AI didn't kill portfolios

I used to think that my GitHub profile helped me because people could read my code.

Maybe I was wrong.

I created GoReleaser to automate the releases of my own software. It wasn’t any technological breakthrough (in the sense of inventing something new), but it solved a real problem: no one wants to keep gigantic shell scripts to release their software, or do it manually.

The code was not the interesting part. The interesting part is that I had the same annoying problem enough times to turn it into a tool, publish it, talk about it, keep maintaining it, and survive the consequences. And, of course, more people happened to have that same problem, and liked the way I solved it.

The portfolio was always a proxy

For most of my career, the advice was the same: have some stuff on GitHub. I’ve given that advice myself, countless times.

The logic was that recruiters would see it, so you could get pre-filtered into the next round, or at least not pre-filtered out of it.

But they never really cared about the code itself, or at least, I don’t think any recruiter ever read my code¹, at least not right away. For the interviews, I’m sure some developer will read it to ask questions about it.

I believe the code is merely a proxy for something else: your ideas, interests, things you’ve learned.

The code was never the signal. It was merely the first thing people could look at. And, for a long time, this worked - mainly because building anything used to take enough effort and time that the mere existence of a repository implied at least some of those other things were true as well.

AI broke the proxy

A model can now write, in a couple of minutes, what used to take weeks of after-work time. So having a dozen repositories no longer means much.

It doesn’t mean you spent a weekend on it. It doesn’t mean you thought about the problem. It doesn’t mean you learned anything. It doesn’t mean you can explain how it works.

“They have some repositories on GitHub, so they must be a builder” was always a lazy interpretation of portfolios. Not because portfolios stopped mattering, but because code got cheap.

What AI didn’t break

The upstream signals are still there. They’re harder to fake, and I think they got more important, not less:

• Taste: I wrote about this not long ago. Models can churn out endless competent-looking software. They can’t tell you which problem is worth solving, or which solution feels right.
• Community: dealing with users complaining is still hard, and now we also have the added complication of bots creating slop PRs all day long.
• Finishing: actually shipping, deploying, getting feedback, iterating on what you built, and being there for the long run. You know, actually caring about what you’ve built.
• Traction: if you made something people like, star, use, and talk about.

Having no portfolio feels worse now

In the old days, “I don’t have any public projects” was defensible. Maybe you worked in private repositories. Maybe you had no free time. Maybe your company didn’t allow it.

The bar to make something was high enough that the absence of it was ambiguous. You were given the benefit of the doubt.

Now, making something is often just a couple of prompts away, which, I think, means that the absence got louder.

You mean to tell me that, as a developer, there is nothing you would like to build now that the time cost is so much lower? Why aren’t your leveraging it to do more ambitious things?

Build something. One thing. Start to finish. Explain it honestly. Make it obvious where your hands were.

Or, you know, just help existing projects. That might be even more awesome, actually.

Good developers can be invisible

Of course, everyone knows someone with no online presence who is really good.

That is the obvious exception: a portfolio is not the same thing as skill. There are plenty of great developers with no public work, no blog, no talks, no GitHub activity.

But that is not the point.

The point is that strangers do not know that.

There is no hidden committee watching everyone’s work, noticing who is talented, and quietly moving them to the next step.² If all your good work happened in private, behind NDAs, or inside companies no one can inspect, then the world has to take someone else’s word for it.

A portfolio does not make you good. It makes your work visible.

A portfolio isn’t your repositories

Your portfolio is evidence that you can:

• notice a real problem
• make decisions
• weigh trade-offs
• improve based on feedback

Whether that problem requires an entirely new project or a PR to an existing one matters very little.

A merged PR to a real, used project is evidence. A bug report with a clean reproduction is evidence. A small write-up about how you solved an interesting problem is evidence.

So it can be side projects, micro SaaS, a blog, a YouTube channel, anything.

Just don’t fake it³ and you should be fine.

In summary

AI didn’t kill portfolios. It killed the lazy interpretation of portfolios. A repository used to imply effort; now it might just imply one prompt. The signal moved upstream: why this problem, why this solution, what did you reject, who used it, what did you learn, what can strangers actually see?

The career advice didn’t really change. If you loved this work enough to do it for free before, you probably still do. Show that.