FearNotDaniel

The instant value I got from this is actually something that resembles journalism and editing, even if it’s just an LLM doing the job: the rewriting of headlines along with a half-decent editorial summary paragraph gives me a clear overview of what each link is actually about without the usual puzzling over what the cryptic HN headline is trying to say.

Respectfully, it sounds as if your assumptions about the market for LLM/Agent tools is out of date by a long way. I don’t have any stats but it’s easy to observe that Claude, Gemini, ChatGPT et al are aggressively marketing their products to non developer business users, and pretty much everyone I know who is non technical has heard of at least one of those and uses it regularly in their work.

Great idea, let’s counter the power of a handful of large private organisations by putting that power in the hands of a single large government. What could possibly go wrong?

Just trying to be helpful. I couldn’t see any other way to interpret the question below except as someone who doesn’t know how tool calls get done and who might therefore value a little insight. Didn’t mean to be patronising, apologies if I misinterpreted your question.

> So if I understand it correctly, the tool calls are not actually done by the model?

Correct, here's an implementation example using the Claude SDK [0] but it's the same approach on other platforms: you tell the model what tools it has available, it sends back a response saying it wants to use a specific tool with given parameters, your code executes that tool then sends the output back to the model on the next turn.

[0] https://platform.claude.com/docs/en/agents-and-tools/tool-us...

In no sense is that true of a Catholic Mass. Nor, from what I hear, of an Orthodox liturgy though I’m no expert there. It might be true of some Evangelical Sunday worship services but they cannot be called a “Mass”

You store the future alarm as a wall clock time and date plus location, without making any assumptions about what the civil time offset is going to be on that particular date in that location. When the time comes, whatever system has to read your record and raise the alarm will know what the local time is.

For the same reason people say "Greta Gerwig made this film" even though it took literally hundreds, or even thousands, of individuals to bring her or any other director's vision to life. Unless you've got a super tiny magnet to flip the individual bits inside your RAM, there's always been some chain of tools that translates your intent into an actual running system. You can't even hand-write assembly language without having to rely on an assembler to convert that into actual machine code that is executed by your CPU. Writing in any high-level language, like we all did three years ago, relied on an entire stack of tools to convert that into something that was compiled, deployed and executing somewhere. Would you say "Visual Studio built it" rather than "I built it"? Even if you did, Visual Studio, or whatever IDE/compiler/etc you are relying on, is only the top level in that very deep stack. Claude is just another layer on top of that, which converts your intentions and ideas into code that still has to go through all those other layers before it does anything useful for anyone.

There are still hobby blacksmiths in the world. And at least a bunch of people willing to say they appreciate the craft and effort that goes into that. A very small number of people and companies are still willing to pay actual money for that handcrafted work. But the economic circumstances that cause “Smith” to become the post popular surname in English are long gone, and never coming back.

Worth checking out the methodology this guy put together using Jev and Haiku at appropriate points to cut the cost and also speed up a deep research job: https://news.ycombinator.com/item?id=50019056

Yeah I keep getting “bell curve meme” vibes when I read about all these super complicated handrolled ways people have of getting multi agent setups to work… I’m more likely to be at the caveman end of the spectrum than the guru end but things are moving so fast, by the time I’ve heard about or vaguely understood these esoteric techniques I can usually trust that if they are any good then the good folks at Anthropic et al have already built it into the default behaviour of the harness. Reminds me of the obsession people have with wrapping a repository pattern around Entity Framework because it proves how clever they are.

> In flat regions such as the Fenlands, the necessary elevation for long-range sighting was achieved by means of Bilby towers... What these looked like in detail is left to the imagination.

Fortunately these are an easy web search away - link below shows some UK examples, modelled on the US originals.

https://wiki.trigpointing.uk/Bilby_towers

This comment makes 1000 times more sense than the blog post intro itself. I’m not one of those people who thinks LLMs should never be used to write blog posts, however it’s up to you to tell the model what audience and tone of voice you’re aiming for - maybe try explicitly saying this is aimed at external developers who’ve never heard of your product and don’t know what you’re trying to achieve… makes me wonder if somewhere out there is a blogger’s equivalent of Matt Pocock skills?

Reads like classic Opus slop from before 5.5 - the first few paragraphs are absolutely impenetrable to someone who hasn’t read the prompt used to generate it, there’s nothing to introduce the context or background and it’s full of invented jargon summarising concepts that are never actually described.

Preserve context in the lead chat - let the subagents fill up their own contexts then only return the necessary information.

I'm never letting any of these models bake me an apple pie.

