Punit Arani

Introducing GrayPane - the simplest way to search for the best time and flight to fly. 1. Search for any one-way flights with filters 2. Find out the best dates to fly 3. Learn about available award flights 4. Set alerts to get daily or price-drop notifications

git worktrees took 10 years to pop off

consumer AI won’t be won by wrapping the smartest model - instead I'm convinced it'll have the following chracteristics: - AI native functionality reinvents the UX enough to move the needle - deliver enough new AI UX with "good enough" models - ARPU reliably outruns inference cost (as the latter goes down) - retention ends up stronger than non-AI incumbents - creates margin to fund distribution channels Thus, am particularly bullish about high ARPU consumer sectors (particularly with whale dynamics) like personal finance, health, productivity, gaming, etc - these categories already have willingness to pay, which means you can afford heavier models, more iterations, and better UX. Particularly variations of these ideas where an agent can dramatically improve the outcomes Meanwhile low ARPU categories get trapped in a race to the bottom, forced into cheaper models, worse experiences, and fragile retention loops. Particularly true for high global / low ARPU categories like content creation tools and communication apps etc As I mentioned earlier, it seems like 18-24 months before we can wrap AI functionality with remnant ads and it just works. Can def see a huge mega explosion of AI consumer in 2027 as this flips on which will be exciting

A CEO from one of our portfolio companies shared this with their team. I’m re-sharing it with their permission, because it resonated and reflects what all founders and CEOs should be communicating. -- We are living through a period of compounding change. And in moments like this, the biggest risk is no longer making the wrong decision. It is moving too slowly while the world moves around you. There are two paths. We can play defense: - Protect what we have - Optimize what works - Wait for clarity It feels safe. It isn’t. Or we can play offense: - Learn faster than the environment changes - Use new tools to solve old problems in better ways - And create entirely new strategies and businesses That’s where the opportunity is. Challenge yourself to do things faster and better than you have ever attempted. Stay uncomfortable. Stay on the front foot.

appreciate the transparency but the model the model is basically the same as GPT-5.4?? what’s the real advantage here?

The reason CLI is superior isn’t because MCPs suck, it’s because CLI requires access to a terminal and file system which is its superpower

@benhylak Claude gets dumber by the day

just me or claude dumber today?

/batch is underrated

@saharmalik111 I’ve never run into any limitations with dia and lowkey only use it moving out of arc It doesn’t save me meaningful time every day, it saves me maybe 5-15minutes not having to copy/paste to ChatGPT/Claude

@punit_arani Tempted to try Dia Pro for unlimited chats does AI integration with open tabs really save you time every day?

who here has paid for this?

I love cloudflare, all my domains and workers are there and still use rest because: 1. Monitoring and Observably sucks 2. Workers AI is limited and inferior 3. R2 is great if you’re in the cf ecosystem You can code up a lot of things and waste your time/money or just use something that works

@punit_arani @PlanetScale @Cloudflare Workers, AI gateway and data lake on R2 will get you a long way on all the others. You can easily code an 80/20 Doppler replacement with Cloudflare access and WARP and run it for dollars a month and have more security

You could do all this with @PlanetScale and @Cloudflare why overcomplicate?

intelligence is still in the hands of the rich

@HrubyOnRails @WisprFlow what do you use it for?

i have used very few products with such retention. @WisprFlow

@bdougieYO @ivanburazin @PlanetScale @Cloudflare @vercel @daytonaio @sentry @OpenRouter @braintrust @doppler people who know, know

@punit_arani @ivanburazin @PlanetScale @Cloudflare @vercel @daytonaio @sentry @OpenRouter @braintrust @doppler Nice seeing Doppler out here. Going at it for awhile

here's all the right answers db @PlanetScale backend @Cloudflare frontend @vercel sandboxes @daytonaio monitoring @sentry llms @OpenRouter ai o11y @braintrust secrets @doppler

its rare to find founders whose next milestone isn’t just hitting X revenue or raising Y dollars

nothing worse than a tweet getting more bookmarks than likes

someone’s out there building a white-glove ai service that’ll be printing cash

@shivam_310 @PlanetScale @Cloudflare @vercel @daytonaio @sentry @OpenRouter @braintrust @doppler @temporalio for any real project @Cloudflare for fun small ones

@punit_arani @PlanetScale @Cloudflare @vercel @daytonaio @sentry @OpenRouter @braintrust @doppler What about BG jobs

I just want LLMs to write as few lines of code as possible to get something working

@ndrewpignanelli what are your top ones?

Fun fact when you do a priced round you have to list every subprocessor + library from your code and their license for your disclosure schedule. Ours was ~5 subprocessors and ~150 libraries. It took 3 hours of founders’ time.

@ndrewpignanelli @PlanetScale @Cloudflare @vercel @daytonaio @sentry @OpenRouter @braintrust @doppler what do you use?

@punit_arani @PlanetScale @Cloudflare @vercel @daytonaio @sentry @OpenRouter @braintrust @doppler ai o11y