The harness that connects to a chatbot, API or voice interaction is the place to route requests to different systems. If you remember the early days of ChatGPT it explicitly said it was routing image generation to Dall-E after embellishing your request itself first.

Determining which tool to use should be a lightweight operation but I’m not expert enough to understand exactly how much lighter than a full LLM call just to recognize it needs a different tool or model.

For sure, I just know it’s tempting given the power of large transformers to throw things at an existing model.

For instance, OCR is something that can be done locally with no access to a GPU but people (including me) still often use cloud hosted multi-modal large language models for it.

Woah dude I took a nap and woke up in 1995.

Heh. Link is wrong but here’s the text:

Then the Lord said to Joshua, “Do not be afraid; do not be discouraged. Take the whole army with you, and go up and attack Ai. For I have delivered into your hands the king of Ai, his people, his city and his land.

Not to be this way but can you not go to a local jeweler and buy a ring with a lab-grown diamond in it, no assembly required?

My local jewelers don't carry lab grown diamonds. So you might be able to, you might not.

The jeweler might have a smaller selection and steeper, retail prices. Many jewelers don't deal in diamonds at all.

I mean, you're right, just ask them to set a lab-grown rock. You're not bringing in ingots of gold for the ring, after all.

Then again, even synthetic diamonds can get expensive so maybe any savings is good savings.

AI coughed up a bunch of references that put their stockpile estimate at around a year. Might have been more at one time but nothing close to 100.

Everything you said is why I don’t agree with the idea that “everyone will be writing their own apps.”

Can’t wait for a world full of apps as janky as most spreadsheets (or Access or FileMaker databases)

In a way if people can’t yet use spreadsheets correctly (but they think they do) then it’s impossible to move them away.

Also the OP has much about the programmatic components, but Excel is also an interface. The front end of an unofficial corporate app. And people know (they think) how to use it.

I just listened to a demo of an insurance industry SaaS that isn’t yet public. Goal to move many things out of spreadsheets. Had AI agents built in and all sorts of bells and whistles. But the screens were chaos. I immediately thought it was just too “complicated” to switch a person off the busy/messy system they know - Excel.

My cynical mind think link you but also comes to the opposite conclusion.

Everyone will be writing apps. Almost every app will be as janky as possible and people will just take the jank as the normal thing and get used to it and working around it, just like they did when every company started using electron and fake mobile apps that are wrapped buggy webpages.

With fast updates that changes every feature across updates, breaking everything and people accepting as the norm.

It's already happening:

- you can see Show HN post making it to the front page full vibecoded with as much jank as possible and the all the comments praising it and not questionning all the jank.

- I've seen it in some popular open source projects when after updates existing feature started to break constantly. The whole thing is ai coded up to the handling of bug reports and feature request and have Claude as co-commiter. I've looked in the internet to see the perception no one is complaining about it

- same at work, managers accepting slop AI solutions that are wrong at so many levels simply because it's fast and complain about the symptoms months down the line without questionning the AI part.

> Can’t wait for a world full of apps as janky as most spreadsheets (or Access or FileMaker databases)

This was already the case even before generative AI

In fact, everyone will ask their llm to write their own app, and it better be correct…

What a cool company. They’ve got the coolest nerdiest things.

I’m loving the non-transitive dice. My son’s birthday is in (checks calendar) ten months, but he’s getting these dice now anyway. So cool.

> Worth stating because four documents now assert

I got one too many chunks of this nonsense and told Claude to knock it off, forever. It acknowledged and wrote out some instructions to its memory about it.

And what a breath of fresh air. Its responses are maybe 20% longer but I read them at least twice as fast. Should have done it a long time ago.

I feel like mine is mocking me. I added an instruction in Claude.md that says "under no circumstances use the phrase found the smoking gun, say I found the problem instead"

What does it do? It says "found the smoking gun! Ooops I wasn't meant to say that - I found the problem!"

It's pretty wild how "reasoning" models now generate like 10 thousand hidden chain of thought tokens in response to a "increase opacity of the logo by 20%" prompt before writing the actual message and yet they still manage to do this.

Why are you using an LLM for "increase opacity of the logo by 20%"? That sounds like the type of straightforward operation a dedicated tool exists for.

any specifics on what you did?

Not the person you're asking, but I did that by explaining to Fable my problem with Opus's gobbledygook and having it write a Claude skill for producing clear explanations in its reports to me. I also had it add notes about the need for clearer writing to CLAUDE.md and other project documentation. Opus's subsequent reports to me have been much clearer.

Here’s an example of one of those Claude skills, in a public repository I manage:

https://github.com/tkgally/je-dict-1/blob/main/.claude/skill...

Fable wrote it specifically for this project.

And then both Opus and Fable will happily ignore these random markdown files (happens to me all the time)

What the American government funds with tax dollars should be chosen by the American people, via Congress. Choosing not to fund speech isn’t the same as prohibiting it.

That said, “chosen by the American people” needs to be broad enough that we’re not singling out individual universities and individual countries based on the power of their political lobbies.

> Their heuristics is opportunity of action

My favorite burglar fact is robberies go way down when it’s raining. Thieves just don’t want the hassle of getting wet.

Business idea: start your sprinklers on burglar detection, not your alarms.