The glass ceiling/ガラスの天井
A dark and honest account of an AI startup job, and reflections on workplace feminism.
•
I was furious on the day I was promoted.
It was a cold and damp Friday morning, and I left home early in the morning to go to my 9 am doctor's appointment. Perched on the waiting room chair, I opened my phone to check work Slack.
"Worth a read team", sent the top executive of the company. Attached was an industry newsletter from an anonymous author. I opened it and browsed the first page, with an AI-generated 2-panel comic prominent.
The first one depicts a printing machine labeled "NYTimes headlines generator", churning out a story about the connection between industry lobbying to the Trump administration and favorable regulation change. In the frame, a Jewish guy raises his hands high in the air, appearing shocked at the paper's distortion of truth.
The second frame shows the same Times printing machine "distorting" the news about "a country defending itself". Below the comics, the front-page article continues to defend both the industry lobbying and Israel's war crimes. I gasped.
"We should sponsor this", said the head of Engineering in the Slack thread.
"We did it a while ago", replied the top executive.
I turned off my phone screen, and attempted to calm my breathing. I told myself I was going to quit soon, and walked into the doctor's room.
It was 3 hours before I met my manager and received the news of promotion.
•
I came to this job after having been driven out of my last one by corporate politics and mounting stress. The new job offer came through a friend, who spoke highly of my then-future manager. The company was a healthcare AI startup, with a pretty good financial backing despite having only a handful of employees. I happily jumped ship.
However, there were always things that could only be seen from the inside. Like the company's sole supporter and funder was a radical Zionist healthcare-conglomerate-owner-turned-billionaire(one hell of a combination, as my therapist commented). Like what I would see once I walked into the office.
The office was at the top of a fancy Hudson Yards tower. The interior was clean, white and sterile. All the startups that the billionaire funded shared the same floor, which was always crowded and buzzing. Here, men were engineers, investors and executives, and women were receptionists, HR and marketing specialists. Almost all were white (mostly Jewish), although a few Asian engineers or analysts were sprinkled into the mix. The head of HR kept at least 3 pairs of heels under her desk, and strutted around the office everyday in 4 inch stilettos. At lunchtime, a circle of finance bros and tech bros would convene in the kitchen area, loudly chatting about macroeconomics and latest AI developments. Occasionally one or two women would sit aside and ask occasional, humble questions.
In the company I was working at, I was the only woman-adjacent employee that do technical work, let alone queer.
Suddenly I was longing for some performative DEI and loosely-arranged affinity groups.
Soon, I developed techniques to perform my version of professionalism and fend off misogyny. First, minimal conversational niceties: No "I'm doing great, how are you" or "have a nice weekend". Second, choose what to wear to work carefully. The universal trans girl's dream to be a corporate fem queen had to wait. Finally, I used my usually-low-and-sometimes-creaky trans voice to my advantage, and learned to not be afraid to cut people off. After a few weeks, I settled somewhere in between the two modes of workplace gender expression. Non-binary, you may call it.
Once I have built a character for myself, work was actually going smoothly, especially the actual "work" part. The good side of being the one-person team in a startup is that there were always things to change and problems to fix, and the solution required was usually not that complex. Easy to "make an impact", as in LinkedIn speak. And unlike in corporate, nobody will stop you from making changes, fight for scope or take your credit. Not long after I joined, I started overhauling technical infrastructure, educating colleagues about best practices, and delivering features earlier than expected.
•
It was not without its fair share of struggles and doubts during that time though.
One month into the job, everyone was asked to go to the large "offsite" event that the billionaire's family office was organizing. As the date was looming ahead, I worried for days about whether I would share a hotel room with a colleague, and if so which gender they would be. Both possibilities were equally terrifying. Thankfully, I caught a cold right before the event, and obtained a great excuse to return home for the night. So after a day of listening to soulless executive talks among men in suits and women in pencil skirts, I gladly took an Uber home from the villa they rented in central New Jersey.
The next morning, after another dizzy and somewhat frightening Uber ride down I-78 in a rainstorm, I found myself at an "engineering productivity" breakout session. In a (supposedly) meeting room where the chairs were mysteriously missing, fifteen engineers and a few managers sat on the ground in a circle around the empty room. All of them dudes except me. After some dry discussion around development processes, the conversation inevitably turned to AI.
"The AI models are getting better every month." "According to the trajectory it will probably be 10x better next year."
"Since we are all building AI use cases, how should we deal with the scenario where AI is going to replace jobs?" asked the facilitator, while sounding like he was treating a serious topic like a softball question.
"The change is inevitable, and the key is how to manage the change." a high-level manager from a behavioral health company volunteered to answer. "Change management", as I had already learned, means staying on the profiting side and containing the losing side.
He continued to share a story from his company, which had started using conversational AI to make phone calls with patients and was planning to use it to replace more human workers. It was originally only aimed to use for routine scheduling calls, while leaving more complex matters to human employees. However, as he described, call records had showed the AI having conversations for up to 40 minutes, which means that the topic had definitely deviated from whatever the original plan had been. "We have to adapt to the change as it's already quickly happening." He said at the end.
