I Never Felt Qualified, So I Learned to Speak Anyway

By the time I was invited to speak at Google’s MP7 building in Silicon Valley, I was already a Google Developer Expert.

The title should have settled the question.

It did not.

An organization does not give somebody that title by accident. Nobody had confused my name with another name, carried the mistake through the selection process, and then continued the administrative error until I found myself with an invitation to speak inside a Google building. There was evidence. It was public, specific, and difficult to explain away.

I explained it away.

The invitation did not feel like proof that I belonged in the room. It felt like the beginning of the examination that would finally reveal I did not.

I decided to give a talk about self-supervised learning with JAX and Flax. It was a field I had studied, implemented, taught, and thought about seriously. I had something to say. But as I prepared, the knowledge I possessed became less visible to me than all the knowledge I imagined the audience possessed.

Every explanation began to look obvious. Every slide seemed introductory. Every missing derivation felt dangerous. I imagined people in the room recognizing some simplification, leaning back, and silently reaching the same conclusion: This is what he came all the way here to tell us?

So I added depth.

Then I looked again and thought the talk was still not advanced enough.

The closer the day came, the more certain I became that I was going to get demolished. Not necessarily with cruelty. A precise technical question would be enough. A pause that lasted one second too long would be enough. One person in the audience would notice the edge of what I knew, and once the edge was visible, everything inside it would somehow stop counting.

I gave the talk.

Afterward, some of the other experts told me that it was too advanced.

I was confused.

We had apparently attended two different presentations. They had heard a talk that assumed too much. I had heard a talk that still failed to prove I knew enough.

I had invented a different room#

There was the real room inside MP7: people with different backgrounds, specialties, questions, and limits to what they happened to know.

Then there was the room I had constructed in my head.

In that room, Silicon Valley was not a place. It was a measurement. Everyone inside it was positioned ahead of me. They all knew the foundations, the latest papers, the implementation details, the mathematical objections, and the exceptions I might have missed. I had turned a collection of human beings into one imaginary intelligence whose only purpose was to evaluate me.

That was the audience I had prepared for.

I did not ask what the people in front of me needed in order to understand self-supervised learning. I asked what I needed to include so nobody could call the talk basic. I was not designing a path through an idea. I was building a defense around myself.

The actual audience paid the price for an attack that never came.

This is one of the less discussed effects of imposter syndrome. We talk about the pain it creates inside the person: the anxiety, the self-doubt, the feeling of being one question away from exposure. We talk less about what it does to the work leaving that person.

Imposter syndrome did not only make me feel unqualified. It sometimes made my thinking harder for other people to reach.

My life had entered the building before my self-image did#

I grew up in a small town in the south of Morocco, near Agadir.

I do not mention that as the beginning of an escape story. A small town is not a measure of a small mind. The people around me carried intelligence, craft, memory, humor, and forms of knowledge no institution had taught me how to name. What the town did not contain was a visible path from the life I knew to the rooms I would later enter.

Silicon Valley belonged to the category of places that arrived from elsewhere. I knew it through screens, products, company names, technical stories, and the biographies of people who seemed to have begun much closer to the center of the world than I had. Research arrived in English. Parts of my university education arrived in French. Authority repeatedly spoke from somewhere else.

Distance began to look like hierarchy.

The farther away a place was from where I began, the easier it became to imagine that everyone there knew something I did not. I mistook geography for a measure of intelligence. I assigned completeness to people because of the institutions around them, then compared that imagined completeness with every unfinished corner of my own mind.

Years later, my work could cross the distance. My self-image could not.

The people at MP7 met a Google Developer Expert invited to speak about a field he knew. Inside that expert remained the boy who had learned to see advanced knowledge as something arriving from far away. He did not object to the talk. He objected to the fact that we were now the person giving it.

My life had entered the building before my self-image did.

Credentials are evidence, not medicine#

I used to imagine that the next achievement would correct the image.

Perhaps confidence would arrive after I became good at the work. Then after I created something people used. Then after I taught it. Then after people sought my guidance. Then after recognition. Then after research. Then after an invitation from a place I once could not imagine entering.

Each achievement worked for a few minutes.

Then the mind moved it from the column labeled proof to the column labeled minimum requirement.

What I had already done became what anybody in my position should have done. The next thing became the real test. I did not allow accomplishments to accumulate into an identity. I converted them into obligations as soon as they arrived.

Every achievement entered the courtroom as evidence for one day, then returned the next morning as the new burden of proof.

This is how somebody can advance continuously while experiencing no motion. The world updates its description of him. His biography grows. The rooms change. The people asking for his opinion become more accomplished. But internally he continues measuring the distance to the next standard rather than the distance he has already traveled.

There is always someone more advanced. There is always a theorem I cannot prove, a paper I have not read, an implementation detail I would need to check, a question whose answer will not arrive immediately. If qualification requires the disappearance of those edges, nobody becomes qualified.

