[Weekly Roundup] - Jan 3rd 2025
Future of Work
https://x.com/clairevo/status/1873099542796878096
I don’t know how the future of work is going to shake out. But I really agree directionally with this. Claire also has a really great podcast episode on Lenny’s Podcast.
Would highly recommend following her on Twitter, and I’ve learned a lot from her.
Code Generation
Lovable is incredible! Based on some cursory prompting, it seems to be better at generating aesthetically pleasing UIs than bolt or v0. However, I’m not sure if that’s because my prompts were too underspecified for the other tools to do a good job. Although I suspect that my tool of choice is likely to predominantly be Cursor. I’m just a lot more used to VSCode and being able to configure it however I want is really useful.
Cursor has an ARR of $48M - link
Some Googling suggests that they have ~12 employees?
It’s difficult to argue with revenue. And it’s certainly true that Cursor has become my IDE of choice mostly because of my familiarity with VSCode. However, I often find myself wanting to see the internals of how their RAG setup works. For example, I’ve seen people on Twitter recommend that it’s worth adding the docs for your most used libraries to Cursor. That seems very sensible! However, the overall process of how the IDE does retrieval is quite opaque. This can make it harder for power users to more effectively prompt the models.
In comparison, zed seems to be open-source and this Zed blog post suggests that they’re actively collaborating with Anthropic on AI features for the IDE. I wonder how many of Cursor’s existing ARR will churn once Zed becomes sophisticated enough. I’ve also personally started wondering if I should start using zed more often, and contribute PRs to zed itself.
Cursor’s in a pretty interesting spot in terms of their overall positioning in the market. Replit, Vercel v0, Lovable seem to be going after non-engineers and “beginners”. Or maybe more precisely, they seem to be going after folks that want to really optimize the journey from idea to “time to first customer”. And I wonder if “power users” will eventually start using zed (or comparable OSS alternatives) because it’s OSS. On the other hand, things in LLM-land are moving very quickly. I can see the value in being able to pay a company $20/month and have them be entirely responsible for providing “sensible defaults” for how to wield these models effectively in a VSCode-like environment.
As I’ve been ramping up on NextJS, I’ve realized that it might already be possible to interleave LLM inference calls with the generation of the frontend. This amazing post explores this idea in some more detail. I’ve talked to some friends about the idea that LLM inference may someday sit between the browser and frontend server. Client-side LLM inference APIs like this one. In particular, this post goes into a bit more detail about what this might look like. I suspect that inference cost will have to fall by 100-1000x for this to be a tractable reality. So…….15-24 months???
Does everything actually cause cancer?????
As an anxious hypochondriac-adjacent sort of person, I’m very grateful that such a list exists! However, I’m not actually sure whether or how I’ll change my immediate behavior based on this list.
Startups
Pattern Breakers by Mike Maples Jr and Peter Ziebelman
This was a great book. I really liked the overall ideas in the book, but found it a bit repetitive. I did find the case studies in the book interesting to read. But I’m quite excited that the authors did the work to distill insights, and would encourage you to buy the book to support them.
Although if you want a quick distillation of the material, this episode from Lenny’s Podcast with Mike Maples is probably the first place to start.
They also have this great substack that I liked a lot more than the book. Although note that I’m a free subscriber. I'm not sure they even have a paid option?
Spirituality
The Upanishads by Eknath Easwaran is an absolutely fantastic book! I just went through his translation of the Gita and really loved it. I’ve listened to a lot of lectures about Advaita Vedanta from Swami Sarvapriyananda on the Gita. And that’s been my doorway to Advaita Vedanta. But it was really cool to get a deeper dive via this book.
I’ve also been learning a lot more about Christianity recently. Although Christianity and Advaita Vedanta make different underlying metaphysical claims, it’s quite striking how similar the nuts and bolts of the “practices” for cultivating the mind seem to be. For example, there seem to be many similarities between Christian prayer and many eastern meditations like the cultivation of loving-kindness from Buddhism.