sethhochberg

The key difference is that while an encyclopedia holds facts themselves, LLMs trained on that source material encode something more like a highly probable facsimile of those facts - the original fact was lost, LLMs are lossy, but can often be generated again with a decent level of accuracy by churning through stats about words, concepts, and relationships between them.

The whole catch is that they can often be regenerated. But LLMs (on their own, in their parametric memory - which is the result of training) don't have any conception of whether what they've generated is a real reproduction of some training material or whether they've invented something false that seemed probable based on their encoded stats. When the probability produces something contrary to what was in the training material, you get hallucinations.

They're very, very good predictive text models and can be very, very powerful when hooked up to other tools or outside databases. But its fundamentally lossy technology and all the books having been fed in doesn't guarantee all of the knowledge from those books can be spat back out.

People tend to mix the terms around a lot, but generally you'll see the manufacturers use "ducted split system" or "mini split", where mini splits are the self-contained wall or ceiling mounted units.

I live in an NYC apartment with a similar Mitsubishi ducted split system - very short run ductwork just because the apartment itself is small, but takes up way less space than traditional American-style air handlers, even the apartment sized ones. The entire air handler is suspended above my bathroom ceiling, in between the drywall and the slab above.

My system was originally installed with the 2-wire serial wall thermostat but I converted it to one of Mitsubishi's Redlink wireless units because I needed to relocate the thermostat. But since Redlink isn't wifi, and the Mitsubishi networked thermostats and cloud platform get pretty poor reviews... this project is super interesting to me.

I think there's an element of this which really breaks down to the type system being the part of formal verification that we've figured out how to do during the course of implementation.

Software engineers (myself included, over the years) often argue their real value isn't just writing code, its figuring out the gaps in requirements and how to resolve them. Sometimes that engineering process gets turned back into a formal spec. But much more often, the implementation functionally becomes the spec and contains many details that were never present in the original statement of the requirements.

Formal verification techniques in general are a harder sell until we get the industry to a point where there's broader agreement that what we call "implementation" is often a blurry mix of spec development, prototyping, and actual implementation all happening at the same time.

The details people gloss over when throwing SOC 2 or whatever other audit costs around are the complexity of the system being audited, the chosen criteria to audit (AICPA defines 5 families of criteria... Security is one, but you can optionally add Processing Integrity, Confidentiality, etc etc) and the reputation of the auditor.

A security-only audit for a small company with a narrow product focus can indeed be very inexpensive. A full SOC 2 examination for a large organization with a mix of legacy and modern systems by a name-recognizable public accounting firm can be many hundreds of thousands of dollars, or more if you need a Big 4 firm.

In my opinion, there's not much value to the "cheap" audits... If you're doing enterprise sales to a certain kind of client, your partners who demand an audit are going to want a reputable auditor or they're just going to put you through their own in-depth procurement due diligence regardless. The segment of the industry where a SOC 2 attestation is mandatory to participate but where any random auditor will do feels pretty narrow.

The real takeaway for projects and companies should be that someone having historically behaved in a logical and responsible way doesn’t guarantee that they’ll continue to do that for forever.

Good security architecture has circuit breakers, even for people who are generally high-trust.

I’m one of those people. Wore watches regularly through my mid 20s, completely fell out of the habit as I spent more years working from home and my routine around “getting ready for the day” loosened, and the Apple Watch was the thing that got me to put something on my wrist again - until I got sick of the screen and kept wearing watches, but now analogue ones.

The concept of "tool building" is one of the areas my team has spent the most time coaching our less-technical employees on since widespread LLM rollout in our company.

Developers and developer-adjacent, technical people tend to think this way on their own... but every business has dark corners where repetitive, manual things still happen. We're leaning a lot on training and even org-wide LLM instructions to try and let the LLM (by its own assessment) be the vehicle use to codify a process and turn it into some good old-fashioned reviewable, deterministic automation.

Honestly whipping up a lexer/parser and a REPL is one of my favorite ways to learn a new language. You can cover a lot of ground in a "real" language by just doing the frontend implementation of your own made-up language grammar and a little eval loop and its great for learning/teaching because you don't get bogged down in trying to solve some actual problem.

Which is to say: no shame in just settling for that simple C backend!

Whats implied there is that the "human touch" will become a luxury.

I have a hard time seeing a future where retail doesn't bifurcate even further into ultra low margin, happy-path optimized megastores and concierge-style high touch boutiques. Places like Crutchfield that split the difference nicely seem to be a dying breed.

