linguae

I had the same experience; grad school is the best preparation for grad school. I had wonderful experiences participating in NSF Research Experiences for Undergraduate projects during two of my summers as an undergrad, and I also had a wonderful eight-month internship in a research lab at a major Japanese IT company, but grad school was a different beast, even though I was a grad student at the place I did my NSF REUs. The internships felt like the “demo version” of the real thing.

Come to think of it, I had similar whiplash in industry of full-time jobs being far more stressful than my internships, which I now realize were more-or-less “demo versions.”

To make a long story short, after grad school and working as a researcher in industry for nearly a decade before becoming a professor at a community college (where the emphasis is teaching), I learned that I love doing research as a hobby but not as a career, where there are many pressures.

I call it the “it’s not delivery—it’s DiGiorno” construct. I generally tune out when I see this construct, which is so prevalent on LinkedIn these days, which is inundated with clichéd writing.

Unfortunately the price hikes are affecting the entire industry, and it's not just high-end rigs for gaming and local LLMs. Game consoles and even the Raspberry Pi have gone up in price. In fact, Apple was one of the last holdouts; it didn't start raising prices until June.

I paid $96 for 32 GB of DDR5 RAM for my Framework 13 laptop back in December 2023. When I checked a few months ago, the same RAM cost around $400; I don't know what prices are like now.

I agree. I would switch to an open source spiritual successor to even Tiger in a heartbeat if it were available.

The difficult part is replicating the ecosystem of applications. From a technological standpoint I would prefer a modern Linux distribution with KDE or Cinnamon over using 2005-era Mac OS X Tiger, though I’d gladly use a version of Tiger that was retrofitted with a modern Darwin core, modern libraries for security purposes, and an up-to-date web browser. However, Linux does not have Microsoft Office, the Adobe Creative Suite, Preview.app, Keynote, Dictionary.app, and other pleasant Mac apps. The FOSS alternatives are not as polished. Even venerable Office 2008 for Mac feels nicer than LibreOffice, and it’s compatible with the XML-based Office formats.

I guess for now I could run Tiger in an emulator, though admittedly current operating systems serve my needs, even if I prefer the polish of Tiger and Snow Leopard. I get really good performance running PowerPC Tiger inside QEMU on my base M4 Mac Mini and on my Ryzen 9 3900 desktop, definitely not as snappy as running Tiger on actual contemporary PowerPC hardware, but fast enough to actually be productive.

This summer I’ve been investigating agentic coding with local LLMs, and while I’m far from an expert, one thought that has been on my mind is leveraging techniques from “old-school” AI such as heuristic search to guide agents when it comes to planning. The use of Datalog in this article resonates with me, since logic programming was a major part of old-fashioned symbolic AI. I’m very curious about this combination of “old-school” AI and LLMs.

I'd love for Apple to bring back the Mac Pro, but Apple hasn't had an affordable Mac Pro tower since the 2012 model. Apple got rid of the tower in 2013 in favor of the "trash can" model. When the trash can languished for years without updates due to limitations with its design, Apple brought back the tower Mac Pro in 2019, but at a dramatically higher entry MSRP, which priced out many users of tower and trash can Mac Pros (I have a "trash can" and was very disappointed by the 2019 Mac Pro's entry MSRP). The introduction of the Mac Studio revealed a large price gap between similarly-specified Mac Studio and Mac Pro models. Given the lack of expandable RAM in Apple Silicon Mac Pro models, and also given the lack of support for NVIDIA GPUs in macOS, there weren't a lot of customers for Apple Silicon Mac Pros when the Mac Studio was available at a lower price, and thus the Mac Pro got discontinued.

In many ways, the Mac Studio is the spiritual successor to the trash can Mac Pro. Apple Silicon's architecture solved the thermal problems that limited the trash can Mac Pro.

Last week I purchased a MacBook Pro M1 Max from www.macsales.com, which is owned by Other World Computer, a maker of Mac peripherals. My computer comes with a 1-year warranty. It was my first time purchasing from them.

Yup. Yesterday it was $799 for the base Mac Mini M4, a $200 increase from the original MSRP before June’s price hikes.

I paid $499 for my M4 back in April with the academic discount since I predicted the next generation base model would cost more. Sadly my prediction was correct: $599 to $899 is steep. I understand why; the prices of DIMMs, GPUs, and SSDs have also skyrocketed. But $899 for a base Mac Mini is painful, even if the alternatives on the current market are also painfully expensive.