Expertise does not remove the boundary of your knowledge. It makes the boundary easier to see.

I interpreted that greater visibility as evidence that the territory inside it was empty.

Before a sentence left my mouth, it had to defend me#

Every talk had two audiences.

The real audience wanted to understand the idea.

The second audience lived inside my head and wanted to know whether I deserved to explain it.

Before a sentence left my mouth, it had to survive an internal peer review. Was it technically exact? Was it too obvious? Had I named every exception? What if someone knew a counterexample? What if simplifying the idea made me sound like I did not understand its complexity? What if the way I expressed it in English made the thought appear weaker than it was?

The sentence would enter this review clean and leave carrying qualifications, side paths, defensive details, and answers to questions nobody had asked.

Sometimes I knew the idea well and still expressed it badly.

I would begin with the central thought, see an imagined objection before the thought was complete, leave the main path to answer it, remember another condition, insert it, then try to return to a sentence whose beginning the audience had heard several minutes earlier. Every branch made sense to me because I knew the tree. The listener received branches without a trunk.

The problem was not always a lack of clarity in the thought. It was the number of checkpoints between the thought and the person hearing it.

Part of my mind spoke. Another part watched me speak.

Did that sentence land? Why did that person stop looking at me? Was the last point too basic? Am I moving too slowly? Did I pronounce that correctly? Should I add the mathematical version so they know I know it? What comes after this slide?

Self-monitoring consumed the same attention I needed to choose words, read the room, pause, and let an idea breathe. I was asking my mind to explain the subject and audit the person explaining it at the same time.

The audience heard whatever remained.

I confused density with rigor#

The MP7 talk was not too advanced because self-supervised learning, JAX, or Flax was inappropriate. A technical audience can engage deeply with all three. It was too advanced because my fear had changed the purpose of depth.

Depth was no longer there only to serve the idea. It was there to protect my image.

I feared that a simple explanation would make the subject look simple, and if the subject looked simple, perhaps the audience would conclude that I had nothing difficult to offer. So I compressed the path. I removed steps I assumed everyone already knew. I added technical material before establishing why it mattered. I treated accessibility as a risk.

But simple is not the opposite of advanced.

A simple explanation can be the final product of advanced understanding. It requires knowing which complexity belongs to the object and which complexity belongs only to the history of how you learned it. It requires choosing what to leave out without lying. It requires enough confidence in the structure to remove the scaffolding.

Density can hide the absence of those decisions.

I had prepared a talk that was difficult to criticize and difficult to follow for the same reason: it tried to include the defense before anyone had been given the claim.

I was measuring the talk by how much knowledge it displayed. The audience was experiencing it through how much understanding it created.

Those are not the same measurement.

Language had already made expression expensive#

In the language I could speak but could not think in, I wrote about the distance between fluency and thought. I grew up with Tachelhit and Darija, studied mathematics in Arabic, reached university and continued it in French, then rebuilt much of the foundation in English.

Expression was already a journey.

A technical idea might be encountered in English, attached to mathematics I had once studied in Arabic or French, tested through the intuition I could express most naturally in Darija, and then returned to English for a talk. This did not make the thought weaker. Multilingualism eventually made my thinking richer. But it meant that speaking clearly required integration I had largely been left to build alone.

Imposter syndrome added another language without giving it any useful words.

The thought no longer traveled directly from understanding to speech. It passed through fear first:

thought → internal judgment → defensive reformulation → technical language → speech

At every stage, clarity could be exchanged for safety.

The effect was cruel because it could make an actual strength look like a weakness. I could understand an idea deeply, see several connections at once, and care about its precision. Then I would try to release all of that simultaneously so nobody mistook selection for ignorance. The result could be a sentence less clear than the understanding that produced it.

I did not lack thoughts. I had too many thoughts about whether I was allowed to express the first one.

The audience was not a tribunal#

The feedback at MP7 did not cure me.

If imposter syndrome could be defeated by one piece of external evidence, the Google Developer Expert title would have defeated it before the invitation. A successful talk would have finished the job. Instead, the mind that discounted the title could also reinterpret the feedback. Perhaps I had simply misjudged the level. Perhaps “too advanced” was a polite criticism. Perhaps the next room would be different.

What changed was not the feeling. It was the authority I gave the feeling.

I began to understand that anxiety was not an assessment of my qualification. It was a prediction produced by an older model of myself. The prediction could be vivid, physical, and completely sincere without being accurate.

I also began changing the question I used to prepare.

Instead of asking, Is this advanced enough to prove that I belong here?, I tried to ask, What should the person listening be able to see by the end?

That shift sounds small. It changed where my attention went.

The first question keeps the speaker at the center of the talk. Every slide becomes evidence in a trial about him. The second places the idea between the speaker and the audience. The talk becomes an act of construction rather than self-defense.

