knowing Steam players are hoarders explains why you give Valve that 30%
I think you may even be missing a lot of things we completely take for granted today. I saw a talk by a long time RPG indie dev a few years ago. Said 30 years ago, he needed a full time employee to do all the things he gets for using steam. Things like distribution, customer support, returns. Said he was more than happy to pay 30% at his volumes to not have to have that head count.

edit: Oh, and payment... I think it was Jeff Vogal's talk I'm thinking of.

https://www.gdcvault.com/play/1024944/Failing-to-Fail-The-Sp...

milesvp··on Writing toy software is a joy
Yes, you have to be very careful when querying LLM's, you have to assume that they are giving you sort of the average answer to a question. I find them very good at sort of telling me how people commonly solve a problem. I'm lucky, in that the space I've been working has had a lot of good forums training data, and the average solution tends to be on the more correct side. But you still have to validate nearly everything it tells you. It's also funny to watch the tokenization "fails". When you ask about things like register names, and you can see it choose nonexisting tokens. Atmel libraries have a lot of things like this in them

#define PA17_EIC_LINE PIN_PA17A_EIC_EXTINT_NUM #define PA17_EIC_BIT PORT_PA17A_EIC_EXTINT1 #define PA17_PMUX_INDEX 8 //pa17 17/2 #define PA17_PMUX_TYPE MUX_PA17A_EIC_EXTINT1

And the output will be almost correct code, but instead of an answer being:

PORT_PA17A_EIC_EXTINT1

you'll get:

PORT_PA17A_EIC_EXTINT_NUM

and you can tell that it diverged trying to use similar tokens, and since _ follows EXTINT sometimes, it's a "valid" token to try, and now that it's EXTINT_ now NUM is the most likely thing to follow.

That said, it's massively sped up the project I'm working on, especially since Microchip effectively shut down the forums that chatgpt was trained on.

milesvp··on High-quality OLED displays now enabling integrated thin and multichannel audio
Yeah, I don’t know, I suspect you’re right. I know that 25 years ago there were big flat speakers thay drove sound with little holes. Those things could only handle high ends despite being very large themselves. But I know that with DSP you can treat an array of microphones as a single microphone, and the diameter of the array sort of dictates the size of the microphone. Speakers and mics tend to sort of be opposites, in that you can drive or be driven by either. For years I’ve wanted to build a wall of cheap speakers and experiment with this. I suspect I’d find out what we already know, which is to get deep frequencies you have to move a lot of air, and an array of small piezos can never move a lot of air…

That said, it might matter a lot what the substrait was. If it was light and flexible, you could maybe get all the piezos to move it in a way to get a very deep frequency. You could probably get deeper frequncies than the power output of the piezos by taking advantage of the resonance frequency of the material. But you’d be stuck with that one frequency, and there’d be a tradeoff in response time

milesvp··on Apple App Store guidelines remove ban on encouraging external payments in US
I feel like you're being disingenuous with your choice of franchise. 30 years ago there were much worse toys. There were capsule machines that randomized what toy your quarter would give you. There were toys that you could buy random assortments at the toy store (M.U.S.C.L.E was one if I remember correctly). You could buy trading cards too. It's not that kids are marketed to (which is arguably its own problem), it's that the randomization is really not good for creatures that utilize associative memories (not sure if other intelligence avenues will be as susceptible to near misses, but likely it's a feature of intelligence in general to be stupid about randomness). And this has only been ratcheted up in the last 30 years.

What you may be missing, if you don't have kids, is just how insidious modern arcades are. They really opened my eyes in a lot of ways to the problem in general, since I just avoid a lot of the other modern invasive gambling mechanics. Most of the games are now just thinly veiled gambling machines. There are a few classics, like pacman still, and they eat quarters, but they are not programmed to randomly modify the game itself. Claw machines these days all have their claw strength randomized and is unknowable value that changes from play to play. And almost all the games I see at kids venues have some similar mechanic.

But it's not just the arcade. The rise of skinner boxes have become ever more weaponized (for lack of a better term?) in the last 30 years, as data collection has become cheaper and easier. I can't even imagine gacha mechanics in any of the games I played 30 years ago. Like, here, send Nintendo a dollar, and you can get a code for a better sword in Dragon Warrior? I would have mailed that dollar faster than you can imagine (I then would have shared the code, so of course this wouldn't work, but still, I would have sent the dollar). And for what? so they can make the games even harder?

This is a real problem beyond just teaching kids to ignore marketing. I don't have a solution other than trying to shield them until they're old enough that they're less likely to develop real addictions.

milesvp··on I just want to code (2023)
I am a firm believer that the brain has finite resources. It is fundamentally no different than a muscle, in that, the muscle has resources at its disposal, and using those resources causes them to deplete faster than they can recover. This means, that at best, you can optimize how you use your brain, similar to how you can optimize how you use your muscles. You can train, to get "more" out of your brain, similar to muscles. But at the end of the day, there is only so much your brain can do. I have found, through fairly good time keeping, that I can do ~6 hours of deep work in a 9 hour period. Maybe 7 in a 12 hour period. This is the amount that puts me at the edge of burnout, and requires no other commitments in my life. 4-5hours in a 7.5 hour day is sustainable. Others may be able to do more, or less than this, but many people are fooling themselves if they think they are doing good work that is also deep for more than this, for more than say 4 weeks at a time.

That said, if you can find ways to use very different parts of your brain... well, then you can squeeze out more performance. In the same way that in the gym, if you find exercises that isolate different muscle groups, you can squeeze more sets in throughout your workout.

