It should be obvious the opamp output is restricted to the range (0,3.3V) or in practice much less (about 2,3.3V given the "741" tag) giving (12V, 13.5V) range or (13,13.5V) on the MOSFET gate.
At both ends of this range the MOSFET is fully ON, connecting 24V directly to the load. Which is what you are measuring.
So : Take a more organised approach to designing, rather than throwing some random bits together and wondering why it doesn't work.
First define what you really want.
At what temperature, what voltage do you want applied to the load?
I expect a function defining the relationship between output voltage and temperature. Or a table with at least three temperatures (too cold, normal, too hot) and the expected output voltages.
What thermistor resistance corresponds to each of these temperatures?
What gate voltage corresponds to each desired output voltage? Or rather, what range of gate voltages can you expect, given the variation in characteristics for your chosen FET? This will involve reading and understanding the datasheet.
Consider a FET with the minimum Vgs(th) and another with the maximum Vgs(th), and work out the gate voltages you expect for each FET at each desired output voltage.
Now you can design an amplifier that can take input voltage from that range of resistances and produce at least that range of gate voltages. You can use negative feedback (sensing the drain voltage) to eliminate the errors caused by the variation in transistor characteristics.
Finally you will have to consider system stability. you have three large time constants : the response time of the load (heater), the response time of the thermistor (sensing temperature) and the response time of the amplifier (dominated by R? and C? ... that's another thing : from now on, give every component a NAME aka reference designator, like R7 and C2 so you don't have to say that capacitor between the negative input and GND).
Anyway, those three time constants could easily make this a phase shift oscillator if you get them (and the gain of that amplifier) wrong, so it oscillates, alternating between too hot and too cold.
The calculations required for system stability are also known as "control theory" and I'm not going into them in this answer.