kfajdsl

These are like $20-50 subs, you’re probably paying your dev a hell of a lot more. Let them use the tools they want. I spend almost all of my time in Emacs or Cursor, but I still haven’t found a database client that I like better than Datagrip.

Wow, that sucks. Getting Claude for everyone wasn’t even the stupid thing, it was thinking that a shiny new hammer meant you could throw away all your wrenches.

To 2., yes - you just have to look in the right places. You're sure to find them in middle/high school or university robotics teams, for example.

When computing was niche, you really only got into if you had a real interest in it (I assume - I wasn't around back then). That's changed, but it doesn't mean that that category doesn't exist anymore. If anything, it's probably way larger in absolute terms, if a smaller proportion of people who work on software in general.

You don’t need to know what proton is to click ‘install’ in Steam.

Not really? I did something similar for a different device recently. It can make files and has access to bash. It's perfectly capable of installing packages and writing small scripts basically entirely autonomously. No MCP needed.

All the LLM logprob outputs I've seen aren't very well calibrated, at least for transcription tasks - I'm guessing it's similar for OCR type tasks.

I’ll say that, I think this varies by language/SDK - at least with the Temporal TypeScript SDK, a simple single idempotent step background task is however many lines of code to do the actual work in an activity, and then the only boilerplate is like 3 lines of code for a simple workflow function to call the activity.

It's a tool. The first time you used a shell you had to learn it. The first time you used a text editor you had to learn it.

You can learn how to use it, or you can put it down if you think it doesn't bring you any benefit.

It’s one thing for you (yes, you, the user using the tool) to generate code you don’t understand for a side project or one off tool. It’s another thing to expect your code to be upstreamed into a large project and let others take on the maintenance burden, not to mention review code you haven’t even reviewed yourself!

Note: I, myself, am guilty of forking projects, adding some simple feature I need with an LLM quickly because I don’t want to take the time to understand the codebase, and using it personally. I don’t attempt to upstream changes like this and waste maintainers’ time until I actually take the time myself to understand the project, the issue, and the solution.

> The music industry already went through this with AutoTune and we know how that turned out.

Yeah, it turned out that almost all mainstream tracks nowadays have post-processing on vocals (the extent varying between genres and styles).

I understand that you're being glib about buses or trains, but the driver is a large part of the operating costs of a bus, and additionally driverless buses might make more frequent but smaller buses more economical.

For your examples, honestly yeah. A dev should familiar with the basic concepts of their language and tech stack. So yes, they should be able to understand a basic snippet of code without Google, an LSP, or even a computer. They should even be able to "write CTEs with a pencil and paper". I don't expect them to get the syntax perfect, but they should just know the basic tools and concepts enough to have something at least semantically correct. And they certainly should be able to understand the code produced by an AI tool for a take home toy project.

I say this as someone who would definitely be far less productive without Google, LSP, or Claude Code.

Yes, but the federal government uses far more than just Office.

Microsoft is very far from being at risk of failing, but if it did happen, I think it's very likely that the government keeps it alive. How much of a national security risk is it if every Windows (including Windows Server) system stopped getting patches?

fyi your (?) css is messed up for the 'Show activities by...' images on a viewport that's the size of a macbook pro 14" display split in half vertically

Google sheets

Do they? Are there any production cars under 20k? Plenty of used ones for that.

You give the LLM search tools.

You mentioned grep-mode, which to my knowledge is just bringing up a buffer with all the matches for a regex and easily jumping to each point (I use rg.el myself). For the record, this is basically the same thing as VSCode's search tool.

Now, once you have that, to actually make edits, you have to record a macro to apply at each point or just manually do the edit yourself, no? I don't pretend LLMs are perfect, but I certainly think using one is a much better experience for this kind of refactoring than those two options.

Finding and jumping to all the places is usually easy, but non trivial changes often require some understanding of the code beyond just line based regex replace. I could probably spend some time recording a macro that handles all the edge cases, or use some kind of AST based search and replace, but cursor agent does it just fine in the background.

I don't think that's the case in TFA, because then why would recent layoffs matter?

This is my number one complaint with LLM produced code too. The worst thing is when it swallows an error to print its own error message with far less info and no traceback.

In my rules I tell it that try catches are completely banned unless I explicitly ask for one (an okay tradeoff, since usually my error boundaries are pretty wide and I know where I want them). I know the context length is getting too long when it starts ignore that.

Yeah I don't think a business is going to try to force the issue when a geoblock is simple to implement. If it happens, it's probably going to be some kind of advocacy group pushing it.

As far as I understand it (IANAL), this ruling decides that the speech restrictions imposed by the Texas ID verification law are compliant with the 1st amendment. It didn't touch on whether or not Texas can enforce its laws on entities that don't do business in Texas.

If an state AG tries to prosecute an entity that has no ties to the state other than content being passively accessible, that's probably another supreme court case if it doesn't get immediately decided in favor of the defendant in the lower courts. You open a big can of worms if entities are required to proactively comply with regulations in states they have zero presence in.

If Texas wants to block content from entities that have nothing to do with Texas, they can build their own great firewall.

What incentive to sell off the plot does LVT create that doesn't already exist, maybe with a marginally lower degree? I'm guessing the reason they can't sell a tiny weirdly shaped lot is that no one wants it. If they didn't want it and they could sell it, they already have ample financial incentive to sell.

Has there ever been a swing back to small, independent producers of anything after an industry has consolidated?

I mean that you should only give the LLM the same privilege as whoever the source of your untrusted input is has. For the support agent example, it should only be able to access records related to user it's talking to, and only be able to do mutations that the user has permissions to do anyways. Though I've hated all the chatbot support "agents" I've interacted with so far, so please don't actually do this unless you have some secret sauce to make it not horrible.

I agree that for most tasks a pre-defined workflow with task specific LLM calls sprinkled in is a much better fit.

However, I really like agents with tool use for personal use (both programming and otherwise). In that case, the agent is either sandboxed or I approve any tools with the potential to do damage.

For the example of the Supabase MCP, it still seems pretty useful when limited to a test environment or read-only access to prod - it's a dev tool. Since it's a dev tool, the user needs to actually know what its doing. If they have no clue but are still running it on prod data, they have no business touching it or frankly any other dev tool. I class this as the same ignorance that leads people to storing passwords in plaintext.

So, I blame the developer for trying to use an MCP server 1) when they have no idea wtf it does and 2) in an environment that can affect real users who aren't aware of the incompetence of the dev whose service they're using. Likewise, in TFA, I blame the dev, not the tool. Unfortunately, no matter how you do it, lowering the barrier of entry for development while still providing access to ample footguns will always result in cases like this.

How many private parks are there? Pretty much every one I've been to has been government run, other than small outdoor spaces next to private buildings and large pay for admission gardens that are usually way out in the boonies on the grounds of an old plantation or manor.

I can't think of how a private, but still public-access, park survives without a rich benefactor eating the losses, even today.

Attracting investors is hard, but it's still the easiest part of building a company. If the company pivots, then it's pretty likely that the vast majority of the previous work no longer has any (monetary) value.

You can use it with live data if you give it read access to prod and write access only to internal channels (whatever that may be, the point is it doesn’t have the ability to leak data to the outside world).

There are plenty of ways to sandbox things for a particular use case.

LLMs are still incredibly useful under these constraints.