My understanding is that the main reason splitting up work is effective is context management.

For instance, if an agent only has to be concerned with one task, its context can be massively reduced. Further, the next agent can just be told the outcome, it also has reduced context load, because it doesn't need to do the inner workings, just know what the result is.

For instance, a security testing agent just needs to review code against a set of security rules, and then list the problems. The next agent then just gets a list of problems to fix, without needing a full history of working it out.

Which, ultimately, is not such a big difference to the reason we split up work for humans, either. Human job specialization is just context management over the course of 30 years.

> Which, ultimately, is not such a big difference to the reason we split up work for humans,

That's mostly for throughput, and context management.

It's context management in that no human knows everything, but that's also throughput in a way because of how human learning works.

I’ve found that task isolation, rather than preserving your current session’s context budget, is where subagents shine.

In other words, when I have a task that specifically should not have project context, then subagents are great. Claude will also summon these “swarms” for the same reason. For example, you can ask it to analyze a specific issue from multiple relevant POVs, and it will create multiple specialized agents.

However, without fail, I’ve found that creating a subagent for a task that requires project context will result in worse outcomes than using “main CC”, because the sub simply doesn’t receive enough context.

So two things.. Yes this helps with context and is a primary reason to break out the sub-agents.

However one of the bigger things is by having a focus on a specific task or a role, you force the LLM to "pay attention" to certain aspects. The models have finite attention and if you ask them to pay attention to "all things".. they just ignore some.

The act of forcing the model to pay attention can be acoomplished in alternative ways (defined process, commitee formation in single prompt, etc.), but defining personas at the sub-agent is one of the most efficient ways to encode a world view and responsibilities, vs explicitly listing them.

What do you think context is, if not 'attention'?

You can create a context that includes info and instructions, but the agent may not pay attention to everything in the context, even if context usage is low.

IMO "Attention" is an abstraction over the result of prompt engineering, the chain reaction of input converging the output (both "thinking" and response).

Context is the information you give the model, attention is what parts it focuses on.

And this is finite in capacity and emergent from the architecture.

So attention is based on a smaller subset of context?

How much do you babysit claude, and how much do you just "let it do its thing"?

I haven't had anything as severe as OP, but I have had minor issues. For instance, claude dropped a "production" database (it was a demo for the hackerspace, I had previously told claude the project was "in development" because it was worried too much about backwards compatibility, so it assumed it could just drop the db). Sometimes a file is dropped, sometimes a git commit is made and pushed without checking etc despite instructions.

I'm building a personal repo with best practices and scripts for running claude safely etc, so I'm always curious about usage patterns.

Even in small companies, its important to discuss what the expectations around AI are. In the absence of any further requirements (i.e. assuming privacy is not a major issue, regulatory issues etc), it can be as simple as clearly saying: "You can use AI, but you are ultimately responsible for what you deliver. It is expected you verify the data, test the code, and otherwise validate the responses."

Something as simple as that gives an expectation, without being overbearing to start with.

Anyone's basic expectation is that your sourcedis a valid one, but like Wikipedia is not a source and neither are LLMs/MMMs.

This is a really important point - my file manager just says "Name" with sorting. So while its not perfectly defined, it doesn't make the promise of saying its alphabetical.

Arr as in pirate.lots of tools that support piracy end with that suffix like sonarr etc

I came from an academic background, and didn't have anyone else in the team with programming experience when I was researching. I found my time as a contributor for the scikit-learn project to be invaluable in learning about the requirements to build not just working code, but more robust reliable code that others can depend on. Having my work reviewed by those with more experience, making recommendations etc, was fantastic.

So in short, see if you can contribute to a well-run open source project with a good community.

There is a risk, however small, that your bonds won't be paid back leading to a large loss.

There's also the very real risk of bigger than expected inflation.

I think one method here is to incorporate your own site into the content as much as possible. For example, if you are a creator, get people to sign up to a newsletter to get the source files. Get people onto your platform/forum/whatever as well as watching through YouTube. Easier said than done, but better than not doing anything.

From there, you also ensure that you have a backup of all your videos. I've talked to people that only had their stuff on YouTube/Facebook/whatever. It is super risky. If you have a backup, and YouTube bans you, you can rehost elsewhere, it won't be as big, but you might still have a business afterwards.

Also something that needs to be noted, you don't need the same original numbers of people in your kingdom to make equivalent money.

When you're making commerce in someone's fief, they will demand tribute as well. In the confines of your own kingdom, all the ad dollars are yours.

Which also means you don't need to chase the same amounts of people to make similar coin, especially if the deals you make with advertisers are between you and the advertiser (not you, the advertiser, and the king of some other fief).

Exactly. You can be huge on Youtube or tiktok and if you convert some of that to direct engagement you are much better able to survive a changing landscape.

Yeah, every YT creator that is serious about their job should have their own website with a copy of the videos, and I find it really curious that this doesn't seem to be much of a thing? At best I'm seeing merchandise webshops. But you'd think these people would be multi-channel and have a website, youtube, all the social medias, etc, and the bigger ones a company to manage them all.

But I suspect that as they get bigger, they enter in exclusivity / no-compete contracts with Youtube, and if they detect the same video hosted elsewhere, they get taken down or something.

This sounds like an opportunity for a product. Apart from eyeballs and familiarity, Youtube does a lot of handholding so that non-technical people van run their own channels. I don't think 90% of youtubers would have any idea how to spin up a website. But I'm sure they'd be happy to pay someone to do it for them (as long as the price was a small fractuon of their ad revenue).

Probably something like "Buy a pass to play"

Sounds good. I’ll know not to ever press that button.

Do you play paid video games now? Do you plan to stop? The idea that this law is going to cause anyone to actually change how they license games is laughable.

How do you know?

I'm a big fan of Grammarly and have been using it, and paying for it, for years.

The advantage is not spell checking. It is grammar and style improvements. It tells you things like "this language is informal", or "this is a better word for that".