sznio

i assume openai is trying to beat anthropic at any cost, and made a training regiment that makes agents manic

you've never walked under a high voltage power line? even driving a car under it you can hear the radio being weird. i assume it can make life miserable for a ham radio enthusiast. also, walking beneath one, you can hear the crackle, which is why nobody places a power line above homes.

as for them causing leukemia i'd not consider that realistic

as a gen z - same thing

didn't know of that. I'll switch it to learning mode at work, wondering how long it's going to last.

it's more like that the IPv6 switchover was so fumbled that we went from fast P2P like with Skype, to shitty, centralized and data-mined Discord.

The internet would be much less centralized if IPv6 happened when it was supposed to.

and if these elites are better aligned with you, siding with them is better than doing nothing

and now you can use AI to create even more unnecessary features even quicker.

i think that having teams for each product is an antipattern. if the team was purely a "mail task force", the workers could be placed to work on Exchange or the Azure related bullshit. But now, the Outlook team has to constantly create unnecessary work for itself.

I think it really could. You can vibe-code efficient software, if you care.

Microsoft's problems are organizational. A developer can't actually do shit correctly when constantly being pushed to deliver more.

hard to sell something people can't have though

Making a website should be the easiest fucking thing in the world by now. There should be a program that you can use to spit out a website as easy as Word spits out words on paper

squarespace?

and we - us fucking foss nerds or whoever - should've made it.

wordpress?

except in ipv4 getting a link-local address means "I fucked up DHCP" and isn't really meant to be a feature it didn't really work in ipv4 land, and as per the OP, doesn't work in ipv6 land too. Just give everything a proper address and leave link-local to mdns or whatever it was meant to support

I use qwen3.5 locally and it often outputs a lot of claude-isms. Claude actually stopped telling me that I'm absolutely right with every message, but Qwen still does.

I think that the convergence of these tics is just a symptom of distillation.

But why the fuck should any company pay for an ad when the user explicitly searched for their product?

If you don't buy out the top-place for the query, someone else will. In a better case a competitor - imagine Facebook Marketplace being the top result for "ebay". In the worst - a phishing site. The latter happens quite often for software searches.

I don't think it would result in anyone getting nuked. It would result in an end to bombings on Iran.

I've been thinking recently of creating a programming language where you write mostly python, but can just "hand wave" away the boring stuff for the agent to do. If you don't want to deal with it, just type in a prompt or pseudocode and it will get filled in. Kinda like using the ai-assisted image editing software.

the main difference being that you don't switch between an agent chat window and the code. Just leave a note to the agent and go back to coding as usual, while the agent fills in the gap.

Iran should have a nuke. Might finally bring some peace.

all that matters is next quarter.

Funniest thing is how this compares to web browsers impersonating each other in the user agent string. Should Mozilla sue everyone?

I just end up sleeping every time I try that.

After using ThinkPads for years I decided to get a MacBook Pro just to see what's up with them. I went on a 2-week vacation so I didn't even touch my work ThinkPad for this time.

After I returned I was immediately annoyed by the shit hinge, but pleased by the keyboard. ThinkPad keyboards are so nicer than MacBooks. Plastic is just nicer to the touch than metal. If Hackintosh was still possible I'd want a ThinkPad with an upgraded screen running MacOS

For Jira I think the limitation is buried all the backend execution and rendering. It's still fast if you go only by the frontend user-executed part.

what kind of metric should we use to filter out that bullshit?

dollarhours?

I want a in-browser Gemini version. For some reason my company doesn't count Gemini CLI use. I guess I'm supposed to copy code between my browser and my editor.

what I find worst is that not all of this seems to be a work of a nation-state actor.

it seems like quite a lot of these are just made by people in 3rd world countries to make ad revenue. making videos for westerners gives you the richest audience, so the best ad revenue. and anger creates engagement, so making polarizing content gives you most reach.

I guess this model doesn't maximize engagement

beyond the general idea that we shouldn't normalize gambling, betting on some real-life events is horrid. think about insider trading on a polymarket bet for someone's death.

amazing, took me 5 clicks of the back button to finally get back from that link

i think that a wasteful but good solution would be to tag each token, not use opening/closing tags.

whatever n-dimensional space the tokens occupy, manually add more dimensions, to reflect user/agent, trusted/untrusted input.

it should be much harder for the LLM to fuck up this way if every single word it reads screams "suspicion" or "trust". with tag tokens at the start it can just forget

Wtf has happened?? How can they mess up a fucking Notepad App?

they made it use Electron

I had to fall back to that to deliver anything recently - but the last two months were really comfy with me just saying "do x" and just going on a walk and coming back to a working project.

Claude is still useful now, but it feels more like a replacement for bashing on a keyboard, rather than a thinking machine now.