Even in a relatively open organization where conversations and work are public/discoverable by default, there's still a huuuge difference between the level of curiosity required to join a convo happening in the office kitchen while you're waiting for a coffee to brew vs needing to spend your idle time at work discovering places (Slack channels or whatever else) to chime in while hoping you're not a distraction for others.

I'm a pretty staunch defender of remote work for most roles, but outside of the smallest companies where the entire organization is on a single conversational thread, you really do lose the organic peripheral vision that comes with an office environment and deliberate effort is required to try and recreate some of that in your fully-remote org if you want some of the same upside. Even with deliberate effort, I'm not convinced you can match it perfectly.

Corporate software in general is often chosen based on the value returned simply being "good enough" most of the time, because the actual product being purchased is good controls for security, compliance, etc.

A corporate purchaser is buying hundreds to thousands of Claude seats and doesn't care very much about percieved fluctuations in the model performance from release to release, they're invested in ties into their SSO and SIEM and every other internal system and have trained their employees and there's substantial cost to switching even in a rapidly moving industry.

Consumer end-users are much less loyal, by comparison.

Manufacturers themselves generally don't want to sell directly to consumers: consumers are fickle and need support and have questions and sometimes want refunds or returns and if you sell directly to them, you need to have the staff and policies to deal with all of that. They're also located all over the world, and you might not want to deal with figuring out taxes and duties etc for shipping your product around and figuring out your warranty obligations everywhere you want to sell.

Much easier if you can sell wholesale (sometimes via distributors) to a retailer or network of retailers, and the retailer is responsible for owning the customer relationship, dealing with their part of import/export, local regulations, etc. Retailers are businesses who will buy hundreds of your product at a time, can accept it as palletized freight, and pay you via bank EFTs instead of credit cards.

There are notable exceptions to this model like Amazon's FBA system, but they're the outliers. I'm sure we can all point to inefficiencies in legacy product distribution networks but they solve some real problems.

I think its at least as much of a working environment preference.

Once I became experienced enough to have opinions about things like my editor and terminal emulator... suddenly the Visual Studio environment wasn't nearly as appealing. The Unix philosophy of things being just text than you can just edit in the editor you're already using made much more sense to me than digging through nested submenus to change configuration.

I certainly respect the unmatched Win32 backwards/forwards compatibility story. But as a developer in my younger years, particularly pre-WSL, I could get more modern tools that were less coupled to my OS or language choice, more money, and company culture that was more relevant to my in my 20s jumping into Ruby/Rails development than the Windows development ecosystem despite the things it does really well.

Or to say differently: it wasn't the stability of the API that made Windows development seem boring. It was the kind of companies that did it, the rest of the surrounding ecosystem of tools they did it with, and the way they paid for doing it. (But even when I was actually writing code full time some corners of the JS ecosystem seemed to lean too hard into the wild west mentality. Still do, I suspect, just now its Typescript in support of AI).

High-protein everything is riding the wave of GLP-1 popularity right now. Doctors are begging people on that class of drugs to chase protein targets more similar to what might have previously been reserved for heavy weightlifters just to prevent muscle wasting.

As a result, the entire packaged food industry is pumping up protein numbers and marketing it as the primary attribute of the food (where they might have previously marketed low fat or low sugar or whatever else in the past).

So, saturated market... but certainly one people are investing in now.

I don't have many regrets about having spent my career in (relatively) tiny companies by comparison, but it sure does sound fun to be on the other side for this kind of thing - the scale where micro-optimizations have macro impact.

In startups I've put more effort into squeezing blood from a stone for far less change; even if the change was proportionally more significant to the business. Sometimes it would be neat to say "something I did saved $X million dollars or saved Y kWh of energy" or whatever.

My debit card is a direct line to my primary bank account. If something goes wrong there and an attacker gains access, my cash is simply gone. Yes, the bank will perform an investigation and yes they may issue some provisional credits as a bridge, but there's a window of time between the theft and that investigation concluding where my actual cash is not in my account.

With a credit card, if the card is compromised, its not my money being stolen - its the card issuer's money from my line of credit, and they were planning on settling up with me when my monthly statement closes. I still have to launch a fraud case with the issuer, but critically, _all of my money is still in my bank account_ and I can continue to pay my other bills and obligations as normal.