Russ Cox is one of the persons behind Xv6 (https://pdos.csail.mit.edu/6.828/2011/xv6.html), an x86 port of Version 6 Unix, which is the version of Unix described in the famous Lion’s book (https://en.wikipedia.org/wiki/A_Commentary_on_the_UNIX_Opera...). Xv6 was prepared for MIT’s 6.828, an undergraduate operating systems course. In this syllabus from 2018 (https://pdos.csail.mit.edu/6.828/2018/overview.html), students are expected to write significant subsystems in Xv6.

I think many good computer science programs offer a similar course where students make modifications to an open source Unix-like OS like Xv6, Minix, Linux, or one of the BSDs. I took such a course at Cal Poly San Luis Obispo in 2008.

However, I don’t know of any undergraduate programs where students are required to code an entire OS from the boot loader all the way up to basic drivers. Building an OS is part of the final chapter of From NAND to Tetris, so perhaps there might be some universities out there that do require their students to write an OS from scratch as part of the larger curriculum.

I’m not a political conservative and I personally wouldn’t move to Idaho, but Idaho has a lot of appeal. There are plenty of conservatives who used to live in California who moved to Idaho in the past decade due to a combination of California’s high housing prices and wanting to move to a more politically compatible state.

Additionally, Idaho has a lot of beautiful natural scenery.

I’m quite optimistic about the long-term future of local LLMs for privacy and cost control reasons. An LLM running on my own hardware, even if it’s not a laptop but a home server, is one where I don’t need to worry about token limits, token fees, privacy, and “rug-pulling” from the vendor.

In the short term, the big challenge is being able to afford hardware that can run a ~30B model. Last month I got to experiment with LLMs on a NVIDIA RTX 6000 Ada Generation as a visiting researcher during my summer break. I see the power of local LLMs for agentic coding; they’re no Claude, but they are quite useful. I wish I had gotten into local LLMs before hardware has gotten prohibitively expensive and in some cases unavailable; Apple discontinued certain Mac Minis and Mac Studios with high amounts of RAM due to the RAM shortage.

Hopefully high RAM prices don’t become a new normal, though the next year or two doesn’t look good.

I also enjoy my Blackwings, though my normal pencil of choice is the Mitsubishi Hi-Uni HB, which is cheaper than the Blackwing but is still an awesome pencil. I think my supply of pencils will last at least a decade. I take a lot of handwritten notes when doing research, and I prefer pencil over pen when working in computer science and math.

In some areas like the San Francisco Bay Area, schools would have to pay astronomically more in order for teachers to compete against software engineers. I’m a community college professor in the Bay Area making a low six-figure salary, but even a townhouse in the neighborhood my college is in is easily $1 million for a two-bedroom, and rent for a two-bedroom apartment is easily $3500/mo. A modest salary boost can help me rent a two-bedroom apartment, but if I hope to buy closer to work, my salary would need to double, and there’s still the challenge of saving for the down payment, since there’s no such thing as RSUs in a not-for-profit institution.

At least teacher housing provides a way for teachers to obtain secure housing closer to work without having to directly compete against highly-paid corporate workers with RSUs.

I want 2006-era Facebook. No algorithmic feed. No viral posts. No AI slop. Just updates from my friends and acquaintances.

There were exceptions, and I use “were” and not “are” because I don’t know any modern Silicon Valley companies that value loyalty, though I know some people at Apple and Google with long tenures.

Hewlett-Packard under its founders believed in loyalty, and even had a philosophy dubbed The HP Way (https://en.wikipedia.org/wiki/The_HP_Way) that seem like an interesting cross between some of the positive aspects of working for a Japanese company (loyalty, sense of care by the company) and the positives of working in Silicon Valley (flexible work hours, profit sharing), plus some other ideas (Friday afternoons devoted to personal projects).

This philosophy was not commonplace even in Silicon Valley at the time; the founders lamented “hire and fire” companies in Silicon Valley. Eventually HP abandoned the HP Way once HP started hiring outsiders as CEOs instead of hiring from within, beginning with Carly Fiorina.

Maybe I’m looking through rose-colored glasses, but while people need money, there was a time where there was more idealism from the founders and the employees. Steve Jobs could’ve quit business in 1985 after his failed coup; he was rich enough to never have to work again. Yet he started NeXT, and he was able to get people to join his company.

I feel we’re in a timeline in Silicon Valley where there’s little idealism. Employees seek money, and who can blame them when a two-bedroom apartment can easily rent for $4000/mo or more, and when a house is easily $2 million? Leaders seek not only money, but power, and their desires for power have been more blatant in the past year.

