You have a ring of vertices that should be a circle. What you have is an eight-sided approximation of one, extruded out of something else twenty minutes ago, and the bevel you are about to run across it will show every one of those eight flat spots. So you select the ring, hit To Circle, and move on.
That operator has lived in an add-on called LoopTools for about fifteen years. Bart Crouch wrote it. It ships inside Blender and always has, sitting in Preferences with its checkbox off until you go and tick it, which is the kind of detail that sounds trivial until you are the person on a support call asking somebody why their right-click menu is missing half the entries it should have.
As of Blender 5.3 it is not an add-on any more. The operators are being rewritten from scratch in C++ and moved into core, and BlenderNation flagged on 12 October that the set is now native in the 5.3 beta.
Why this one, and not the other four hundred add-ons
The reason is in Blender’s own tracker task for the port. LoopTools is the most downloaded Blender add-on there is — the task puts it at roughly 1.5 million downloads. For a thing that ships in the box and only needs enabling, that number means an enormous amount of people went looking for it by name.
The task also says, in as many words, that the port is an opportunity to fix long-standing bugs. Which is the honest reason to do this work. An add-on accumulates a decade of edge cases that nobody wants to touch because the Python is load-bearing and the author moved on. Rebuilding it in C++ means somebody has to decide what each operator is actually supposed to do when you feed it a non-manifold loop, and write that down.
Named in the port: To Circle, Space (Space Edge Loops Evenly), Flatten, To Curve, and Relax. GStretch appears as a rename — To Annotation — which makes sense, since what it does is snap a loop onto an annotation stroke. Bridge Edge Loops is a separate thing and has been a native Blender operator for years, independent of any of this.
The part I cannot pin down
Here the sources disagree, and I would rather say so than pick one.
Blender’s 5.2 LTS release page is reported to state that three LoopTools operators — To Circle, Space, Flatten — are built in to 5.2, which shipped in July. The 5.3 release notes describe the same three operators as now native and rewritten from scratch in C++ for improved performance. The roadmap marks the LoopTools port as done and slated for 5.3. BlenderNation, today, says all of them are native in the 5.3 beta.
blender.org does not resolve from where I am writing, so I could not open the release pages and settle it.
My reading of the muddle, offered as a reading: “bundled” and “native” are two different claims and the word native is doing both jobs. LoopTools has been bundled since forever. What 5.2 and 5.3 are doing is converting it, operator by operator, into core C++ mesh operations, and a staged conversion described by three different documents will produce three different answers about where the line currently sits. If you need to know for a production decision, check the build you are actually running rather than any page about it.
What a C++ operator buys a modeler
Python in Blender is fine for a tool that runs once on a selection and fine for anything that mostly calls into C. It gets expensive the moment it has to walk a lot of geometry itself, element by element, in the interpreter.
Which is precisely what these operators do. Relax, Flatten and Space are loop operations: they look at every vertex in a selection, work out where it should go relative to its neighbours, and move it. On a twelve-vert ring, nobody notices. On a retopology pass over a scan, where you are relaxing loops across a mesh in the high tens or hundreds of thousands of verts, the interpreter is the whole cost. Retopology is where Blender’s modeling toolset has been getting the most attention lately — there is now an add-on that reads a mesh and picks which of five retopology algorithms to run on it — and it is the same pressure showing up in a different place. Dense meshes, interactive editing, and a tool stack that was written when meshes were smaller.
So the performance claim in the notes is the plain consequence of moving a per-vertex loop out of the interpreter. It lands on the meshes where you felt it.
The other thing it buys is boring and worth more. Add-ons break. Every Blender release moves some part of the Python API, somebody has to update the add-on, and in the gap your tool is gone. Core operators do not have that failure mode. If your studio is pinned to one Blender build for a project, or your render farm runs an image somebody baked six months ago, an operator in core is there and an add-on is whatever was ticked when the image was made.
The thing to re-learn
Relax does not come across unchanged. Per the port’s design notes it becomes contextual: in vertex and edge mode it runs the LoopTools relax, and in face mode it runs Blender’s own Smooth. One entry, two algorithms, chosen by what your select mode happens to be.
I can see the argument — face mode relax never did what people expected anyway, and Smooth is the operation they wanted. But the muscle memory for this stuff is deep. Fifteen years of people hitting the same menu item and getting one behaviour, and now it depends on a mode you changed with a keystroke and forgot about. Expect a week of confusion and a few forum threads.
5.3 is in beta now, with release reported for November. The other change in that cycle I have written about is spectral dispersion in Cycles and why it renders so slowly, which is the glamorous one. This is the other kind of release note: a tool thousands of people use every day, moved somewhere it cannot fall off.