The first compromise was the title. David Castellano had called his screenplay "The Glass River," which was not a river at all but a metaphor — the way a small Pennsylvania steel town looked in winter when the ice formed on the Monongahela and caught the light from the blast furnaces, a sheet of dirty crystal stretching from bank to bank, beautiful and poisoned in equal measure. The story was about a father who worked third shift at the mill and a daughter who was slowly going blind from a degenerative condition, and the whole thing was quiet and sad and true. David had written it in six months in a one-bedroom apartment in Silver Lake, typing on a secondhand IBM Selectric that hummed when it was plugged in and sometimes shocked him when he touched the metal casing. He had come to Los Angeles in January of 1987 with the manuscript in a leather satchel his father had given him when he graduated from NYU film school. The satchel was too nice for his circumstances. Everything he owned was too nice for his circumstances, which were that he was sleeping on his cousin Marco's fold-out couch in Van Nuys and eating instant ramen three nights a week and pretending not to notice when Marco's girlfriend left passive-aggressive notes about the dishes.

The man at the production company was named Jerry Fleischman, and he was the kind of man who had been in the movie business since the decline of the studio system and had survived by never betting on anything that had not already been a hit. He had a tan that looked like it had been applied with a spray gun and an office on the Paramount lot that smelled of cigars and air conditioning. He read "The Glass River" over a weekend, or said he did, and on Monday morning he called David into his office and told him it was the best script he had read in five years. David remembered the exact words because he had written them down afterward in a spiral notebook he used for phone messages, the way you write down a doctor's diagnosis or a love confession. "The best script I have read in five years," Jerry Fleischman said. And then he said: "The title has to go."

"Glass River sounds like a PBS documentary," Jerry said. "Glass River sounds like something your grandmother watches on a Sunday afternoon. We need something with heat. Something with momentum. Something that tells the audience this is not a movie where nothing happens." He leaned forward across his desk, which was the size of a small car and completely empty except for a single script — David's script — and a brass cigar cutter shaped like a miniature guillotine. "What about Steel Town? Steel River? River of Steel?"

David said he would think about it. He went back to Van Nuys and sat on the fold-out couch and looked at his script, which he had called "The Glass River" for three years, through two drafts and five readings at friends' apartments and one staged reading at a black box theater in the East Village where the actress playing the daughter had cried real tears during the third act. The title mattered to him. The title was the first thing he had thought of, before the characters, before the plot, before he knew who the father was or what the daughter's disease was called. The image of the frozen river with the furnace light beneath it — that was the story. That was the whole story. He called Jerry the next morning and said he would like to keep the title.

Jerry paused for exactly three seconds. David counted them: one Mississippi, two Mississippi, three Mississippi. "Okay," Jerry said. "Okay, let's keep the title for now. But we have to talk about the ending."

The second compromise was the ending. In David's script, the father died. He died in the mill, in an accident that was not dramatic but industrial — a burst pipe, a cloud of steam, a body that could not be recovered for six hours because the temperature in that part of the mill was still too high. The daughter, now almost completely blind, stood at the window of their house and looked toward the river, toward the light of the furnaces that she could no longer see but could still feel as a faint warmth on her face. That was the last image. The warmth on her face. The light she could not see but knew was there.

Jerry said the ending was too depressing. "The audience has to leave the theater feeling something," he said. "Not necessarily happy. But something. You give them a dead father and a blind daughter and a frozen river, and what do they have? Nothing. They have nothing. Give them something, David. Give them one thing."

David gave them the dog. The father still died, but the daughter had a dog now — a golden retriever named Rusty that the father had given her in the second act, a guide dog in training, a symbol of hope. The dog was never in the original script. David had written the dog in a single afternoon, sitting at a Denny's on Sunset Boulevard because Marco's girlfriend had finally kicked him out and he had nowhere else to type. He hated the dog. He hated everything about the dog — its name, its symbolic function, the way it nudged the daughter's hand at the end as if to say everything would be okay. But the dog was reasonable. The dog made sense. If you were a studio executive spending four million dollars on a movie about a dying steel town, you wanted the audience to leave the theater with something. A dog was something. A dog was the least he could do.

