Mental State
In one of the Next Token podcast episodes Dylan talks about the current developer mental state. Dylan talks about how his tweet on this topic was popular on X (I am not on X, so came to know from the podcast about the tweet). It resonated with me as well. With agentic coding it is hard to hit flow state, there is a lack of purpose in the day-to-day agentic prompting. It makes me uncertain about the future, which leads to anxiety, and a very confused state.
Few weeks ago I thought the best way to handle the current situation is to get closer to one of the GenAI companies that are the producers in this space.
I am only a consumer of GenAI tools, mainly agentic coding ones.
So strengthening skills to use the LLMs better, by building agents, loops, products etc., I thought was the best way to future proof my career.
I tried a bunch of job openings, interviewed at 2 places, but did not get an offer. This only ended up strengthening my uncertainty, or increasing the probability of some worst case events I was thinking of.
Career State
I have been at my current dev job for 4 years. I am being paid to write software for almost 7 years.
Before this I was paid to be a mechanical engineer, to build things, deploy physical things. It was in the Oil and Gas industry, it has its own set of uncertainties. In 2016 I was fired from my job at an Oil and Gas service provider, I was 29 at that time. I vaguely remember the company downsized quite a bit at that time, 10% of the workforce or something? It was the first time the economic situation had an impact on my work life. As someone who finished undergraduate studies and started working in 2008, it was a good run. I was blindsided by that episode, I did not anticipate it, it has put a sense of fear in me that I should get better at foreseeing such life changing episode.
The current impact that LLMs are having on coding, is bringing back this dynamic. As an almost 40-year old with a mortgage, wife and kid, the financial uncertainty is daunting. My wife works as a lecturer and researcher, her job seems to be a bit more insulated compared to mine? But her employer has had other financing related uncertainties. We have a decent emergency fund, about 10 months of monthly expenses as savings. I was willing to take some risk by moving jobs, but since the first 10–15 applications didn’t yield results, I had to either dive deeper to get serious about job change or put it on hold. My wife preferred the later as most of my actions felt knee jerk reactions.
What next?
I started exploring open-weight models using openrouter. Setting up this blog was meant to be a motivating factor to build and publish stuff, true learning can happen only by building. I think some version of RAG could be useful, I have not found the right use case yet.
There is the alternative approach of Andrew Kelley that he explains in don’t take the black pill. Its very nuanced and not a binary for or against the AI developments, its actually much deeper and roots on the human-computer relation. There is something to it that resonates with me. I need to dwell with this idea a bit more to have my take well formed.