Shashank Indukuri

one memory for all your agents is nice. but if one of them saves something wrong, or mixes up two similar projects, the rest just trust it. curious how Notion deals with that 👀

OpenAI is acquiring Pierre. Their code review tools are already used in Codex. As more agents work on code at the same time, I’m curious how teams will review all those changes without missing things.

Tana now lets you bring Claude or ChatGPT into meetings. The agent can follow the conversation and see what’s on screen. I’m curious if it can tell when someone is just sharing an idea vs when the team actually makes a decision. that’s where meeting notes often get tricky.

Happy Monday, and it’s time to hustle after the weekend with Halloween and a lot of food. Let’s connect and make the best of this week. #letsconnect

Spent this week in Bellevue, near Seattle, attending an ML conference. A few topics kept coming up: • Evals for AI agents • Using a jury of LLM judges instead of just one • Agents that learn from feedback and improve over time Good to see more people talking about how we measure these systems and make them better.

Wispr Flow’s meeting MCP lets you connect your meeting history to an AI assistant. That could make finding things from old meetings much easier. But before connecting it, I’d check what the assistant can access. Meeting notes can contain a lot more than we remember.

Granola started with meeting notes and just added voice typing. Wispr Flow started with voice typing and added meeting notes. Interesting to see both moving toward the same workflow. For me, the deciding factor would be how well each handles names, technical terms, and corrections.

let’s connect and make this day to hit 100 followers #connect

confused on why the analytics shows 33 even if I just literally had 5 likes 😅 Am I missing anything?

this is a nice opportunity to test the waters! Time to run a voice agent evaluations

GPT-6.1 Terra looks interesting for computer use and visual tasks. For builders, the real test is whether it can handle a messy interface, recover when something changes, and finish the task. I haven’t tested it yet, but that’s what I’d be looking for.

Tailoring my resume used to mean bouncing between ChatGPT and Word and fixing the formatting for every application. So we built Vanara to make it easier. I used it for 5 applications and got 3 interview emails that week. We later open-sourced it so others could try it too.

I’m an AI engineer who enjoys building and shipping products. I’d love to meet more mobile app developers, founders, and people building with AI here. What are you working on right now?

Early flight to Seattle. Laptop open at the gate while I wait to board. If you’re building a mobile app or indie product, do you use travel downtime to make progress, or take the time off? I’m still figuring out the balance.

At AI Build Club DC Demo Night, I shared how I use Muse day to day: what saves time, and where it still gets in the way. The best part was talking with founders, engineers, and people just getting started. Building a mobile app around DC? What are you working on?

Hey X, I’m Shashank AI engineer by day, building mobile apps outside work. Looking to meet: • solo founders • mobile app developers • indie builders Let’s learn from each other and grow together 🤝 What app or product are you building? #connect