A bug which I don’t know if its related to the new lauch or not, but I experienced it today.
The setup is simple:
Table with 6 canvas columns. In each cavas column I embed a graphic so I can resemble a dashboard with tabs experience.
The bug
When trying to apply table locking for some reason it always switches to the locking of the embedded table (not the main table with only canvas columns. If I click on a new tab (ONLY AFTER the table locking right menu is open) then it switches to the correct table settings. Somehow it switches between table options for the embedded table first rather than the options for the main table (which Options menu I use).
My colleague reported earlier that they are unable to edit content and it was because I accidentally clocked 62 views of an embedded table. Then I tested this several times and the behaviour explained above and visible in the screenrec below (will be available only for short) repeated several times.
Thanks for flagging, @Stefan_Stoyanov — let me bring to the team now
NOT A BUG - BUT A HUGE IMPROVEMENT !!!
My clients and I are seeing a HUGE improvement in the performance of the AI Columns feature of the Superdocs tables.
The model seems to be vastly more intelligent - so things we couldn’t attempt are now working extremely well.
And the performance is VASTLY improved. Inferences that took 10 seconds before are now happening in 1 or 2 seconds!
We have now started bringing in complex prompts and skills files from Claude Code that only worked on their recent models, and they are working very well in Superdocs.
The difference between the side-panel AI and the built-in AI columns is much smaller than ever before. Only the lack of web search and MCP access seems to remain - and we are ok with that. In fact, we would NOT want these inferences to suddenly try to access the web or do any MCP wizardry.
So a big round of applause to the team for all that work.
Well done indeed.
This puts Superdocs back on the top podium for the best platform for building AI Agentic Business Workflows for non-technical managers.
Impressive!
thank you for the feedback Max, we really appreciate it!
we have more updates to AI Columns coming very soon 
URGENT ISSUE
Unable to run ANY AI inferences.
Claims I have used up my BETA credits ???
And there is NO ERROR MESSAGE generated for AI Column inferences.
It just leaves the previous value in place.
THIS IS DEADLY: MY AUTOMATIONS SEEM TO BE WORKING
BUT THEY ARE SILENTLY GIVING STALE RESULTS
This is exactly what I warned about in several calls with the Codans…
NEVER LET MY AUTOMATIONS STOP WORKING DUE TO AI CREDIT
They run overnight and at weekends and we will PAY for extra credits!
Not a good situation
@Max_OBrien Sent you a DM, but let me know here as well if the issue persists!
Enterprise user Analytics were working, stopped on July 9, and have been at 0 for all Sessions by Day and Views by Day
Hey @Kate_Keller!
Bringing this to the team now and will follow up ASAP.
Everything is now working as before.
Thanks to @Ruggy-Joesten and the Docs AI team for resolving this.
Max
Great to hear it, @Max_OBrien!
NOT A BUG - BUT A HUGE IMPROVEMENT !!!
Once my lobster clunkers (Claude Code, Cursor, Antigravity, Copilot, etc) discover they can test out any CFL Formula via the MCP, they go bonkers and try out all kinds of things and spiral-in on a solution brilliantly!
So even though the volume of training data they have on CFL Formulas remains limited, they are suddenly WAY more capable of creating excellent CFL solutions to problems.
They look up the Superhuman documentation to find likely candidates, then test them out using the MCP tool to evaluate a formula, and ‘learn’ from the results.
For example Claude Code was able to figure out the gnarly inner workings of both the ParseJson() and the RegexReplace() functions beyond the (limited) documentation available online.
BOOM! - most impressive!
Well done!
Max
I’m sure I’m not the first to mention this, but in the detail view of a row, select list chips can be difficult to read. The dark background, chip colors, and dark text results in poor contrast.
FWIW, the screenshot below uses the default select list colors for reference.
Playing with the AI helper just now. This isn’t a big deal either way:
I tried to be “helpful” by adding column IDs to the prompt. However, the AI-generated formula used the literal column IDs, instead of the proper field formatting.
The generated formula was correct after reentering the columns.
Hey @al1388Bek!
There should be a fix for this anytime now. I’ll follow up with you again in the morning to see if you are still seeing this.
Thanks, @Ruggy-Joesten. I just checked this morning. The update looks great.
Ruggy, Im sure I was just way late to the party for this one. But I’d prefer to think it’s because you’re on top of it and great at making stuff happen at SH.
Cheers to you and the dev teams at SH.
@al1388Bek When we say that we take community input seriously, we mean it! I appreciate you flagging this and hope it’s a far better experience for you.
I’m not sure when this started happening, but I’ve noticed that my embeds to Google Drive video files no longer work. I’m getting this “Unable to load video” message in the image below.
Has anyone else experienced this? I had them confirmed working before, and it seems across the board they no longer work. Could be something on Google’s side, or perhaps with the Superdocs launch?
