@Wendy Breakstone From what I can tell, there's still an unanswered questions of whether Cowork running via web/mobile will only be limited to what it can do online via web connectors/online info - or if it will also link to and can use your desktop files and Cowork tasks. Because the backend is still separate via where both Chat chats and Cowork sessions are stored locally - though whether it stays that way is yet to be determined.

Someone else had pointed this change out - and I immediately started digging into it, as it directly impacts my cross-LLM archive project that hinges on the desktop app (and the locally stored files!) The "merging" of project folders, yet the inability to move chats/sessions to a folder pulled in from the opposite interface is rather odd. Feel like it's more of a hiccup than intentional though, because why merge them if you cannot move chats & sessions into a project folder. (And I just realised that I haven't tried setting up a NEW project post merge yet - I wonder if a new project would have the structure to allow both in?) What I CAN add here is that for now, your the local file locations have NOT changed. Without getting too technical (partly because I only JUST understand enough of it myself!), the desktop app stores local history files for your Chat chats; it's not just a web interface pulling in the web chat history. Naturally Cowork has always kept its sessions saved locally thus why it requires the Desktop App. With this merge, it's currently only showing in the apps interface (and online when it's rolled out to all of us on lower plans). But unless they plan to FORCE the desktop app to relocate the physical locations of both chats and sessions, they will still stay separate on the backend. Or they may create a third location moving forward... oi! The renaming of "pulled-in" projects is irksome, but again so far on the backend, they remain separate.

One thing I've noted with Sonnet 5 is that all of the "thinking" in between responses (which you can click the down arrow to view and follow along - esp. as my Claude is well-known for disregarding rules that I have to babysit IT!) are now hidden. AND IT'S Not just from the interface either - if you open up the actual transcript of the chat, the entire thinking section is one encoded chunk! Actually thought some other setting had changed, but figured out that it's Sonnet 5 itself doing this. Starting a new chat in 4.8 showed the thinking steps as before.

I live in NZ, so unless they are evening trainings (or early morning), I generally cannot attend live regardless of how interested I am in the topic. And yes, it’d be nice to just attend a masterclass or training without expecting a new pitch at the end. (Part why I’m tired of those who host free live events and then lock replays behind paywalls or memberships.)

@Wendy Breakstone 🤣 except our upside down is stunning - and lacks monsters.

Looks great for those who have gone full LLM or dev tool builds! Though I'm still sticking with my fully fleshed out them with Kadence, blocks, and more... mainly because I have a MASSIVE backend database with ACF and more on my site. BUT - I have used Claude to draft up some basic wireframe layouts for redesigns or to update pages such as moving existing blocks around. Just shows there are still ways for AI to help without giving up full control. (And Kadence's AI is horribly abysmal and don't anticipate it getting any better under new management... )

@Wendy Breakstone Advanced Custom Fields. Creating custom post types (CPTs), custom taxonomies, and much more.

After weeks of on/off attempts, I finally discovered, setup, tested and scheduled a Claude Cowork task that grabs ALL of my Claude Chat chats, summarises them into a markdown file that is then added to my GitHub along with updating the Project's project-brief file. The goal of the project was to be able to be able to "hop" between LLMs and keep the general gist of things/plans going by simply pointing it to the GitHub projects. (Along with keeping. a summarised archive). Set to run weekly, updating only chats without certain prefixes. Tried SO many different things... and blocked by the fact that neither Claude nor ChatGPT have a public Chat API which would require running a prompt or skill in EVERY chat. And of course even Claude would tell me one method would work - then it wouldn't. Finally realised that through the Desktop app, the chats ARE accessible! Now, its onto the latest method for ChatGPT - though I'm still stuck on the individual chat angle without having to paste or deal with the output of a prompt in every chat.

@Wendy Breakstone Oh... it's been fraught with a LOT of struggles (and plenty of swearing at them both!); I'm not quite sure I'd even call it "done" yet. The ChatGPT side is set up but created different hurdles to fix that I'm still tweaking. I have already given some thought into packaging it up into something once I let it run a few weeks to work out the kinks, but it's utterly irrelevant to my site's content so not sure how I'd get it out there. Be mindful that with long chats, it often truncates the memory - especially noticed this in ChatGPT. So unless you scroll the entire way back and copy from there, it might not grab it all. BUT that is also what helps create a more accurate summarisation of the chat. Lastly, I have mine set to not only include a description and summary of the chat, but to also note what worked, what did NOT, any key decisions made, and cite/list any references used. My goal isn't a purely archival, but those points help make the chat useful in another LLM without having to rexplain everything!