Sai Tharun Kakirala

This is a genuinely interesting vision — one more people will feel as AI agents become daily-use tools. We're building Hello Aria (launching on PH April 10th) as an AI productivity assistant via WhatsApp/iOS, and the social identity question shows up immediately. When your AI remembers preferences, has a consistent voice, and acts on your behalf — it's already behaving socially whether the...

My workflow has evolved a ton this year. I used to dump research into Notion, but context gets buried fast. These days I pipe key findings directly to Hello Aria (our AI assistant, launching on PH April 10th) — it keeps context across conversations, so when I ask "what did we figure out about X?" it actually knows. For longer deep-research sessions I still markdown-export and tag by project....

Custom domains are a serious trust signal for anything consumer-facing — we noticed the same thing building Hello Aria (an AI assistant launching on PH April 10th). For onboarding, the difference between a branded domain and a random cloud subdomain is night and day. On the infra side, TLS issuance is usually the hairiest part, especially at scale. Does Inquir provision certs on-demand per...

Love seeing more voice-first tools coming to iOS — it's where people actually live. We're also bringing Hello Aria (our AI assistant launching on PH April 10th) to iOS. The whole bet is that messaging is the most natural interface for AI — you just speak or type and it handles your day. Curious what the interaction model feels like on mobile when there's no physical key to hold. Excited to see...

The sidebar approach for creating everything makes a lot of sense. Dashboard tools often fall into the same trap - they're powerful but route you through a wizard-style flow that interrupts your thinking. Putting creation in-context is the right fix. We thought about something similar building Hello Aria (AI assistant via WhatsApp/Telegram, launching April 10th on PH) - instead of making users...

Love the concept! The thing we keep coming back to building Hello Aria (AI assistant via WhatsApp/Telegram, launching April 10th on PH) is that most AI tooling forces a context switch - you have to leave what you're doing to interact with the AI. Supercharging the textbox itself is the right instinct: bring the intelligence to where work is happening, not the other way around. Main question:...

The self-evolving coding agent concept is genuinely fascinating. 31 days of autonomous commits with no human pushes - that's a novel test of what AGI-adjacent tooling actually looks like in practice. I'm curious about drift detection: does the agent have any mechanism to flag when it's moving in a direction that contradicts the original design intent? We're building Hello Aria (AI assistant via...

That insight - early users aren't just customers, they're believers - is one of the most underrated things in early-stage building. The crowdfunding angle makes a lot of sense if you can get the regulatory structure right. We're seeing something similar with Hello Aria (AI productivity assistant launching April 10th on PH, ~3k users) - early adopters feel invested, giving feedback, referring...

The paradox of choice applies to AI too. More options often lead to worse decisions. We ran into a similar pattern building Hello Aria (AI assistant via WhatsApp/Telegram, launching April 10th on PH) - when the AI had access to everything, it tried to optimize for everything and ended up doing none of it well. Our fix was identical: explicit constraints. "Pick the ONE most important thing right...

The voice cloning angle is what makes this genuinely interesting. Hearing content back in a familiar voice is a fundamentally different experience than generic TTS. The commute use case is exactly right - passive consumption time is massively underused, and most people never think to reclaim it. We're exploring similar ambient intelligence ideas with Hello Aria (AI assistant via...

The value prop makes a lot of sense. Managed serverless hits real walls when you need isolation or custom domains, and most people only discover that after they're already in production. The cron jobs + webhook combo is especially relevant for AI agents that need to run reliably in the background. We've been building Hello Aria (AI assistant on WhatsApp/Telegram, launching April 10th on Product...

The "politely furious" email is perfect — AI misses all of it. It optimizes for surface tone, not actual emotional temperature. A few more I'd add: Knowing when to stop. AI is terrible at recognizing when it's outside its competency boundary. It hallucinates confidently rather than saying "I don't know." Understanding that silence is meaningful. A human notices when someone stops replying...

This is a fascinating experiment — and I'd guess the agreement rate was lower than most people expect, especially on anything involving nuance, recency, or value judgment. The more interesting question for me: on the questions where they disagreed, who was right? Or was there even a definitive right answer? That's where the real research is. We interact with multiple models daily building Hello...

I haven't used Prosaic yet but I'm curious — the terminal-first, distraction-free angle is compelling for anyone who's tried to write seriously in an environment full of notifications. My current writing workflow is pretty ad-hoc: I draft in Obsidian for anything long-form, then move it to wherever it needs to live. The context-switching between writing environment and publishing destination is...

This kind of post is what this community is for — honest reflection, not just the highlight reel. Thank you for sharing it. The part about not cleaning up old things that still have some life in them resonates. There's a tendency to constantly chase the new and deprecate what 'works but doesn't sparkle,' and you lose compounded trust with the users who still depend on those things. $2,398 in 48...

This really resonates. We went through the exact same phase building Hello Aria — shipping fast, celebrating deploys, treating velocity as the metric. The real turning point came when we slowed down and did user interviews. Turns out people didn't want more features, they wanted the ones we had to feel more reliable and personal. Since shifting to that mindset, retention has improved...

This is a great exercise — and I think the results usually end up being more useful as a forcing function than as actual predictions. The AI surfaces blind spots you already knew existed but hadn't formalized. We did something similar with Hello Aria (our AI productivity assistant for WhatsApp/Telegram/iOS, launching on PH April 10th). Fed it our positioning, feature list, pricing, and ICP...

Subscription is almost always the right call once your core use cases are recurring — PAYG creates budget anxiety that kills conversion, even when the actual spend ends up the same. A few things to consider: 1. Your $0 tier will either attract your best leads or your worst support burden. Worth being intentional about which features are gated there. 2. The gap between Pro ($49) and Business...

The "Todd from Finance" archetype is universal 😂 Every team has at least one. This is a genuinely interesting direction for Krisp — moving from technical noise removal (background sounds) to semantic noise removal (rambling content). That feels like a much harder and more subjective problem, and honestly a braver product decision. Curious how you handle edge cases: someone slow to the point...

Haha this is so real — what seems like just send a notification turns into a full distributed systems problem! We ran into similar chaos building Hello Aria — our AI productivity assistant that manages your day via WhatsApp/Telegram/iOS. WhatsApp delivery receipts + FCM tokens on iOS + Telegram update channels — keeping them all in sync and cleaned up is genuinely non-trivial. The token expiry...