The third compromise was the love interest. Jerry's boss, a man named Sid Kaplan who had produced three hit comedies in the 1970s and had not produced anything since 1982, read the revised draft and said the father needed a reason to live. "He's working at the mill, his daughter's going blind, his wife is dead — why doesn't he just jump in the river?" Sid said. "Give him a woman. Give him a reason to keep going." David gave the father a woman, a waitress at the diner where the father ate breakfast every morning, a single mother with a sharp tongue and a soft heart. Her name was Lorraine. David wrote three scenes for her, then five, then seven. She was good. She was actually good — better than the dog, better than anything he had written since the first draft. He found himself looking forward to writing her scenes. He found himself imagining the actress who would play her. He was, against his will, becoming invested in the version of the story that was not his.

The fourth compromise was the genre. Sid Kaplan took the revised draft to a marketing meeting and came back with a single sheet of paper covered in bullet points. "The research says the audience for this is women over forty," Sid said. "Which is fine, women over forty go to movies. But they don't go twice. They don't bring their friends. They don't buy the soundtrack. We need the eighteen-to-thirty-four demographic. We need action. We need stakes." The bullet points were: stronger antagonist, higher stakes, faster pacing, younger leads, broader appeal. David read the bullet points three times. They made sense. They all made sense. The father was forty-seven years old and worked in a mill. The daughter was fourteen and going blind. Of course the target audience was small. Of course the movie would not make back its budget. He was not writing a blockbuster. He was writing a small, quiet story about people who did not matter to the economy of American entertainment. And the economy of American entertainment was, as Sid had pointed out several times, an economy.

He made the daughter sixteen. He made the father a former union organizer who had been blacklisted and was now working at the mill under a false name. He added a villain: the mill owner, a silver-haired industrialist named Harlan Cross who wanted to close the mill and sell the land to a shopping mall developer. Harlan Cross was a cartoon, a mustache-twirling villain in a three-thousand-dollar suit, but he gave the story stakes. He gave the father someone to fight. He gave the audience someone to hate. David wrote Harlan Cross in a weekend and hated every line of dialogue but wrote them anyway because they were what the story needed. He was learning, slowly and without quite noticing, that what the story needed and what he wanted to write were no longer the same thing.

The fifth compromise was the money.

David Castellano had been living on nothing — on the dregs of a student loan, on Marco's grudging hospitality, on the calories in instant ramen and the caffeine in gas station coffee. In February of 1987, Sid Kaplan offered him thirty thousand dollars for the script. Thirty thousand dollars was more money than David had ever seen in one place. It was a year's rent. It was a car that started in the morning. It was the ability to eat in restaurants where you sat down and someone brought you a menu. He signed the contract in Sid's office, sitting in a leather chair that probably cost more than his entire education, and when he walked out into the California sun he felt something he had not felt in months: the sensation of not being afraid. The money was the cross. The money was the silver object he thought would save him, the transaction that would prove his worth. He did not understand this at the time. He understood that he had been paid for his work. That was what writers wanted. That was the dream.

The sixth compromise was the credit. In April of 1987, Sid Kaplan brought in another writer — a man named Chuck Fiore who had written three episodes of Miami Vice and had a reputation for punching up dialogue. Chuck read David's script and made 147 changes in three weeks. Some of them were small: a line of dialogue here, a scene transition there. Some of them were not small. Chuck changed the daughter's name from Claire to Amber because he said Claire sounded too old-fashioned. Chuck changed the setting from Pennsylvania to West Virginia because he said West Virginia was more cinematic. Chuck changed the father from a union organizer to a former CIA operative because he said, and David remembered the exact words, "everybody loves spies." David fought Chuck on the CIA thing. He called Sid and said the story was not about spies, had never been about spies, could not be about spies without becoming a completely different story. Sid listened. Sid sympathized. Sid sent him a check for five thousand dollars as a "goodwill payment" and told him Chuck's changes were going into the next draft.

The seventh compromise was the premiere.