I think its reasonable to consider giving up that buffer to be additional risk for the debit card approach, setting aside any other advantages or disadvantages between the two.

I've personally had a decent amount of luck with trying to reframe this sort of sentiment from "being useful" to "having purpose".

Right now, yes, its true that a lot of my day to day purpose is driven by participating in the economy and setting myself up for the life I'd like to have in my later years, and I get genuine validation from solving problems and collaborating with people in my day job.

But sometimes, my purpose is to go snowboarding and forget about work. Or to help a friend fix their bicycle. Or to get lost in conversation with a new person I'm dating. As far as any of us know, we only get one turn to be alive on this rock, so we might as well purposefully enjoy it as much as we try to purposefully be useful.

If you look at Ginny Oliver from the article, it might be fair to question whether she was as useful on a lobster boat at 105 as she might have been in her youth. But I doubt she was concerned with usefulness since she had such sense of purpose.

Its less about torrents being the delivery mechanism and more about bringing data from a potentially unknown source, under potentially unknown licensing, and distributed for a potentially unknown reason into the corporate computing environment.

Torrents would be a perfectly valid way for Google to distribute this dataset, but the key difference would be that Google is providing it for this purpose and presumably didn't do anything underhanded to collect or generate it, and tells you explicitly how you're allowed to use it via the license.

That sort of legal and compliance homework is good practice for any business to some extent (don't use random p2p discoveries for sensitive business purposes), but is probably critical to remain employed in the sorts of giant enterprises where an internal security engineer needs to build a compelling case for spending money to upgrade an outdated protocol.

The thing about trademarks is that, if you want to prevent other people from using them, you generally have to still be using it yourself and be able/willing to justify to a court that you're still using it. (At least in most legal systems that I'm familiar with)

Since the original company both changed names and was subsequently liquidated in bankruptcy nearly 20 years ago... that seems unlikely. There's only so many names out there, and occasionally they get fairly recycled.

I have no insider knowledge here but it doesn't seem outlandish to think that the negotiations would go a little differently for an established product vs a brand new one. Goldman may have simply been the only bank willing to work with Apple when the customer base (in size, demographics, spending patterns, whatever) was hypothetical.

Game mode being latency-optimized really is the saving grace in a market segment where the big brands try to keep hardware cost as cheap as possible. Sure, you _could_ have a game mode that does all of the fancy processing closer to real-time, but now you can't use a bargain-basement CPU.

I think there's some real sample bias in that definition of "the community" though, because people who are passionate Ruby programmers giving conference talks, running meetups, etc are often a distinctly different group than the regular-old programmers making business software go 'round every day. The big players writing tools for bringing various flavors of type safety into Ruby are doing it because they're experiencing the pain of having lots of programmers working on large, complex software over years-long periods with the tools that Ruby gives you out of the box. They often employ some of those community fixtures, but thats not the majority of an engineering organization.

The reality is that there certainly are enthusiast programmers who can thrive with the lightweight elegance of stock Ruby, but most people writing code professionally aren't enthusiast programmers under ideal conditions. Everything is always a little more distracted, a little less well-defined, and a little more coupled to legacy than anyone would want. And those are the conditions where I want my tools working as hard as possible, automatically, for me / my teams.

Shorter lifetimes means more renewal events, which means more individual occasions in which LE (or whatever other cert authority) simply must be available before sites start falling off the internet for lack of ability to renew in time.

We're not quite there yet, but the logical progression of shorter and shorter certificate lifetimes to obviate the problems related to revocation lists would suggest that we eventually end up in a place where the major ACME CAs join the list of heavily-centralized companies which are dependencies of "the internet", alongside AWS, Cloudflare, and friends. With cert lifetimes measured in years or months, the CA can have a bad day and as long as you didn't wait until the last possible minute to renew, you're unimpacted. With cert lifetimes trending towards days or less, now your CA really does need institutionally important levels of high availability.

Its less that LE becomes more of a single point of failure than it is that the concept of ACME CAs in general join the list of critically available things required to keep a site online.

"Internal" is a blurry boundary, though - you pick integer sequence numbers and then years on an API gets bolted on to your purely internal database and now your system is vulnerable to enumeration attacks. Does a vendor system where you reference some of your internal data count as "internal"? Is UID 1 the system user that was originally used to provision the system? Better try and attack that one specifically... the list goes on.

