The title of the figure is ambiguous about what "SD" really means but I guess it is plotting the number of standard deviations of the 1991-2020 data measured from the mean of that data and plotted per day for the 1982-2026 data.
That shows measured temperature and 2026 as very hot (surprise) and in fact is the hottest to date for June+July. The standard deviation of that data is not 3.5 C but something less than 1.0 C. It is plausible that current temperature is about 3.5 sigma from the selected mean.
It's worth recognizing that the analysis is applying a biased conclusion prior to making the plot. It singles out post 2020 data to compare to pre-2020 data. and then concluding the held out data is a significant deviation with a cause. It almost certainly is but this is not a proper way to analyze data (unless one is pushing an agenda, be that for good or bad).
I think the sst_daily plot stands on its own without crafting this SD plot to emphasize the point. Especially when the accompanying text doesn't even explain it. It's a disingenuous message.
You're reading this graph wrong: we're currently 3.63 standard deviation above the mean.
It's clearer on the original article[1] that this AI-generated blog is taking the graph from, the average temperature on the period at this time of the year is around 27.5°, the ocean is almost at 29.5°, just short of 2°C above average, and the standard deviation is 0.55°C.
According to one comment on the site, the 3.5 means "3.5 times the SD", which makes much more sense to me.
I initially tried to make sense of "SD being 3.5 on that day of the year", which seems to be a wrong interpretation.