Where are the Steve Wozniak and Gary Kildall figures of Silicon Valley these days?

I agree regarding the West Coast in general. Seattle and Los Angeles are also similarly expensive, and even Sacramento (my hometown) isn’t much of an improvement. A few months ago I was floored to see a $13 burrito at a place that was just $8 pre-pandemic. Sacramento is not inundated with rich people like the Bay Area, so what gives? The average Sacramento resident can’t afford those prices.

I went to St. Louis, Missouri on a business trip two years ago and was pleasantly shocked by the food prices there. I felt like I entered a time machine to 2016 California.

Regarding sushi in Japan, prices run the gamut from cheap conveyor-belt sushi chains (like Hama Sushi and Sushiro) and grocery/convenience store sushi (which is surprisingly good here) to ultra-expensive, reservation-only restaurants with sushi served by world-renowned chefs, and everything in between. I’m not an expert on sushi, but I find that $5 grocery store sushi in Japan often beats what costs $20 back in the Bay Area. To be fair, though, the Bay Area has many excellent sushi restaurants, and while it’s more expensive in the Bay Area than Japan for the same quality, the counterargument is a meal at a Bay Area restaurant is much cheaper than a flight to Japan. I just happen to be fortunate enough to go to Japan twice a year, so I just wait until I’m in Japan to have sushi.

I don’t know where you live, but in the Bay Area prices have exploded since COVID. I remember a decade ago when it was still possible to get a super burrito for under $10; in fact, the price was $7.75 for a carne asada super burrito at El Farolito in San Francisco. A burger was under $10, a deli sandwich was about $8-9, and tacos were around $1.50 at Taco Bell and no more than $3 at a proper taqueria. The prices weren’t cheap, but they never felt bank-breaking. Unfortunately prices have skyrocketed since COVID. Unfortunately my income hasn’t matched; I made the individually fulfilling but poor financial decision of pursuing a research career and not working for OpenAI/Anthropic/FAANG/NVIDIA. Thus, two years ago I quit eating out in the Bay Area, except when socializing. It’s sad making six figures and feeling priced out of dining out, just like I’m priced out of housing and RAM since I can’t compete with Big Tech. I now cook, which I guess is healthier but is also time-consuming.

Japan, on the other hand, is heaven for those with US dollars. A bowl of ramen is around ¥1,100 here (I’m here on a research visit), and they generally taste amazing! Even a Big Mac meal from McDonald’s is far cheaper than in the Bay Area.

Additionally, as a former industry researcher myself who now teaches at a community college, there is less research freedom in industry, especially with very free places like 1970-era Bell Labs and Xerox PARC half a century in the rear view mirror. Focused industry research is definitely valuable, but there seems to be fewer places in modern society that allow a researcher to be driven primarily by curiosity.

I’m an noob to Lean myself, but my best explanation is that asserts are checks that happen at runtime. Imagine compile-time asserts, where the compiler is able to determine the validity of assertions without needing to run the program. Formal verification techniques make this possible. In Lean, it is possible to write a proof that, if true, guarantees that the code works.

I’m out of the loop these days in systems research since I largely focus on programming languages and AI these days (though I still love systems) and I treat research more as a side hobby rather than a full-fledged career. It’s disappointing to hear about toxic systems labs. There’s also the “funding-or-perish” and “publish-or-perish” pressures of academia. This is one of the reasons why I teach at a community college, where 8 months of the year I focus on teaching, leaving me 4 months of break per year where I could do research without having to worry about my tenure chances or about funding, though it would be nice to be able to pay some students to help with research projects, and it would also be nice to have the funds to buy expensive equipment such as GPUs with large amounts of RAM.

I may be biased since I interned at Google in 2013 and 2014, but Google in the 2000s and early 2010s felt downright magical as someone who wanted to pursue a career in systems software research. They made impressive technologies that still hold up today, like MapReduce, BigTable, and Spanner. They hired many legends of computer science and software engineering, such as Rob Pike and Jeff Dean.

I’m concerned about the power that Google and other Big Tech companies have, but from a technical point of view Google has a lot of impressive technologies, and from a workplace standpoint, it seemed idyllic back in the early 2010s, though I’ve heard the work culture has changed in the past decade, and I may have rose-colored glasses from only being an intern there, never a full-timer.

