I started clicking a 'next page' link before I'd actually finished reading something (so I kept holding the mouse buttown down), and a couple of seconds later Firefox popped up a 'link preview' box informing me that I was clicking on a link to a web forum. Wow, thanks, couldn't have figured that out myself. (It did not actually summarize the next page in any way.)

> Even when the model is explicitly instructed to pause due to insufficient tokens rather than generating an incomplete response

AI models can't do this. At least not with just an instruction, maybe if you're writing some kind of custom 'agentic' setup.

Yeah, it does. It was possible with 2.5 Flash.

Here's a similar result with Qwen Qwen3.5-397B-A17B: https://chat.qwen.ai/s/530becb7-e16b-41ee-8621-af83994599ce?...

Devtools is seemingly partially broken in this version, if I have devtools open on a reasonably dynamic web app Chrome will crash within a minute or two

It's also been ridiculously slow for a month or two now :/ not a good time to be working on some relatively intricate performance optimisation with DevTools taking 1-4 seconds to even start the performance recording.

Why bother supporting Safari when they aren't interested in supporting the modern web? They're five years behind.

In a way I agree with you; but practically 100% of iOS/iPad users are forced to use Safari. Plus, it's nice to have a browser engine that's not Chromium.

If that browser engine is safari, I have to seriously question the validity of that final sentence.

The fact that safari refuses to support modern features and is forced on ios devices makes it even worse.

Archive.today don't use Cloudflare, the admin mimics their captcha page because he hates them. He also used to captcha-loop anyone using Cloudflare's DNS resolver because they don't send the IP subnet of clients to upstreams.

I don't think it's a honeypot, though, it's not like he's learning much about me other than I like not paying for news sites.

The -codex models are only for 'agentic coding', nothing else.

They could simply say that, then, instead of saying you might be able to use it under the AGPL.

They didn't say you "might" be able to use it under the AGPL, but that you "may" be licensed to use it. Which, as a native speaker of American English, seems to be relatively clear in its meaning along the lines of what the GP poster stated. Of course, the various meanings of "may" in English might be subtle enough that I'd readily believe it's less clear to non-native speakers (or maybe even speakers of a different dialect), and it's unfortunate that Mattermost's lawyers aren't interesting in cleaning up the language.

Yes, many. As a random example, see: https://www.servethehome.com/ubiquiti-flex-mini-2-5g-review-...

The last image on the page shows various chips in the switch, the top left is an ESP32.

Interesting, seems like they're just using it as a MCU? Specs don't mention anything wireless, and I don't see an antenna.

As a result of this the official install is now installing a squatted package they don't control: https://github.com/moltbot/moltbot/issues/2760 https://github.com/moltbot/moltbot/issues/2775

But this is basically in line with average LLM agent safety.

It's even worse than I guessed - moltbot updated their official docs to install the new package name ( https://github.com/moltbot/moltbot?tab=readme-ov-file#instal... ), but it was a package name they have not obtained, and a different non-clawdbot 'moltbot' package is there.

It's been 15 hours since that "CRITICAL" issue bug was opened, and moltbot has had dozens of commits ( https://github.com/moltbot/moltbot/commits/main/ ), but not to fix or take down the official install instructions that continue to have people install a 'moltbot' package that is not theirs.

pandoc is big (nearly 200MB binary on Debian) and does far more than just Markdown -> HTML

Wow I never knew pandoc was 200MB! It's only a 26MB download, but 200MB binary is kind of insane...

It is by far the biggest binary I have in `/usr/bin/` (the second being `blender` at 90MB - understandable I guess! - and the third being `stack` at 75MB - haskell again!)