--- publish: true --- See also: [[_02022 Amble 4 – Plan]] _4 weeks from 23 May until 24 June._ My focus for this period was: software development and writing. Alongside that, my number 1 goal was to maintain high energy, optimism, and get my productivity and work hours back to 90-100% of normal, after a difficult start to the year (but a [promising recovery in Amble 3](https://notes.pjh.is/02022+Amble+3+%E2%80%93+Written+review)). I succeeded! Outline summary: - In week 1, I released updates to [Inbox When Ready](https://inboxwhenready.org/), continued user testing features I developed in Amble 3, and did a push on marketing. I also prototyped [Comment Helper](https://comment-helper.org), a Chrome extension for Google Docs, and shared it for feedback. - In week 2, I reviewed feedback on the Comment Helper prototypes, did some BOTECs, and decided to build an MVP. I worked long hours and finished the MVP. - In week 3, I did user interviews for Comment Helper, then another release based on feedback. I also reviewed Inbox When Ready feature tests and did some things in response. - In week 4, reviewed Comment Helper feedback, released some improvements, made a low-fi [marketing website](https://comment-helper.org/) and started the public beta. I also wrote a blog post on [naturalism and impartiality](https://sun.pjh.is/naturalism-pragmatism-impartiality) and a short one on [digital minds](https://sun.pjh.is/digital-minds-descendents-or-rivals). This was a productive period on software work, and I'm happy to have taken Comment Helper from first prototype to public beta in just 4 weeks. I'm hoping to see first sales within the next 2 weeks, and I have a decent [roadmap](https://comment-helper.org/roadmap/) for the next few months. I've tracked 70 hours of focussed work on the project so far, and I expect I'll have a fairly clear "this will work, or it won't" signal before I reach the 100 mark. Some more reflection on this in Appendix 1, below. I sorted a bunch of personal stuff during this period, including renting a flat for the summer, buying a car, reviewing my investments and sorting a bunch of administrative matters. Things with CK continued going well—our love is deepening, I reflected on things and felt good, and she met my parents for the first time for a 5 day holiday. The main disappointment this period was that I missed my goal of doing 20 hours on writing and P32 review (I only managed 13). I had one great writing day, one okay one, and one where I was low on energy and motivation and gave up part way through the morning. I did 5-10 hours of serious reading, most notably a late draft of [Will MacAskill's new book](https://whatweowethefuture.com) (it's good). The main reason for this, I think, was that I was working like a maniac on Comment Helper, making good progress, and I made several deliberate decisions not to "break the flow". When energy and motivation and obsession for a project are high, I think it is pretty good to focus hard on it, if you can (actually it can be kinda hard to think about anything else). My energy during the period was good or high, for the most part. In the last two weeks I was clearly starting a mild hypomanic phase, which is my usual pattern after after a low. Usually these phases are just great in all respects, but I'm monitoring this one more carefully than usual, since I'm trying new medication (first try of a mood stabiliser, probably a decade too late). In theory this medication should *not* contribute to hypomanic highs (or, god forbid, trigger a first episode of mania), but who knows. I had a couple of low energy days (tired?) and lost 2-3 work days to a heatwave I was poorly prepared for. Finally I did a 3 day silent retreat with a dear friend from Iceland. It was my first "self-organised" retreat (i.e. without a teacher, or other people) and I liked it. So... I'm happy with how Amble 4 went overall. 9/10! --- In this review week, I will aim to do a bit more writing. That said, I also need to plan Amble 5, keep an eye on initial response to Comment Helper release, and move apartments, so it's 50:50 whether I make decent progress here. I expect Amble 5 plan will include a bunch more time on witing. I really want to return to themes from P32, write up some stuff, and decide whether to call the study period "done" or whether to start another phase. I don't want to feel like the study period ended with me "stuck in a ditch". What next? See [[_02022 Amble 5 – plan]]. ### Appendix 1. Reflection on first 4 weeks of Comment Helper In the early stage of trying to make something people want, the top priority is: maximise rate of learning. First up, you want to get confident that lots of people actually have the problem you think they have, and that it's a big problem they really want to fix. Then you want to generate and test ideas for a solution. You also want to ask, fairly frequently: am I wasting my time here? If this is going to fail, how can I (correctly) realise that quickly and move onto the next thing? With [Comment Helper](https://comment-helper.org/), I had the problem myself, and I'd seen a bunch of lamentations about the problem from my colleagues at 80,000 Hours over the years. So I was going in reasonably strong. I made a Figma prototype in a couple hours, and shared it to these colleagues for comment, along with a couple questions. Their responses, plus some BOTECs, made me decide to invest 10-100 hours in taking this forward. Before deciding to build the MVP, I wondered if I should first do another round of protyping and user feedback, since that's much faster than building an MVP. I felt it was a somewhat close call, but decided against, mainly because I knew the problem well (since I face it myself) and was fairly confident in the outline of the solution I had in mind, and I know