I have seen this with people who can produce a lot of output, they tend to not do the same thing all day, but find ways to work on unrelated projects. So, I'm not terribly surprised that you have an urge to code now that your brain is focused on non code related things for a large part of your day.

milesvp··on Windows 2000 Server named peak Microsoft
XP also inexplicably required at least twice the ram as 2000. when XP came out that was a significant cost, and I personally was able to salvage many laptops at the time by downgrading them from XP. Eventually XP became the default for me because ram got a lot cheaper and the service packs and driver support made it more viable.

But then, tangentially, I started using ubuntu at work, in a sort of misguided belief it would make me a better sysadmin, and it was only a matter of time before I couldn’t stand windows at home as well.

I thought win7 was pretty solid, though I didn’t upgrade until well after win8 was shipping. But lucky for me, proton finally got really good, and that allowed me to basically skip win10+. Now it’s only for the rare tool that I even boot into my windows partitions anymore. When I do, being bombarded by random attention grabbers is completely jarring and I want flee as fast as I can.

milesvp··on Trump temporarily drops tariffs to 10% for most countries
Normal economics almost certainly does talk about exporting currency. It’s a key point in freshman macro economics, and central to modern keynseian theory. And while there are some solid alternative to keynes, I doubt that even hayak would claim that you can’t export currencies.
milesvp··on Trump temporarily drops tariffs to 10% for most countries
> Instead of being seen as a safe haven, U.S. debt itself started to look shaky

There's also a more basic answer to why US debt got more expensive. There is no such thing as a "trade deficit". You cannot import something without exporting something else. That something else, in almost 100% of the situations, is dollars. What do you do with dollars, if there's nothing you want to buy? You buy debt. So, every trade deficit ultimately tends to export debt. We have long known that government debt is proportional to trade deficits.

The problem with tariffs, in this context, is that they are being used as a signal that we don't want to sell people dollars anymore, explicitly saying "no more trade deficits". In doing so we are choking off the supply of dollars used to buy debt (rather abruptly), so that US debt is chasing fewer dollars and has to give a bigger yield as a result.

What's worse, is that because we are signalling that the tariffs are meant to be permanent, and we're seeing reciprocal tariffs as a result. When that happens, it means that using dollars in the future will purchase fewer US goods, so the time discount of money has gone up too. Further driving up yields to make bonds attractive enough to sell.

milesvp··on Ask HN: Do you still use search engines?
There is a class of problems I no longer use search for. I find LLM’s give really good results for things like command line usage. Or even things like configuring an application. Basically anything that can summarize lots of disperate sources.

Conversely it’s a huge mistake to rely on LLMs for anything that requires authoritative content. It’s not good at appropriately discounting low quality sources in my experience. Google can have a similar problem, but I find it easier to find good sources there first for many topics.

Where LLMs really replace modern google is for topics you only kind of think should exist. Google used to show some pretty tenuously related links by the time you got to page 5 results and there you might find terms that bring you closer to what you’re looking for. Google simply doesn’t do that anymore. So for me, one of the joys is being able to explore topics in a way I haven’t been able to for over a decade

milesvp··on US tax return filings drop by nearly 1M, extension requests rise
The way I understand it is that some people put off filing as long as possible because they get a better return on other investments. They’re willing to pay interest and other fees on their taxes in order to pay as late as possible. I haven’t looked into this personally because I tend to be risk averse enough to pay off relatively low interest rate loans, but I’ve heard at least one personal anecdote once removed to this effect.
milesvp··on A Small Violin Part Highlights Bigger Problems for the Global Economy
I don't know about any credible economist currently making positive predictions (marginal revolution might be a good place to keep your finger on the pulse), but I did see a paper a few years ago that showed protectionist tariffs in britain several hundred years ago had a strong correlation to current day prevalance/strength of current industries. Which is to say, that industries with higher tariffs had a stronger pressence today.

All things being equal, I might expect that strategic tariffs on important sectors to be vital in maintaining them. But, given how global the economy is, and how many more non-obvious substitutions there are today, I'm not sure these protections provide as much value going forward as they did 200 years ago. Worse, is blanket tariffs are not likely to do anything in the short term but drive up local prices to match foreign import goods costs, and is likely going to be a major wealth redistribution event towards physical capital holders.

milesvp··on RIP Val Kilmer: Real Genius .. the Film Nerd Culture Deserves (2015)
War games was incredibly unusual too. The script writers, this was their first screenplay, they had no idea thay no one in hollywood cared about technical accuracy in scripts and spent a whole lot of time trying to get the tech details correct.

Also, not sure if you missed it, but Manhattan Project was part of the trifecta for me from that era that fed my desire to grow up to do intellectually difficult things.

milesvp··on When the Dotcom Bubble Burst
My experience with LLMs is that they are vastly superior to google at the moment for doing research on topics you know very little about. They make doing research fun and productive again, in a way that google has failed at for over 15 years. I used to be able to google a topic I only vaguely knew existed, and after a few tries and scolling through 10’s of pages of results, I could find a term that would bring me much closer to the term of art that would crack open the topic for me and give me good results. That is nearly impossible today. Google simply doesn’t populate page 4 of a search term, and it doesn’t have booleans to filter down either. But with an LLM, I can poke and prod it until it gives me something that I can then use to reference real sources.

LLMs also seems to be really good at giving sort of the average (modal?) answer on a topic, so I find it useful for trying to get an average understanding of something. One example I’ve found useful is getting it to interpret ambiguously worded datasheets. I don’t necessarily trust its answers, but it may show that I’m totally thinking about something wrong.

← PreviousPage 3 of 32Next →