Zambyte

Do you think there will be no consolidation in the pop?

Training is expensive. That seems like an important incentive.

"AGI" is an expression on how widely the intelligence can be applied, not how effective that intelligence is. I think it's hard to argue that current (as of the last few years) intelligence tools cannot be applied generally. What can be easily argued is that we have not achieved superintelligence (where AI is better, faster, and cheaper than any human at all intelligence related tasks).

Most do. The mentioned rootkits don't.

I'm not reading aloud most of the time that I'm talking.

You put 'published' in quotes because you thought it was a weird word to use for radio broadcasts. Did you maybe stop and consider that I wasn't talking about radio broadcasts?

It's actually okay for people to have their feelings hurt about seeing people literally explode and listening to their final moments as they know what's about to happen.

It's being published in the sense that news outlets are redistributing it. They are under no obligation to redistribute it. They are doing it because a) they believe it will drive more traffic to their site, and b) they don't care how disrespectful it is to redistribute.

I don't use AI for my mind map, but if you're actually trying to map your mind, I recommend against using a frontier model. I recommend seeding the mind map by hand using open formats (ie: something markdown or org based) and then slowly introducing AI tools to automate the high friction mechanical processes you bump into.

The reason I recommend against a frontier model is that any halfway decent model (like, single card consumer GPU level models) can automate the mechanical processes of this kind of tasking.

It's amazing that it seems like only the people on board lost their lives. The crash looked devastating, and into a fairly high population (residential, on a holiday weekend) area.

It also seems very disrespectful to me how the ATC audio has been published. I haven't listened because of how many people I've seen saying they found it haunting or regret listening. Could you imagine that circulating if the pilot was your family? If that was how you found out? I get that all the news outlets want to be the one to report and capture attention, but these are real people. My heart goes out to the families.

I haven't really done Common Lisp, but I have played around with Scheme. The big problem that I run into is the fact that ) and )) (and probably more) are both a single token, and thus it's tricky for models to balance parentheses correctly. I've been curious on exploring possible tool setups in my harness to make it easier, but I'm not sure what exactly would be best yet.

When smartphones became popular, a lot of social functions developed under the assumption that all participants would have a smartphone, and have it with them at all times, connected. There has been no need to consider the idea that someone might want to participate without one.

You're arguing that car dependency is fine (for you). I'm arguing that car dependency is a good analogy for phone dependency.

Depending on your application. If you're not pushing the model to its limits, 5% is not very much.

Are you imagining a modern society that developed without cars as being more of a burden to live in, or are you imagining the current society, developed with cars, as being more of a burden to live in when just getting rid of cars? Because yeah, getting rid of a car is going to make most people's lives harder. That's kind of what makes it a dependency.

Herdr is a harness aware terminal multiplexer, not a harness itself.

Car dependency is an excellent parallel. People buy cars on the promise that they should expect freedom of movement, but then they arrange their lives so they can only go where roads have been paved, and arrive where buildings have been demolished for parking lots. They become financially dependent on oil extraction in the Middle East. They become incapable of getting anywhere alone when they lose the ability to drive, due to ailments or age.

So why do we do it? Because we can't imagine living without the convenience of our phones.

Or have a tablet. That was my main reason for using Molly.

There is a very interesting divide in how people use AI to create things. The prevailing narrative focuses on a (large, to be fair) group of people who essentially use AI as a megaphone for their lack of care. It can be used by people as a megaphone to amplify the existence of care as well though. If you care deeply about making something, such as a tool to solve a problem or to make people happy, you can do really great things that people will appreciate. If you simply pipeline prompts from your head, to AI, to other people, without iterating on the output of the AI, you will annoy people with how loudly you don't care. That's slop, no matter how good AI gets.

I don't think the person you responded to has anything to do with ao3. They quoted the content of the update on the linked status page.

I had to check the spelling - they spelled "loser" correctly at least. Would have been a great irony if they didn't.

I have absolutely no insight into their financials, but yeah, their incentives point entirely to saying the funds are safe, regardless of if they actually are or not. Either they are indeed safe, and it eases panic, or Binance folds, and lying didn't matter. It's do or die for them right now.

> Codex, Claude etc already do subagent management on their own

What about together? Can you reasonably do agent to agent interactions across harnesses? herdr makes that easy.

One of my favorite reasons for using herdr is what I'll call "everything is a pane". Running a new agent? Runs in a herdr pane. sshing into a system? Runs in a herdr pane. Anything that runs in a herdr pane, your agents can interact with. I will often do things like ssh into a system in one pane, and then ask pi in another pane to help me do diagnostics on that system. It never has to try to ssh in directly in the harness with my credentials, it just sends keyboard input to the pane, and reads the content of the pane. I can watch the interactive session as it goes by.

Another workflow that I have sometimes is I'll kick off some work in an agent, and instead of making a new sub agent, I'll actually make a new parent agent if I get worried that it's going down an unnecessary rabbit hole. I'll tell the new agent something like "periodically watch the agent in pane X, it is working on Y but I am concerned that it is losing track of its tasking. Periodically review what it's doing, and make sure it doesn't go off the rails. If it does seem to be going off the rails, give it guidance to bring it back on track" and that seems to have helped.

Binance is centralized.

Well at some point they were.

"racing cars" actually would not be stationary. "Race cars" are the class of vehicle, which can be stationary. If the cars are racing, they are not stationary.

"High-speed" is ambiguous. It can be used to describe the class of vehicle, or it can be used to describe the state of motion of the vehicle.

In most languages, async is implemented as a special feature in the compiler, not as userspace code. You can't just jump-to-definition of async/await keywords in languages that use them. async/await are not keywords in Zig, they are regular function calls. There is obviously a categorically different level of insight between jumping to library code and digging into compiler internals.

It's all implemented as explicit userspace code that follows the same rules as the rest of the language. There is quite literally no implicit logic.

Exactly, so the title wrongly frames git as trusting certs that it shouldn't, when it's correctly trusting the system certs, and MacOS is wrongly trusting certs.

You along with everyone else was trying it. Hand crafted artisanal code wouldn't have stayed online either. Fluxer works now.

By Genie specifically? Or by generative AI in general?

If you mean world models like Genie,I don't know. If you mean the latter, I have definitely used software that I have enjoyed that I do not believe would have been possible without generative AI.

Fluxer is the first Discord competitor that when you show to someone they can't say "well Discord does X better", and it was largely made using AI.

Herdr is designed for AI usage, but I actually really like it even for just normal terminal multiplexing. It was surely made with a lot of AI.

I have seen people make incredible things at my work that have actually solved multi-year long pain points that we've never had the resources to address before.

I have used AI to patch and revive a bunch of small services like chat bots or web services and such that I had made and had decayed into failure over the years.

Sure, it's easy to point at people using AI as a megaphone for their lack of care. But I don't think it's impossibly hard to find people doing good things with it either.

How does Discord compare to Zulip? Fluxer is functionally equivalent to Discord from a users perspective.

AFAIK, Zulip doesn't have any sort of built in calling, streaming, etc. right? Fluxer has that I'm addition to the Discord style chat features.