This monthly goal system went well in both january and february, so I am once again hoping to extend the momentum.

In the last two months I’ve built up a significant capacity to dedicate productive effort towards new goals, and now I want to cash in on some of that capacity. I’m quickly approaching an inflection point in my professional career, in which I can choose what computing subfield I want to spend the next couple of years working on. This currently feels like a high-impact but low-information decision, as I don’t actually have that much breadth of experience in different subfields. With that in mind, my main priority for March is getting some quick feedback by exploring my top 3 candidates: Security, machine learning, and formal methods. I chose these mostly because they feel like the most impactful fields moving forward (particularly as part of my desire to see future AI systems more robustly secured and aligned). The fact that all of them have a significant opportunity for math is an added bonus.

In order to get this signal in limited time, I want to do some preliminary trials of all of them simultaneously. By that I mean I’d like to spend at least ~40 hours trying to do some work in each of the 3 above fields over the month of March. Ideally I’d get as much of that time as possible done during the course of approved work time, and only have a small amount of overflow into my limited personal productivity time (I currently spend ~40 hours a week on work, and ~7 hours a week on personal productivity). As was pointed out to me by my challenge network, 40 hours is nowhere near enough time to actually get ramped up in a new area sufficiently to get a good idea of what it’s like on the inside, and a better timeframe is something like 6 months of full-time work. I fully recognize that this is a reasonable criticism, but I just don’t want to spend 18 months trying to make this decision (in that time I could pick a random field, work in it, and then be ready to start switching again!) I’m willing to spend an extra month to try and improve my decision, but not much more than that.

And with that motivation, on to the specifics.

Goal 1: On-the-Clock Exploration

I work full-time in computing and have some influence over the projects I work on. My team’s work is adjacent enough to security and machine learning that I can pick up projects in both without too much trouble. Thus my first goal for the month is to start working on both of these during work hours, with the target of spending >40 hours on each by the end of the month. Doing this during work hours allows me to get some of this exploration by using time that I would have already spent on different, likely-less-informative projects, so it feels very efficient.

My previous project is just coming to a close at the end of February, so I believe I have all the affordance I need to pick up the appropriate next projects, and have already discussed this with my manager, who is supportive. I predict I will successfully spend at least 40 hours working on a relevant ML project (70%) - I’ve actually already worked with a project lead on scoping out the appropriate subproject under their initiative, so I just have to make sure that my previous project doesn’t overextend by so long that it drowns this out. I would put (60%) on spending at least 40 hours working on security-related projects. This is in a similar boat of having already been worked out with the relevant project lead, but I expect the responsibilities to be more interleaved. That is, I’m much more confident that I’ll spend at least 5 hours working on security than that I’ll work on machine learning (95% vs 80%, respectively), but am less confident that I’ll spend a full 40 hours (60% vs 70%, just to restate the previous numbers).

Goal 2: Off-the-Clock Exploration

The above goal targets security and machine learning, but I don’t see a plausible way to spend approved on-the-clock time working on formal methods this month. In order to get that exposure, I’d like to invest some of the personal productivity time I’ve been cultivating the last two months. There so happens to be a formal methods study group that conveniently starts at the beginning of March and plans to cover Formal Reasoning about Programs, and the associated MIT course that started a couple weeks earlier (and is instructed by the author of that book).

To try and get a feel for formal methods, the goal is to keep up with the study group each week. Assuming the group progresses at the same rate as the class, and that the class has a typical workload for college classes, I’d estimate this translates to 3-8 hours per week of effort. The necessary time commitment depends heavily on whether I just follow along the reading, or whether I try to solve the problem sets. Given that I’m trying to actually understand what it feels like to work on formally verifying programs, I am aiming to do the problem sets. I predict (85%) that I’ll be able to spend at least 3 hours a week working through the book, plus an extra hour per week to attend the study group. This level of investment is probably enough to at least let me keep up with the reading, as it’s usually the amount of time that students spend in lectures each week for a given class. I would give a (75%) estimate that I’ll spend at least 5 hours per week, or fully keep up with the problem sets (whichever is sooner). In my experience college courses have a huge variance in the amount of time required to complete all the problem sets, and the 5-hour maximum allows me to filter this out a bit better and attempt to focus on the most available and informative parts of the problem sets. Giving a 75% chance of consistently spending 5+ hours a week seems too high given my current budget of ~7 hours a week in total; this estimate is taking into account the following goal.

