Weeknote #37 [W43] - Return of the weeknote, looking for jobs, and dancing landscapes

Return of the weeknote, after 14 weeks of absence

My last weeknote was a while ago.

I don’t remember precisely why I had stopped. But somewhere in my mind I probably had this idea that I should write more stuff beyond weeknotes.

But stopping weeknotes, evidently, hasn’t helped with that goal.

So, after a bit of prodding from Tanvi (who has also resumed writing weeknotes recently), here we are again.

One of the benefits of writing weeknotes is that I remember the weeks I write about much better than the ones I don’t. So the last few months have certainly been blurrier than the rest of the year.

Looking for jobs

I completed a year at my current job two days ago.

I got stuck at my previous job for almost 5 years because it was rather comfortable. I don’t want to make that mistake again.

Switching jobs — and switching frequently — is probably the most reliable way to increase your salary. Particularly early in one’s career. This is because employers often use your previous salary to determine their offers (which sounds completely idiotic to me, but it is what it is). Staying in one place for too long can have a compounding cost.

So, for the last week or so, I’ve been applying to as many open Product Manager roles as I come across.

I don’t have a lot of experience with job hunting. I’ve managed to avoid actively looking for jobs until now. I’m generally far too lazy to go through the tedium that job searches entail (like making a CV, filling up a hundred forms, writing cover letters, and so on).

Friends and colleagues tell me that applying through job portals (LinkedIn and the likes) is mostly useless. What works, apparently, are referrals from within a company. It’s only been a bit more than a week, but my experience so far hasn’t given me any reason to disagree.

Maybe it is too early to conclude anything from these numbers. Most applications haven’t received a response yet. But a quick Google search tells me that apparently only 3% applications through jobs portals get a response.

Which probably means I should start “networking” more. I dread the idea more than how much my cat dreads his baths (which he has been getting frequently lately, by the way, because: fleas).

Dancing landscapes, and the benefits of LLMs

As a part of the job search, I finally “finished” my portfolio site for the first time ever.

It has been in various states of unfinished-ness for the past 6 years. (I find compiling a portfolio for myself rather tedious.) This time, I figured that I needed to add something more engaging to keep me from abandoning it yet again.

Around the time I started working on the portfolio again, I was also listening to Understanding Complexity, a lecture series on complexity theory by The Great Courses. (Worth checking out if you have an Audible subscription.)

One idea I came across was that of dancing landscapes. The idea is that a complex environment is like a landscape that is constantly changing with time. In such an environment, the path to achieving a goal changes along with the circumstances.

I found the idea interesting enough to build a page of interactive illustrations around it: Dancing landscapes. That page eventually became the anchor around which I rebuilt the portfolio. Having a smaller project I was actually interested in gave me a reason to stick with the larger project of finishing the portfolio.

I’ve never been very good with making animated data visualizations. I’ve used a bit of R and matplotlib on Python for a few projects in the past. But, as fun as it was, I never went back to it because there was no way to justify to myself the time it took to work on those things.

But we live in the age of LLMs now. And they are really good, and incredibly fast, at generating code.

I used to think that I liked writing code as a hobby. I’ve realized now that I probably didn’t. What I actually liked was creating aesthetically pleasing outcomes, and sometimes that meant writing code. (Which probably explains why I like CSS so much.)

Generating aesthetically pleasing outcomes from LLMs is still quite frustrating. The ratio of garbage to good output remains high. But the amount of tedium involved in reviewing LLM outputs and prompting them to get what you want is more tolerable to me than the tedium of writing code from scratch.

The results of LLM generated code have also improved way beyond anything I expected even 6 months ago. LLMs now write, on average, better code than what I or my colleagues at work would produce ourselves. We’ve reached a stage where we don’t read most of the code that goes into production, while also getting relatively fewer bugs than human written code (which was definitely not the case for us 6 months ago).