My chest tightened as I was listening to his words. I imagined in what kinds of scenarios would a mental health patient call the provider and end up having a 40-minute-long call with an LLM. Not the good ones. I remembered one of my friends who had sunk deep into "AI psychosis" after using ChatGPT to help them with mental health issues. There were so many things he could do differently to reduce risk, but instead he had chosen to release an untested and unguarded AI, put down his hands, fire his workers and say "this is inevitable". It never was.
I was probably the only person there to think this way. But that is why I should be there, isn't it? To be present, to be visible, and to be the different voice when needed. As good old second wave feminism has taught me.
On the rideshare back to the city, I reflected on the work that my company was doing, and it sounded pretty ethically safe. I vowed to be the voice of reason when things were going to the wrong direction.
•
Two more months later, and the promotion came unexpected. I was only two years out of a graduate program, and had just been working at the company for three months. To be offered a senior title which they ostensibly declined to give me during my interviews citing my lack of experience, meant that my work was being recognized and respected. It even came with a modest salary increase (although I was probably still underpaid compared to my male coworkers). The day was an emotional roller coaster, and I didn't know whether I wanted to laugh or cry at the end.
I decided to take the good news, stay and wait. On the surface, I was on track to break the glass ceiling, as a trans, immigrant professional who could build great things and lead a startup AI team to success. Maybe I could even try to change the workplace environment to be nicer?
So there was this most senior software engineer in the company, whose speech was always laden with profanity, even in completely calm and normal technical meetings. I suspected it was a performing masculinity thing, as the top executive would follow his suit at times. I was not okay with that, so one day I DMed him a polite but firm request to stop doing this. He politely apologized, and afterward he never used the F-word again in team meetings. Yay for girls who speak up?
I even stayed true to my internal commitment to be the voice of reason in the AI age. Once my manager vibe-coded a bunch of changes at night, bypassed the review process and was going to wreck some havoc. I confronted him the next morning, and made him walk back and begrudgingly apologize.
To be honest, those moments felt great. Things went downhill from there.
•
For a while, there had been rumors that the company was finally hiring a CEO after functioning for years without one. First it was about negotiations happening with the billionaire funder, then there was his name, and some vague words about his "vision" for the company. Oh and in the middle of this, US and Israel launched another war in the Middle East.
I typed his name into the Google search bar. According to his LinkedIn profile, he came from a popular baby-monitoring device company, and before that another health-tech one. Scrolling further down however, I saw "Israeli Ministry of Defense" pop up. Looks like he had spent his younger years there, as an engineering lead. I decided to dig deeper. A helpful Crunchbase link pointed me to his Twitter account, which has an anonymous hooded black figure as the profile picture. And there I had it, both a hot pile of propaganda and hate speech, and a rare glance beneath the civilized facade of the incoming CEO.
In a flurry of late-night replies, he was the secret keyboard warrior of the Israeli nationalistic state, firing equally at liberals, right-wing Americans and fellow Jews with a conscience. "You are stupid", "You anti Jewish m***n", "You are committing treason"... Not shy on language.
Some of his tweets come with the humor of a 20-year-old manchild. "I had a fight with my wife this morning, which is another genocide happening", he posted, under a tweet pointing out Israel's genocide records.
All his tweets have basically no views.
I closed the Twitter page when it was getting too triggering. It was hard to reconcile the Internet troll and the successful healthtech executive as the same person. I was going to check out this dude face to face.
After quite a long period of absent communications, one day he started coming to the office and an all hands meeting appeared on my calendar. At the meeting, he finally presented his vision for the company. After peeling off the narrative decorations, the plan was straightforwardly copied from his old playbook, which was moving from software to selling smart wearables, and the end goal was installing cameras to collect more data. The software that my coworkers and I had been working so hard on, would be no more than bait, used to build relationships with clients before selling them hardware.
After the speech, I caught the chance to rapid fire a long list of questions at him. From the obvious technical limitations and risky business assumptions, to the ethics and privacy question around his plan. He skirted around the questions, offering little more than aggressive assertions and empty promises.
As if to loosen up the mood, during the personal introduction part of the meeting, he said, "I don't usually speak in very nice words. It's not that I don't like you guys, it's just my MBTI. I am not good at talking, but I will be working hard to lead the company." Man, we have the same MBTI, it's not that, I promise.
I started planning for my last day right after the meeting. Have fun running an "AI" company after your one-person AI team quit.
Sometimes instead of trying hard to break the glass ceiling, it made more sense to doubt whether the shiny office tower should stand there. Continue showing up every day and devoting my energy to it in the hope that my 0.5% equity blows up meant something more than personal growth and girl power. It means complicity in enriching a genocide supporter and surveillance profiteer. Here, "lean in" feminism is powerless.
Days after the all hands meeting, the new CEO started creating new high-level positions for his old friends. "With expertise from elite IDF units", as they introduced themselves.
After sending out the resignation letter, I turned down his invitation to "talk and understand more" of my decision. More talking would not help with anything. On my last day, when I was chatting with my manager in the almost-empty office, he walked in, said nothing and quickly lowered onto his seat, as if trying to hide himself. Soon, I walked out of the office, backpack light and heads up.
Maybe one of my coworkers would remember a tall and awkward girlie, who wrote good code, and did not fear to let her thoughts and grievances be heard.
Thank you for reading it all. You can find all my essays using the tag.
P.S. A locally-deployed Gemma 4 LLM was used to check language and grammar issues. I am trying to migrate away from Grammarly.