End of April - reflections
(subscribe to Masala Dew on Substack)
- This was our first month of ops at Masala Dew. The original vision suddenly not only seems achievable, but inevitable. If not us, someone else would go on and do it.
- When I said "we"/"us" whenever I mentioned what's happening at our company, I meant me and my agents. And no, I don't feel weird about it at all.
- Increasingly getting happier about my decision to move out of Bangalore and avoid the doomerism encapsulating the tech world. We're living in a wonderful country: so many unsolved problems and opportunities that lie outside of the blr/sf tech bubble.
- Frontier labs are hyperfocussed on developer tooling, which is good news to anyone who's building outside of that market. A quick stroll through the markets of Delhi will tell you that there's SO much more to be built
- Shitty models work like shit even on good harnesses. Good models work amazing on excellent harnesses. It's often better to spend more on SOTA models than to waste time weeping over crap produced by a "cost-effective" configuration.
- Everything's about writing text files and executing them - which is why frontier labs are hellbent on getting it right. But new capability only makes new businesses possible, it doesn't automatically make them happen.
- I went (really) deep into the rabbithole of new AI tools this month, dug through almost every kind of hot new thing that has come in the market - upgraded to a Claude Max subscription and got converted to a Paper Pro user within the first two days of using it. My wallet's weeping rn
- Our goal for this month was to develop that "execution muscle" and try out the breadth of tools that exist to ship. But you don't always need the perfect tool for every little thing - some of the most productive people are building stuff out of copy-pasting stuff from ChatGPT.
- That being said, well-crafted tools do help. Both in getting stuff done faster, and also in giving inspiration on what's possible to build now. There's a delicate balance here.
- Open source is showing a lot of promise, it may soon become one of the best ways to generate distribution on products, esp tools used by passionate builders
- Still been procrastinating for stupid reasons; not being able to install Linux made me procrastinate on setting up OpenClaw. Old habits die hard.
- Quite inspired to see a lot of my friends starting up or on the fence about it: someone's gotta create all the lost jobs, and so many more!
- There are a lot of things to work on for the coming month - legal entity, tighter org structure, getting closer to our go-live date. Not being too hard on myself as we're in the middle of a seismic shift on how software gets shipped and distributed, and still personally adjusting to the cadence of working nights and weekends after being in a traditional corporate structure for so long
- aur kya kya karwaoge iss nanhi si jaan se?