I’ve been thinking about this over the past few days. There was an exciting keynote at PLDI 2026 by MIT professor Saman Amarasinghe where he talked about the intersection between compilers and machine learning, including LLMs. One of the works discussed during the keynote was the use of Claude Code by his colleague Martin Rinard to implement a compiler known as a credible compiler that outputs both compiled code and a proof that the output code correctly matches the input.

https://youtu.be/Fc3cW0nqAQ0

I'm a professor at a community college in Silicon Valley, and my students use online textbooks. I try to use Creative Commons or other libre textbooks, but sometimes I use paid textbooks when they are heads-and-shoulders better than their libre alternatives. Some e-textbooks can be accessed on a subscription basis. I admit I prefer non-subscription materials, but a colleague advised me that often the book that students learn from is different from a good reference book that students can use once they've already learned the material. For example, my colleagues and I have had great success with an online, interactive textbook for discrete math. While the subscription is unfortunately only valid for the duration of the course, once students have learned discrete math, they could buy a used copy of Rosen's discrete math textbook as a reference.

The nice thing about e-textbooks is not needing to carry around a bunch of heavy books. I remember the tomes I had in my college days, such as Stewart's Calculus.

I may be naïve or completely uninformed, but given the federal government’s vast resources, including supercomputers, national laboratories, the NSA, and many talented employees, why does the federal government need OpenAI or Anthropic for that matter when it has the resources to build its own LLM, even one exclusive for government use? The federal government has a long history of technical feats, such as the atomic bomb, the ARPANET, and the moon landing. Couldn’t it build its own state-of-the-art LLM?

When you have monopolistic or near-monopolistic control over platforms and other key pieces of software such as office suites, taste is irrelevant since people are still buying licenses out of necessity. Now, Microsoft is capable of writing impressive, tasteful software; I have fond memories of the Windows 95-2000 era, and I also enjoyed Office 97 and Visual Basic 6.0. Even Internet Explorer wasn't bad when it was competing against Netscape Navigator. The problem is Microsoft has a tendency of getting complacent and lazy once acquiring dominance, which breeds bad software.

Apple is more tasteful, IMO, but Apple also has a tendency to abuse its users, taking the attitude, "What are you going to do? Switch to Windows or Android?" Adobe has taken a similar attitude since the Creative Suite became subscription-only. "What are you going to do? The GIMP? Seriously?"

Come to think of it, modern cars have a lot of electronics such as touchscreens, cameras, and sensors. It wouldn’t surprise me if new car prices are not immune to what’s happening with RAM and storage prices.

This pressure works for pure software companies that don’t depend on hardware sales and that have competition. Unfortunately not all software vendors will respond to inflated RAM and SSD prices, since there are many important software vendors who have a vested interest in having users upgrade their hardware frequently. Microsoft still makes a good deal of money on OEM Windows licenses, Apple’s App Store and services revenue is built on regular sales of Apple hardware, and Google benefits from the sale of Android devices. The software needs to perform well enough on new hardware to not cause bad reviews, but sluggish enough (or with enough missing features) to motivate users to upgrade their hardware.

Additionally, software is often chosen based on market effects and not necessarily based on quality. If my colleagues use Zoom, then I need to use Zoom to avoid being difficult. If they use Microsoft Office and take advantage of features that LibreOffice and other competitors can’t support well, then I’m pressured to also use Microsoft Office for compatibility reasons.

The only silver lining I see is that these price hikes will effectively freeze current software requirements in the near future, since purchasing power has been diminished. The MacBook Neo has set 8GB of RAM as the standard for casual users. I’ve found that I don’t have a good time on Windows 11 with 8GB of RAM, but 16GB provides more breathing room and 32GB is great. I don’t expect software companies to revert to the days where they needed to squeeze every kilobyte of RAM like back in the 80s and 90s, but I do expect them to be more mindful of the fact that a lot of people will be using 8GB and 16GB configurations through at least the end of the decade.

The only other event I could remember in the history of Apple that is remotely comparable is the release of the original Power Mac G4 towers in 1999. They were originally going to have 400MHz, 450MHz, and 500MHz models, but due to issues regarding processor availability, Apple lowered the specifications by 50MHz for each model, but without lowering the prices.

https://lowendmac.com/1999/power-mac-g4-yikes/

I have a 350MHz model that I purchased used for $40 back in 2009.

I’ve never seen across-the-board price hikes from Apple that were not accompanied with some type of upgrade.

That’s terrible. I purchased my M4 Mac Mini (base 16/256 model) two months ago because I wanted an ARM Mac for a software project. I feared that the M5 Mac Mini would have a price bump, but I would’ve never guessed that Apple would dramatically hike prices for existing models.

I have some choice words for Sam Altman for destroying the personal computing marketplace by cornering the memory market…