I did not need to make every possible objection impossible. I needed to choose one path a listener could follow. The deeper derivations could remain available. The caveats could appear where they mattered. Questions could open the branches the main talk did not have time to explore.

The audience was not a tribunal. They were participants in the explanation.

I learned to let the idea arrive before its defense#

Becoming a better speaker required practices that looked almost too ordinary to matter.

State the central idea before qualifying it.

Give the intuition before asking people to carry the notation.

Pause after a difficult point instead of interpreting silence as exposure.

Do not answer the question you fear when the audience has asked a different one.

Let one slide do one job.

Move optional depth to the end.

Say “I don’t know” when I do not know.

None of these rules reduces the seriousness of the work. They prevent my fear from spending the audience’s attention on my self-protection.

The hardest practice was accepting that somebody in the room would know more than me.

This sounds obvious. Of course they would. A room full of experts contains many lifetimes of knowledge. But imposter syndrome treats another person’s expertise as evidence against yours. If they know something you do not, the mind does not say, Here is a person from whom I can learn. It says, Here is the witness who can invalidate the invitation.

I had to learn that expertise is local.

I can be the person guiding the room through one subject and the person asking a basic question about another. I can know a field deeply and still encounter an edge inside it. I can give a useful answer without possessing every answer that could follow.

Saying “I don’t know” did not make everything I had said before it false.

It made the boundary honest.

Speaking became easier when it stopped being about me#

In the previous writeup, I wrote about how giving talks, creating content, and guiding people in Darija helped me build a coherent foundation in deep learning research. Teaching forced me to reconstruct the path through ideas I had first encountered as disconnected techniques.

It also gave me a way out of constant self-observation.

When someone was trying to understand, their need was more important than my image. The question was no longer whether I looked like an expert. The question was whether I could make the next step visible.

That responsibility was more useful than confidence.

Confidence asks me to feel certain about myself before I act. Responsibility asks what the work requires while uncertainty is still present. I could prepare carefully, speak honestly, listen to confusion, and improve the explanation without first resolving every doubt about who I was.

I stopped waiting to feel qualified before doing the work qualified people had already trusted me to do.

This did not mean trusting every positive opinion and ignoring every criticism. The public image is incomplete too. Other people see a compressed biography: the title, the talk, the paper, the result. They do not see the failed attempts, the uncertainty, the preparation, or the questions still open in my head.

But my private image failed in the opposite direction. I saw the entire messy construction and refused to count the thing it had constructed.

They saw the outcome without the struggle.

I saw the struggle and discounted the outcome.

Neither view was the whole person.

“Not enough” can mean two different things#

I still often feel that what I know is not enough.

In research, this feeling is not entirely an enemy. Every serious question reveals missing knowledge. The closer I look, the more structure appears. A field becomes larger as I learn it because I gain the ability to see distinctions that were previously invisible.

There is a useful version of not enough:

What I know is not enough for the question I want to answer.

That feeling sends me back to mathematics. It makes me read, derive, test, and ask better questions. It is intellectual humility.

Then there is another version:

What I have become is not enough for me to deserve this room.

That feeling does not improve the research. It turns every missing fact into a judgment about the person missing it. It converts curiosity into surveillance and preparation into punishment.

For years, the two voices sounded identical to me.

Dealing with imposter syndrome did not mean silencing the first. I do not want to become a person who mistakes recognition for completion. It meant learning not to let unfinished knowledge become an unfinished right to speak.

My knowledge can be incomplete without my presence being fraudulent.

I spoke before the feeling gave me permission#

There was no morning when I woke up and finally felt qualified.

I accepted invitations while uncertain. I prepared while afraid. I stood in front of rooms before my internal image had agreed that I belonged there. I answered what I could, admitted what I could not, and discovered that the world did not end at the boundary between them.

Action updated me slowly.

Not every talk went well. Not every thought arrived in the form I wanted. There were times when fear made me dense, fast, defensive, or difficult to follow. There were times when I left a room replaying one weak answer while forgetting everything useful that had happened around it.

But speaking gave reality a chance to disagree with the private prediction.

At MP7, reality said the talk was too advanced.

The lesson was not that I was secretly more advanced than everyone listening. That would preserve the same hierarchy and merely place me on top of it. The lesson was that I had failed to see the people in the room as people. I had assigned them complete knowledge and assigned myself complete deficiency. Then I had built a presentation for that fiction.

The better talk would not have proven more.

It would have connected more.

I still sometimes enter a room as the boy who cannot imagine being invited into it. He sees the building, the titles, the distance from the small town near Agadir, and assumes somebody must have made a mistake.

I do not need to erase him. He remembers where the journey began. He protects the part of me that still wants to earn the trust behind an invitation.

I simply no longer ask him to decide whether the person I became is allowed to speak.

My life reached some rooms before my self-image did.

So I learned to speak while waiting for it to arrive.

← all writeups