that one learns a lot from actually trying to use a scrappy MVP. All hail the huge benefits of working on something that solves a problem that you personally experience! I made a very scrappy private MVP in 1 day, and began using it. From that, I learnt a bunch, and then developed it through the week. I set myself the goal of getting a rough version ready for initial user testing by the end of the week, which would mean some hacky solutions, focussing on essentials, accepting a high probability of embarassing bugs. I made that goal. I agree with the YC advice that "if you don't feel somewhat embarassed, you shipped your MVP too late" and I kept this firmly in mind during the week. Nonetheless, I probably polished a few things unncessarily, but I guess only 5-10% of the work was (arguably) unnecessary at this stage (though another chunk was "wasted" as I discarded things during iteration). I found myself feeling high energy and quite obsessed during the work. I take this as a very good sign, and generally think that one should capitalise on these potentially very productive periods when they arise. So I doubled down a bit, punting a bunch of other things that had been on my plan. At the end of week 2 I was starting to feel some wrist and back pain, so I was glad that I could rest on holiday the next week. Generally my desk setup in Penne has not been ideal for intense work periods like this. Now that I've reached another level of commitment to living in France, it's probably time to invest in a better setup. In expectation, I probably should have done this back in January. Week 3 I made a screencast demo and started user tests. I arranged calls with a few 80,000 Hours colleagues who had expressed interest, and had them screenshare then install the extension. Standard stuff: I watched them start using it, then discussed their first impressions, and grilled them on some questions I'd prepared beforehand, based on my key uncertainties. Second half of week 3 was mostly useless, as I was sweating in a heatwave I was poorly prepared for. Week 4 I reviewed feedback, and made a roadmap to the public beta. Then I worked on all the things I'd need for that. I made decisions about initial pricing, setup Stripe payment links (super easy!) and made a minimum viable activation flow (for now, any license key will work... don't tell anyone!). I'm happy with the progress made in 4 weeks' calendar time (5 counting the holiday week) and 70 hours of focussed work. This process was greatly informed by previous projects, things I learnt from Y Combinator, and various product books and podcasts I've absorbed over the years. How could I have done better? - Some people seek waitlist signups and even payments before they even start building the MVP. That's pretty cool, but I didn't have the guts (or marketing libido) to try it. Plausibly I should have done. - What might this have looked like? If I had come up with the marketing pitch (based on a clearer problem statement at the start, which oddly I forgot to write out), I could have made a landing page saying "50 page Google Doc with hundreds of comments?" and listing out the envisioned features of Comment Helper. Then have a waitlist signup, or even a payment link. Hmm. - I did a lite version of this by asking colleagues at 80,000 Hours, and insodoing, effectively recruiting 5 people to the wailtist. I also asked about willingness to pay, to get a better read on severity of the problem. - I could have made high fidelity prototypes in Figma then hired a Toptal developer to build the MVP. This might have sped things up, but it easily might not have done. Andy Matuschak and I both believe that one learns a lot in the tight iteration cycle of building and testing initial MVPs oneself, so to be honest I think getting a Toptal developer in at this stage would have been a mistake. Once the beta testing phase is done, I will consider hiring a developer to build the final product. - I made the initial marketing site very quickly. In my mind, the key thing was to make the text good, especially the first few sentences people will see. The only message testing I did was to ask my girlfriend, who is absolutely not a potential user. If my judgement is bad here, then the first days of the public beta will go much worse than it could have done with a few hours' extra effort. Let's see. - I thought a bunch about marketing and distribution, but was way short of the 50:50 time split (between marketing/distribution and product development) that Gabriel Weinberg suggests in [Traction](https://www.amazon.fr/Traction-Startup-Achieve-Explosive-Customer/dp/1591848369) (excellent book). - I think trouble with marketing/distribution is the mostly likely reason this project fails. I have a few ideas for [product-led acquisition](https://www.julian.com/guide/startup/product-led-acquisition) but they aren't obviously going to work. Writing this I realise I should add testing these to the roadmap for July. - What else? **Dear reader, I'd love to hear your thoughts.** - To finish off this reflection, I will skim a product book I haven't read before, probably [Shape Up](https://basecamp.com/shapeup/) by Ryan Singer since Andy Matuschak's latest prototype for the mnemonic medium is using that book. - Update: I didn't like Shape Up! Poorly written, a bunch of unnecessary and badly coined neologisms, there are much better introductory books on product development. - The suggestion I most appreciated was: run two teams in parallel: "a build team" and a "figure out what to build team" (in his vocabulary, a "shaping"). The shaping team should mostly **not** talk to the rest of the team about product ideas until they've settled on high convinction views about what to build next. - (I read the first 2 chapters in full and skimmed a couple later parts.)