etse

I'm not quite understanding how different prompts for different models reduces the attractiveness of a framework. A framework could theoretically have an LLM evals package to run continuous experiments of all prompts against across all models.

Also theoretically, an LLM framework could estimate costs, count tokens, offer a variety of chunking strategies, unify the more sophisticated APIs, like tools or agents–all of which could vary from provider to provider.

Admittedly, this view came just from doing early product explorations, but a framework was helpful for most of the above reasons (I didn't find an evals framework that I liked).

You mentioned not having this problem yet. What kind of problems have you been running across? I'm wondering if I'm missing some other context.

If you wanted to compare OpenAI models against Anthropic or Google, wouldn't the framework help a lot? Breaking APIs is more about bad framework development than frameworks in general.

I think frameworks tend to provide an escape hatch. LlamaIndex comes to mind. It seems to me that by not learning and using an existing framework, you're building your own, which is a calculated tradeoff.

Go's debugger has worked just fine for years. What do you prefer about gdb (which also seems to support Go)?

Go isn't pretty to look at, but I feel so much more confident in a successful Go build than starting up Python–not knowing when exceptions will throw or crash until it's run is frustrating.

I'm sure Rust folks feel exponentially more safety than Go at compile time, but the jump from Python to Go is dramatic for me.

I can imagine. An AI "copilot" could mean quick error correction... or maybe a snowballing of tiny errors into something catastrophic.

Harrison Ford says, "I know," instead of "I love you, too" (or something like that) in Empire Strikes Back's carbonite freezing scene. That's an immensely significant and meaningful update.

Well, without charging for it, right?

Nice results! It reminds me of the Portal song about Aperture Science: https://www.youtube.com/watch?v=Y6ljFaKRTrI

Entrepreneurship in semiconductors is its own world. Perhaps to a lesser degree, pharmaceuticals, but it’s a valid consideration that credentials and background help overcome barriers when needing to raise massive amounts of money to get started.

By "It challenged my thinking and improved it." OP probably meant 'improved upon it', or 'improved on it' where "it" was the originally desired output.

Well put. Unfortunately true!

Git-versioned marketing templates seem unfriendly to presumably a marketing user.

That said, I’ve been integrating and supporting similar platforms over the past several years, and the feature set and maturity is low for the cost of these services. If Dittofeed puts the pressure on the market, that’s fantastic.

Wasn’t this supposed to be Kurian’s prerogative—sell enterprise services? Google product seem fine to me, but I agree that they have no customer care DNA. I was expecting someone from the outside to bring it in.

PowerBook G3 had expansion ports, too. Maybe it was the 1-inch thick titanium powerbook g4 that changed the game.

Thanks for sharing that. I liked your note on doing the trail in your mind because it reminds me of when I stumble through a piano piece that I learned decades ago as a kid. It somehow brings me back in time, so that I feel like I could guess what was cooking for dinner or what was stressing me out at school/home.

Let's view this from a different angle. The money will always be coming in. Since the movie is beloved for non-racist reasons, let's assume the racist and non-racist portions are in competition. Removing scenes that don't affect the beloved aspects of the movie is easy and cheap.

Meaning, it is entirely possible to easily produce a version of the movie that is less racist and still popular. If the only reason you don't is to appeal to the integrity of the work, I argue that the all-or-nothing approach is more about moral posturing than market demands.

People are so convinced of a slippery slope, but I haven’t been able to piece together the plausible evil end game from the comments here. What am I missing?

Stay on the old version, because you can.

How would you compare Swift with Kotlin in terms of being a good stand-in for high-level Rust?

I've not paid attention to Kotlin in a while, but it seems like there's more progress at supporting Kotlin across more platforms (Kotlin native, web?, etc). I'm curious if people feel that the language features and design are at about parity, or if one is significantly stronger/weaker.

What about the modeling sound in particular fails indistinguishable? My ears aren't trained enough.

I agree that machine-generated passwords are better, though it spends half of the article on it, based on this weird statement:

> The LastPass account password “best practices” advice linked to in their announcement says nothing about using a password generator, so it would be incorrect to assume that users are generating their LastPass passwords using a strong password generator.

A more compelling criticism would be to identify some flaw in the UX instead of the absence of something in documentation, which is a really poor signal for user behavior.

Did you even read the article?

Change doesn’t have to come at the pace of npm. Let’s be charitable, people. :)

Alternative to the recursive function question, was also one testing mathematic induction (1999). I didn't know enough CS or Scheme to get into CS61A, but they were fine with getting the underlying concept in a language that was safer to assume people were familiar. (I wasn't lucky enough to have any CS or programming courses in high school unlike most of my classmates.)

I think Apple was motivated to switch to Intel because of the stagnated development of PowerPC. You may be thinking further back, like in the 90s, regarding Macs that could dual boot into PC

Could you elaborate on what you meant about the sports team applying only to key people?

Is it about a difference between operational work and strategic work?

By resources the parent comment seemed to emphasize lack of “community”, and not training.

Toyota Sienna minivan. Having driven only sedans and coupes, and grew up riding in 90s-era SUVs, I was shocked to discover what modern "minivan life" is like... as a parent. For me, my reference for modern is the past two weeks of a mundane 2010 Sienna LE. Every couple days, I notice something new and thoughtful.

The side doors are powered and can be controlled from the remote, since kids won't often open doors for themselves or you.

The last row of seats can fold down flush with the floor, since we alternate between having more cargo space and passenger capacity (like trips with grandparents or for instance, equipment for sports or music).

The car windows come tinted. One less worry/discussion about harsh glare or sunlight on our young kids.

There's a second lighter outlet right beside the first, which is handy for our electricity-dependent lifestyle.

There is a compartment on the driver's side, near your head, that reveals sunglasses.

There are cupholders upon cupholders on every interior panel, which seemed absurd to me, until I saw them used by my own family. And when you've filled them all, pop open the coin tray and a slow reveal unfolds... yet another cupholder, like digging into the hesitation before the punchline of a joke.

Newer minivans cover more use cases, like automatically popping the rear door by waving your leg, or changing the middle row positions to accommodate side-by-side car seats, and I'm curious to see which features have staying power and which don't.

We often applaud elegant solutions to a well-defined problem as a good design, or at least I feel like software engineers tend to, but I have a growing admiration for designs that solve a problem complex enough to resist being defined once-and-for-all. Those problems tend to be "human" problems that are as deep as human psychology and change as our society changes.

I don't think Google Cloud monitoring and alerts only supports Slack besides SMS and email. No sign of Google Chat.

Abide | Backend Software Engineer | Full-time | San Francisco Bay Area, USA | REMOTE friendly

Tech: Go, MongoDB, Kubernetes, GCP, React

Abide is an app for meditation and sleep using the Bible. (Yes, the Bible has been putting people to sleep for millenia.)

We're looking for a senior developer with strong ability to design and implement RESTful APIs and work effectively with our databases and other systems.

Apply at https://apply.workable.com/abide/j/B225C828D9/