UUIDs or other similarly randomized IDs are useful because they don't include any ordering information or imply anything about significance, which is a very safe default despite the performance hits.

There certainly are reasons to avoid them and the article we're commenting on names some good ones, at scale. But I'd argue that if you have those problems you likely have the resources and experience to mitigate the risks, and that true randomly-derived IDs are a safer default for most new systems if you don't have one of the very specific reasons to avoid them.

There are examples of the warehouse-based model working, but they clearly require both density _and_ mindshare. Its not clear Kroger had either based on the other comments in here. FreshDirect in NYC has been operating since the early 2000s with a fleet of tiny trucks with a couple of employees in them and a giant FC with essentially zero retail footprint.

(As an aside, they also have some of the best meat and produce you can get in the city without going to a farmers market. So many retail grocery stores here lack loading docks, the food handling getting from the truck to the sidewalk to the basement of the store to the shelves is really, really rough especially during the summer months. Skipping that and going warehouse-to-home has advantages)

Its hard to have good enough requirements gathering and documentation and product design practices to let an engineer really wrap their head around a problem well enough to come up with and then consistently follow a thoughtful, long-term-maintainable design for a system during implementation.

And its even harder to make sure everyone who reviews or tests that code has a similar level of understanding about the problem the system is trying to solve to review code or test for fitness for purpose, and challenge/validate the design choices made.

And its perhaps hardest of all to have an org-wide planning or roadmap process that can be tolerant of that well-informed peer reviewer or tester actually pushing back in a meaningful way and "delaying" work.

Thats not to say that this level of shared understanding in a team isn't possible or isn't worth pursuing: but it IS a hard thing to do and a relatively small number of engineering organizations pull it off consistently. Some view it as an unacceptable level of overhead and don't even try. But most, in my experience, hope that enough of the right things happen on enough of the right projects to keep the whole mess afloat.

We're just using Github Copilot as our primary entrypoint for all of the model families. Its the only way we can easily offer our devs some level of Claude, Gemini, and Codex all in one place.

I always find the idea that there's something to navigate kind of curious - as you say, its lots of managed versions of open source tools and a mix of proprietary management frameworks on top. Some of what they offer are genuinely unique products for niche use cases, but if you have that niche you probably know what services can support it, like the people in the other comments here mentioning the IoT APIs.

But me (or my teams) are rarely asking the question of "how should I run my service on AWS" in general, its much more typically "I need a managed Postgres database, what AWS product offers that" or "I have an OCI image, what managed platform can I run that in" or even "I want this endpoint to be available all the time, but its usage is very unpredictable/intermittent, so I don't want to pay for idle compute". There might still be a couple of possible answers for those questions, but by the point I arrive there I'm solving for a specific problem.

Its sort of like walking into a kitchen hungry and seeing 3 knives and a stove and oven and a dozen peelers and can openers etc etc and being very overwhelmed by all of this (do I need the knife with a smooth edge or the serrated one?) until you decide you want to eat a grilled cheese, and then grabbing a skillet to put onto a burner and everything making sense once you actually start to cook a specific thing.

They're quite popular for distributed audio systems in general (of which sound masking is one type). "Constant voltage audio" comes in a few flavors and 70v is very common in the US, other parts of the world often use 100v. Background music systems in retail, voice paging systems, etc use constant voltage hardware because its much better technology for very long cable runs, daisy-chained speakers, and centrally located amplifiers.

The cost is fidelity. Full-range audio transformers aren't cheap, so these systems usually make some compromises because your announcements or smooth jazz over the pasta aisle don't need to be true hi-fi.

Its cool technology. Most of the speakers have variable power taps, so you can run a bunch of them in parallel on a single line and control the actual volume as-needed based on where the speaker is deployed by varying the transformer tap on each speaker.

I know this is written to be tounge-in-cheek, but its really almost the exact same problem playing out on both sides.

LLMs hallucinate because training on source material is a lossy process and bigger, heavier LLM-integrated systems that can research and cite primary sources are slow and expensive so few people use those techniques by default. Lowest time to a good enough response is the primary metric.

Journalists oversimplify and fail to ask followup questions because while they can research and cite primary sources, its slow and expensive in an infinitesimally short news cycle so nobody does that by default. Whoever publishes something that someone will click on first gets the ad impressions so thats the primary metric.

In either case, we've got pretty decent tools and techniques for better accuracy and education - whether via humans or LLMs and co - but most people, most of the time don't value them.