logs

This week we’re focusing on two main goals. The first is to have a small but polished beta launch. This is related to our second goal of finishing our OpenAI Converge application.

Day 217: November 21, 2022

With the Converge app due on Friday, we decided that our goal would be to do a mini launch within the Neo community tomorrow. We had a long meeting detailing what’s important/what needs to get done before Friday, and then set out to do those very things. With the need to have a polished beta by Tuesday morning, today was especially brutal.

Some things that we did today:

  • Write drafts for emails/messages to be sent out
  • Lots of prompt engineering for new templates/culled old ones that were bad
  • Added autocomplete button back in
  • Added diffs back to suggestions
  • Revamped templates to have more guidance for prompt engineering (this way, even complete beginners can write higher quality prompts).
  • Windows support
  • Onboarding guide
  • Did some writing regarding OpenAI app
  • Collected more user feedback
  • Countless bug fixes/mini UX improvements

We all slept super late. Honestly a brutal day in terms of how tired I got by the end of the day—working while this tired has never been that fun, even dating back to high school and college. I guess that’s life though.

Obsidian plugin

11-21

Ended up with 14.5 hours of screen time today, which was suboptimal for sure.

Day 218: November 22, 2022

We launched to the Neo slack today, just to potentially gain a couple more users/additional feedback. From a number of signups perspective, the response was underwhelming—I had far more success with getting people to sign up/try the product through messaging them personally. This will be something to keep in the back of our minds though—it seems like our messaging isn’t enough to intrinsically grab people’s attention. However, this is far from our one and only launch, and we’re going to have a more official launch later this/next week when we have more polished features.

We have two main goals for today. The first is to generate a list of bloggers to reach out to and the corresponding drafts of these emails. John and I are focusing on this. Secondly, Kevin and Tae are focusing on building out our next key feature, which is our passive suggestions. In fact, I’m actually using that feature to write this log right now—it’s not quite live yet (I’m testing it out in my dev environment). We will push it out very soon.

AI Diplomacy

11-22

We went to hotpot tonight with our friends, but we ended up ordering way too much.

Day 219: November 23, 2022

Today was a slower day in terms of development. We did our usual bug fixes/minor UX improvements, but I spent most of my time today drafting emails and reaching out to various famous bloggers/writers. We essentially have two major goals at the moment: one is to create a product that’s actually usable for legitimate writers, and the second is to get our product into those people’s hands. We also have the Converge grant app due tomorrow, so we spent a long time discussing what our roadmap looks like. I personally have a lot of concerns about how we’re positioned relative to the big players in this space: Microsoft Office, Google Docs, and Notion. We’ve settled on a bottoms-up strategy that focuses on students, writers, and knowledge workers—this strategy will definitely take years to execute. I really hope it works haha.

Jasper AI Interview

11-23

another day, same room

Day 219: November 24, 2022

Happy thanksgiving! Today was a weird day in terms of work because we were nonstop working the whole day, but we wrote no code. We wanted to finish and submit our Converge application today. We wrote our script, recorded our lines/a demo, which took a surprising amount of effort. Some people have been posting their applications on YouTube as well—if what we’ve seen is our competition, I’m not super worried. However, I have a feeling that the general level of competition is going to be much much higher than what we can see on YouTube (Converge is only taking 10 startups, and I’m guessing that there’ll be hundreds of applicants). It also feels super weird that this application is for a million dollars, but it’s also laughably short/trivial. I guess you can tell a lot from a 60 second video, but at the same time it’s probably the application with the biggest implication for our future so far (aside from maybe Neo)—the length of the application does not reflect its gravity whatsoever.

swyx take on prompt engineering + generative AI

11-24

Friendsgiving in SF

Day 220: November 25, 2022

Today was also a slower day. We did a lot of reach outs and general bug fixing/UX improvements. The main thing I did today for work was conduct a user interview on one of our fellow Neo accelerator cohort founders. There was a lot of useful feedback—it’s clear that there’s still a long way to go before people find our product usable, but it’s given us clear things to focus on (namely prompt engineering to try to get better AI results).

Human-like GPT-3 summarization of websites

11-25

Eating at home is always nice