I wonder how many of us had the exact same experience (down to those damn three little dots!)

Mobile Safari has been giving me a complete loop on these in the past couple months, I have to switch browsers to get through. Anyone else?

My guess is that this is a direct response to all the claw stuff running on macs. I used to never get cf captchas from a mac + home IP (while getting plenty on my linux ws + work vpn). Now i've gotten 2 sites in the past week that not only show the captcha, but also loop once I click the human thing. Most likely mac + resIP is not a good signal anymore...

Worked for me just now on mobile safari. You get the cloudflare human test but I just clicked the box and was in. This was despite accessing the site while vpn’d from home and using multiple adblockers.

Maybe it’s the blocking of 3rd party cookies, because I experience similar issues with Chrome on desktop from time to time.

I occasionally get those loop even on chrome.

Is there a way to use Anthropic subscription with hermes-agent?

Also one or more examples in this blog post would have been great to see

Great point; I wasn’t sure if anyone would see this post so I spent most of my time on the docs. I just added a few visual examples to the post. Thanks!

Jesse Genet has been posting some cool use cases of OpenClaw for homeschooling that are somewhat along these lines. Using the assistant to inventory the physical manipulables, the curriculum pages, and how they intersect. Printing pages automatically for certain lessons. Updating e-ink screens with other lessons.

It's been 10 hours and all the links in this comment section still work...

If you use Google AI mode, it'll be happy to find you any kind of PII information.

since the solution here doesnt appear to be open source, I think you can get something similar by asking your agents to take AgentFS and replacing the DB with ChromaDB

If you don't follow anemll, they also have a usable version of OpenClaw running on iPhone.

With hardware and model improvements, the future is bright.

This is really promising. Are they now going to scale this up to hundreds of billions of parameters? Why stop at 1.5B if they found a potentially SOTA architecture?

Probably constrained by training resources. It's much easier to experiment with a smaller architecture. You may need many training runs to figure out hyperparameters for example. If each run needs multiple GPUs for a week the cost adds up quickly. I think it makes a lot of sense to start small.

This just happened up north, apparently the largest proposed AI data center in Canada (Synapse's $10B 1GW campus in Olds, Alberta) was just put on pause after the Utilities Commission rejected its power application on March 6, 2026 due to noise pollution concerns from 20 gas turbines, 10 steam generators, and up to 600 diesel backups near 800 homes (just 200m away). The assessments failed to model cumulative worst-case noise. The proposal will be revised and resubmitted of course but the concern isn't going to go away.

Seriously, why do they have to build a DC 200m from homes? Is Alberta out of space?