Hard disagree, I’m finding Opus 5.5 so much more pleasant to work with in comparison to its predecessor. Much less of the invented jargon that I got from the last version, or Fable 5.1. Still a bit wordy, sure, but you can dial that down in the settings. I’m wondering what I’m doing differently from you, or is it just a Geschmacksache?

One excellent nerd-sniping side-effect of those political reasons is that the good folks of Austria built an entire nuclear power station and then never put it into service [0]. Open for guided tours on Fridays [1].

[0] https://en.wikipedia.org/wiki/Zwentendorf_Nuclear_Power_Plan...

[1] https://windows2008.zwentendorf.com/en/location-npp/

Important to note is that what is being measured here is the ability of the models not of the chat tools themselves, which combine model completions with other tools that the models can call upon. The mainstream labs already know this about models, it's no secret, and in fact training materials from e.g. Anthropic are at pains to point out that users, or analysts designing workflows, have the reponsibility to ensure the correct tools are used and that human verification takes place at appropriate stages depending on the risk/consequences of the task at hand.

Of course a language-completion model with a training cutoff date won't have up-to-date information on tax rules or the ability to carry out correct numerical calculations, but when you combine that with (in Claude terminology) web search and code execution tools invoked by the chat agent, you immediately have much more reliable results.

You do you. Just carry on blacksmithing in your forge. People are always going to need swords right? Any seismic changes to society that change that fact can only be the result of evil, selfish people which surely someone else will put a stop to before you find yourself out of a job. Meanwhile all the other people called Smith also have families to feed...

> I don't care if it's "intelligent", I don't care if it "has a mind"...

Agreed. The AI is useful for the particular tasks that it proves itself useful for. And if it occasionally spits out a claim that it is "genuinely curious" about a piece of research that I will have to do myself because it turned out to be beyond its mechanical capabilities, then it's more productive for me to simply ignore that claim as a statistical anomaly - a mere hallucination - rather than allowing it to burn a ton of extra tokens outputting what may or may not be the current state of the art on theory-of-mind applied to LLMs because I make the mistake of telling it that it might not actually be capable of experiencing emotion.

True, that’s why I didn’t have a single Fable thread running for the whole time. Each phase had its own orchestrator that ran about 4-5h, in Fable or Opus depending on complexity, delegating to sub agents along the way - they had tasks broken down in a granular enough way that each returned in anything from 5 to 30 mins approx, then the lead agent did some validating and status updates before delegating the next step. So the main agent’s cache never went cold. Each phase raised its own PR and had an opportunity for human validation/code review before the next phase kicked off in a new context.

I love the way people romanticise this hippie utopian idea of “The Internet” like it wasn’t originally built by, and for, the US Military

Models are already good enough. Last week I had Fable plan out a project that took approx 4 days end to end with each phase orchestrated by a supervisor agent delegating individual tasks to other agents, coordinating everything and checking status by simple text files in the repo. I didn’t have to tell the agent to do it that way, it just came up with it and set up the infra as part of the planning overview. Great that everyone’s posting their “my secret sauce” cookbooks just to jump on the hype train but it looks like the models have already figured it out for themselves.

One piece of anecdata: after first getting into Claude Code and then discovered the Claude app I briefly got excited about using it in voice chat mode, demonstrated this to my Austrian wife who found it immediately cool then of course asked if she could chat with it in German. On finding the answer was no, she immediately signed up for ChatGPT so she could use the same feature in her native language. I’m sure Anthropic models are decent with German in text chat, but the lack of that one feature is what drove her straight to a competitor.

I mean, technically it was built for "the management of general information about accelerators and experiments at CERN". The fact that it also supports a whole bunch of peoples' passion projects as well as a multi-trillion dollar commercial marketplace is just a bonus...

https://www.w3.org/History/1989/proposal.html

Or for Western Europe there is DCF77 https://en.wikipedia.org/wiki/DCF77

Sorry you're struggling. There are tons of resources out there from people who've been through the same pain and built up techniques to mitigate them. Matt Pocock's YouTube channel is one good starting place, there are many others. One key tip though is - you own the architecture of your application. If your files have become bloated over several rounds of LLM-generated code, that is primarily your responsibility to observe and push back on that, to ensure your repo as a brief but firm AGENTS.md or CLAUDE.md to describe architectural non-negotiables and some kind of /code-review skill that you give to a second agent to review the first agent's work against those standards. Worried about LLM changes introducing regressions? That's why you have a test suite, and you read what the agent changes to ensure the tests themselves are still checking for correct behaviour.