Seena LabsSeena Labs

New York, NY

Between Us

You've just stumbled upon the correspondence of co-founders Ax and Grant. These are their exchanges detailing their work, decisions, and highs and lows of building Seena Labs, as well as their opinions on life, tech, and business.

Get copied in.

You'll get notified of new threads. Privacy Policy

Skip to the open thread
The Correspondences3 threads

(003)

From: Grant

To: Ax Ali

CC: Those who dare to dream

The Danger of deciding what you’re not good at

1/2

Hello Ax,

Quick vent:

To me, there’s nothing more annoying than the overconfident, internally-inconsistent founder.

“If you want it done right, do it yourself.” Ugh. One sentence later: “My job is to empower my team and make myself unnecessary.” Okay. “Pay for bookkeeping/accounting? Don’t be an idiot. Waste of money. I do that myself with AI.” Next sentence: “Yeah I pay Cooley $100K/day. The peace of mind on legal is really worth it.” “Engineers actually slow you down.” “There’s no replacement for a 10x engineer.” “Design is commoditized.” “Design is what we are selling.” You get it. Talk to a founder long enough, and one of these will come up.

If I take a step back and try to be more empathetic, I realize that this is kind of endemic of a bigger thing in today’s tech world: we are all increasingly struggling with questions of relative advantage.

Relative advantage is as simple as: if someone else can do something better than you, they should do it, not you. Of course, there’s time and money factored in and all kinds of other stuff too. The underlying question is, “am I good at this?” Design: am I good at this? GTM: is this my superpower? CX: is this my “zone of genius?”

In the dot-com era, I could see how you could know these things. Maybe you scored particularly well in some subjects and poorly in others–and these subjects kind of mirrored professions. Maybe someone told you you were really good at something. Over time, you make small decisions about your strengths, which end up guiding what you do at work, which guides your career path. Industries changed slower back then, and you could spend your career profitably building on a few self-perceived strength areas.

Critically, you also make decisions on what you are not good at. This guides what you outsource. This limiting felt rational–how could it be efficient to be proficient at everything? But the times have changed, Ax.

Hot take: we can no longer reliably infer what we are bad at or good at. Legal is not like English class. Sales is not golf-schmoozing, nor is it banging on a CRM. The nature of our work has changed so much (yes, with AI), that our self-ascribed identities are no longer predictive. The underlying tasks here are now totally, totally different. Your mental framework for what they are is built on what they used to be. And the belief that they are your relative disadvantage makes them further away from you than ever before. Now you really won’t know what you’re talking about.

So what? So, deciding what you are good at is still fine. Deciding what you are not good at is very dangerous. That small decision drives knowledge areas you willfully avoid. In today’s scary job landscape, that area you fully outsourced, your “relative disadvantage”, could actually be the thing that makes your career a bit safer (downside) or wildly successful (upside).

Concretely, this does not mean that you must be the person who thinks it’s better to do everything yourself.

It does mean that you must try everything before you decide on it–and you can’t try it just once. You need to retry every few quarters–because the underlying tasks are changing so quickly. Because you used Cooley at your last startup does not mean you should use them again (or vice versa! (nothing against Big Law here)).

Example:

I have a strong and cohesive product vision, but…

2023: “I suck at product design” - yes, I sucked at design. I had no fluency in any design software, my ideas were cohesive but unoriginal. I avoided it and let the experts do their thing.

2025: “I suck at design” - yes. AI design tools suck. It’s easier and better for a human to do it, and that human needs to tightly integrate with engineering.

2026: “I suck at design” - no. Now my cohesive product vision is the biggest strength–and the bare execution is provided by AI. Now I can massively accelerate my team’s pace by stepping in and contributing here.

I’m glad I explored this area and allowed myself to develop some skills and domain knowledge. Beyond the business benefits, it ignited a lot of newfound passion for my work.

By contrast–even with AI, I do suck at legal.

One more risk: sometimes we think we are being rational in deciding on our relative advantage, but we’re just as often being avoidant.

Example: I tell myself that among the household chores, cleaning the fridge is something I’m just not as good at as my wife. Da truth: I just hate cleaning the fridge because old food grosses me out. And I think unconscious avoidance is the greatest plague of the founder. But that’s a topic for another day.

What do you think, my friend?

From: Ax Ali

To: Grant

CC: Those who dare to dream

RE: The Danger of deciding what you’re not good at

2/2

Hey Grant!

Well that didn’t take long! Two letters in and we’ve found something we disagree on. I’m glad! This is more fun than the alternative.

Let me start where we do agree. The line between "outsource this" and "let me take a shot at it" has gotten much harder to find. You're right that most people draw that line once early and then never redraw it.

But here's what I'm thinking: I don't think AI is what made you good at product design. I think you're already good at product design. You sucked at artifact creation. You said it yourself. In ‘23, your ideas were cohesive but unoriginal, and you had no fluency in any design software.

Someone who can look at their own work and correctly identify it as unoriginal is not bad at design. That's a part of the skill. You could see the gap between what you'd made and what looked good, but you just couldn't close it with your own hands. That's not a talent gap, that's an execution gap, and an execution gap is exactly the kind of gap an AI can fill.

Where we diverge here is that I would amend your framework instead of arguing with it. There are two different things hiding inside. “I suck at this”:

  1. I can't tell good from bad here, which is a judgment gap.
  2. I can tell I just can't make it, which is an execution gap.

Only one of those was ever hard. AI just took the other one to roughly zero.

And here is what scares me about that.

The reason you can tell good design from bad design in ‘26 is that you made bad design in ‘23 and noticed it, then made slightly less bad design and noticed it. You've put in enough time to recognize the shape of the wrong design. That's the apprenticeship, and I think we're at a real risk of losing it because AI lets you skip straight past the part where you're bad at something.

That's what I see when I look at all the AI slop. Somebody creates a 60-page deck and ships it as good work. The problem is not really the deck or the layout of the deck. The problem is that nothing in that person's process will ever tell them that it's bad. They got the output without ever building the instrument that checks the output, and the thing looks fine. That's what makes it so hard to catch.

And it gets worse the second a human being is involved.

Chemistry has right answers, so does physics. Design scholars Rittel and Webber called these wicked problems back in 1973. There is no stopping rule, no true or false, only better or worse, and every solution you try changes the problem you are solving. That's design.

An AI model can give you 100 good options, genuinely good ones. What it can't tell you is which one is right for these specific people in this context right now, because that was never in the training data. It's in the people.

And I think we might have arrived at a better question. Instead of asking, "Can I do this now that I have a model that can execute what I think?" I think the question is,

"Can I tell when it's wrong?"

Maybe that's the thing worth re-trying every quarter, and I don't think we can shortcut it. Now, you might argue: how do you know which gap you have? How can you tell good from bad? I also don't know what I don't know, and that's fair. That's why we can't run these tests on ourselves and that's when we need the professionals.

You're right about legal: neither of us should be drafting contracts with AI. Not because we can't produce one. We absolutely can, and it will look great to us. That's exactly the problem. We won't be able to tell if there is a missing clause, a loophole, or a line that leaves us exposed. No amount of good-looking outcomes fixes that. The judgment gap: there are only two ways to close it:

  • We hire someone who has it.
  • We go to law school.

I don't know about you, but I'm not going to law school.

Do you agree with this distinction that I'm making?

Cheers,

-Ax