The movie came out in the summer of 1988. It was called "Steel Thunder." The father was a former CIA operative played by Nick Nolte, who had accepted the role because his agent told him it was a chance to do something serious. The daughter was a seventeen-year-old played by a twenty-four-year-old actress who had been on the cover of Seventeen magazine. The dog was still there. The dog had survived every draft. The villain was still Harlan Cross, but now he was not just closing the mill — he was selling weapons to the Soviets, and the father's CIA past was the key to stopping him. The frozen river was still in the movie, for about twelve seconds, a single helicopter shot that the director had insisted on keeping because it looked beautiful. David watched the premiere at a theater in Westwood, sitting in the third row because he had not been invited to sit with the cast. He watched the credits roll, and his name was there, wedged between Chuck Fiore and someone called "Marcia Delacroix" who had apparently been brought in for a final polish, and the credit said "Screenplay by David Castellano and Chuck Fiore" which was the minimum the Writers Guild contract allowed and which meant, legally, that he had written at least thirty-three percent of the shooting script. He had, in fact, written zero percent of the shooting script. Chuck Fiore had rewritten everything. Marcia Delacroix had rewritten Chuck. The only thing of David's that remained was the frozen river, twelve seconds of beauty in a movie that was not his.

He sat in the theater after the credits ended, after the lights came up, after the audience filed out talking about Nick Nolte's performance and the car chase in the second act. He sat in the dark silence and tried to remember what "The Glass River" had been about. He could not. He could remember individual scenes — the father teaching the daughter to make pancakes, the daughter describing the river to her father because he had stopped looking at it years ago, the moment when the father first realized she was going blind and did not tell her — but he could not hold the whole thing in his mind at once. It had been taken apart. It had been rewritten, reorganized, reattributed, and what remained in his memory was not a story but a collection of fragments, beautiful and disconnected, like the pieces of a broken mirror.

That was the horror. Not that he had failed. Not that the movie was bad, though it was bad — a forgettable action thriller that made fourteen million dollars in its opening weekend and was never spoken of again. The horror was that the original story had existed. He knew it had existed. He had written it, every word, the Selectric humming and sometimes shocking him, the winter light coming through the window of the Silver Lake apartment, the daughter's name was Claire and the father was not a spy and the river was frozen and the light from the furnaces came through the ice like a promise. The song had been real. It was still real. It existed somewhere, in the leather satchel, in the spiral notebook, in the version of himself that had written it before the compromises began. And the compromises had all made sense. Every single one. The dog made sense. The love interest made sense. The villain made sense. The CIA made sense if you accepted the previous compromises, and each previous compromise had made sense if you accepted the one before it, and the first compromise — the title — had made sense because it was just a title, just words, just the thing you called the thing, not the thing itself. Except the title was the thing itself. The title was the image, and the image was the story, and the story had been replaced by seven reasonable decisions that had accumulated into something unrecognizable.

David Castellano walked out of the theater into the warm August night. The palms were swaying in the Santa Ana wind. The sky was orange with smog and city light. He was thirty-two years old. He was a credited screenwriter on a fourteen-million-dollar opening weekend. He would work steadily for the next fifteen years, fixing other people's scripts, taking studio notes and making them work, becoming known as a reliable closer, a script doctor, a man who could take a broken third act and stitch it back together. He would buy a house in the Hills. He would drive a Porsche. He would be good at his job. And he would never write another original screenplay, because the first one was still in the satchel, waiting, and he could not face the knowledge that it was better than anything he would ever write again — not because he had become a worse writer, but because he had learned too much, learned exactly what the machine wanted, learned to anticipate the compromises before they were asked, learned to kill his own songs so the studio would not have to.

The song existed. That was the thing he held onto, in the years that followed, when he woke at three in the morning and could not go back to sleep. The song existed. Claire was still blind and her father still worked at the mill and the river still froze in the winter and the light still came through the ice like a promise. No one would ever see it. No one would ever pay for it. But it was there, in the satchel, in the words, in the version of the story that had not been rewritten. It had existed before the compromises and it would exist after the compromises, not because it was good — though it was good — but because it was his. Not his to sell. His to have made. And the making was the thing no one could rewrite.