Golang interface vs. pointer semantics in interfaces

semantics.go

package main

import "fmt"

type Thinger interface {
	DoStuff()
	DoThings()
}

type thing1 struct {}

func (t thing1) DoStuff() {
	fmt.Println("thing1.DoStuff")
}

func (t thing1) DoThings(){
	fmt.Println("thing1.DoThings")
}

type thing2 struct {}

func (t *thing2) DoStuff() {
	fmt.Println("thing2.DoStuff")
}

func (t *thing2) DoThings(){
	fmt.Println("thing2.DoThings")
}

type thing3 struct {}

func (t *thing3) DoStuff() {
	fmt.Println("thing3.DoStuff")
}

func (t thing3) DoThings(){
	fmt.Println("thing3.DoThings")
}

func main() {
	var ti Thinger

	ti = thing1{}
	// ti = &thing1{} // ALSO WORKS
	ti.DoStuff()
	ti.DoThings()

	ti = &thing2{}
	// ti = thing2{} // DOES NOT COMPILE: "thing2 does not implement Thinger (DoStuff method has pointer receiver)"
	ti.DoStuff()
	ti.DoThings()

	ti = &thing3{}
	// ti = thing3{} // DOES NOT COMPILE: "thing3 does not implement Thinger (DoStuff method has pointer receiver)"
	ti.DoStuff()
	ti.DoThings()
}