Goal 3: Adding Time

Recently I’ve been becoming more and more aware that it’s very difficult to simultaneously keep my backlog clear of high-priority tasks, keep up with these monthly documents and posts, and still have time to fulfill commitments like those laid out in Goal 2. As such I’d like to add some more productive time to my schedule, and believe that it’s feasible to do this without messing up anything else.

I currently take the bus to and from work. The bus into work is spent productively (from my January goals), but on the way back I usually just unwind and relax. I’d like to change my habits around this return trip from “relax” to “do whatever productive-feeling thing you feel most interested in” - which I’ve come to refer to as “productive meandering”. This time is explicitly NOT for looking at my task tracker and forcing myself to do whatever the highest-priority item happens to be. That’s what the bus-in and weekend productivity blocks are for. During this block I shouldn’t feel like I’m really forcing myself to do anything. I expect that if I do this right and make it habitual (so that it doesn’t conflict with my expectations of how I’ll spend the time, possibly the subject of a future post) then I’ll feel at least as good about that time as I do about my current habit of relaxation. The hope is that it will feel as relaxed as free as it currently does, but that it will also feel more fulfilling in some long-term way. To extend the goal of not feeling like an onerous commitment, if I have some unusual night-time activity that would interfere with my other relaxation time, I will not hold myself to this goal (though of course I maintain the option of doing so anyways).

So with that specified, I predict (70%) likelihood that I will successfully spend my bus-out time doing productive meandering on every day covered by the terms above. The highest risk feels like it will come in the first couple of weekdays of the month, along with the Monday after the first weekend. To mitigate this, I’ve set a phone reminder to go off every day right before I get on the bus, so that I won’t suffer from forgetting. The remaining 30% is mostly just on the possibility that I have a really bad or tiring day and really don’t want to be productive. That being said, I have a really low focus bar for this time, and I’m hoping it will still feel relaxing, so I’m optimistic that it won’t be problematic even on such days.

In addition to this productive meandering time, I’d like to slightly increase the amount of focused, prioritized work I’m doing each weekend. This block is useful for getting more long-form tasks done, like blog posts or formal verification problem sets. I have currently committed to one 2-hour block per weekend. I often spend a bit more time depending on how I’m feeling and whether I’m close to completing some task, but I never force myself to work more than 2 hours if I don’t feel like it. This month, I’d like to officially extend the commitment to 2.5 hours. This doesn’t feel too burdensome, and gives me a good amount more of my most productive focused time. I’ve also noticed that (up to a limit) it’s easier to tack on time to the end of an existing work-block, when I’m already context-shifted into the task and (hopefully) in the zone, than it is to try and get myself to start a new work block.

For these extended weekend blocks, I predict (90%) chance of successfully completing the full 2.5 hours each weekend, given that the work block happens at all (so this would preclude things like full-weekend time commitments that prevent me from doing a block that week). The biggest risk factor here feels like cases where I just don’t have enough time in a weekend to fit a 2.5-hours block, but this feels unlikely given my current schedule. It also doesn’t feel like a failure if I still spend what time I can.

Parting Thoughts

My predictions have all turned out to be underconfident in the past two months, so I’m trying to update towards confidence. That being said, I still feel like it would be a mistake to significantly update my confidence. It could be that my concern for my own potential to fail has a protective effect against that failure, and that changing my confidence will make me less likely to meet my goals. It could also just be that I’ve gotten lucky the past couple months, or that I will “run out of steam” in some sense after a fixed time period. I’m doing my best to consider these seriously and avoid those failure modes.

As I have mentioned before, I’m no longer making it an explicitly numbered “goal” to call out that I plan to continue habits built in previous months, but as a brief point I predict (85%) likelihood that I will not mess up on any of those goals. Here I’m using “mess up” in a way that explicitly does not include anything unexpected that might come up, but which I would have said was a valid excuse if I had known about it in advance. This is just so that my estimation isn’t dominated by trying to predict whether some completely reasonable countervailing force will show up, and I can focus on the part that I actually care about.

edit: you can find my retrospective for these goals here.