Akshay

@unkonfined Let's try

People literally gain followers just by replying to me.

@AshwiniNK21 this is exactly the best kind of internal tool not fancy..! just removes one painful repeat task so well that nobody wants the old way back

lately it feels like a lot of systems break the same way they follow the rules perfectly until the situation changes and then nothing adapts fast enough

a single git push leading to RCE on GitHub is wild the scary part isn’t just the bug it’s how one unsanitized input crossed a trust boundary and suddenly millions of repos were in scope security failures usually start small, then scale brutally 👇👇👇 wiz.io/blog/github-rc…

@saguppa turning comment into a $25k/year deal is insane that’s not luck that’s having the right system ready when the signal shows up

One LinkedIn comment turned into a $25,000/year white label deal for my SaaS. Someone from BusinessCoachAcademy commented on one of my LinkedIn posts. I had no prior relationship with them. They had just dropped a comment on a post about LinkedIn outreach from one of my team members. Here's what happened next. Our AI Appointment Setter picked up the comment automatically, reached out via LinkedIn, delivered the relevant resource, and asked one qualifying question. By the time a human from my team got involved, the conversation was already warm. For some quick context: BusinessCoachAcademy runs a lead generation coaching programme for business coaches. They were already on Expandi. The outreach was running, replies were coming in, but their clients weren't following up on those replies fast enough and leads were going cold. The ROI on their coaching service was taking a hit because of it. Three things Expandi didn't have that we do: video messaging, voice notes, and an AI Appointment Setter that follows up on every reply automatically. Those three things were exactly what their clients needed at the time to stop losing warm leads in the inbox. They signed as a white label partner, for $25,000 a year. Their clients now run outreach under BusinessCoachAcademy's own branding. The deal started from a comment on a post. Someone read something, found it useful, and left a comment. That comment was a warmer signal than most things in a typical outreach list and the system acted on it within 24 hours. A comment on your LinkedIn post is not just a vanity metric. Whether you have a system to act on that signal before it passes is the important part.

@HelloPrerona this is the part most people miss better model doesn’t fix vague inputs, it just makes vague outputs look more polished the leverage is still in knowing what to ask and why you’re asking it

the more I read engineering blogs the more I realize the best lessons are rarely about the specific outage they’re about the hidden assumptions nobody noticed until the system got big enough to expose them

the gap isn’t “can AI write the message?” it’s whether the system understands why that message should exist segment timing context intent that’s where generic workflows start falling apart

you don't really know how your system behaves until 3am when something breaks and nobody's on call.

the bug that takes 4 hours to fix takes 30 seconds to introduce and it's always on a friday

google just shipped two different tpu chips one for training. one for inference. same problem, different enough constraints that one chip couldn't serve both well that's not a hardware decision. that's a systems thinking decision. cloud.google.com/blog/products/…

MCP crossed 97 million installs every major AI provider now ships compatible tooling what started as anthropic's experiment is now the default way agents talk to the world protocols win. products don't.

x is shutting down communities on may 6 something I was actually part of -- gone in two weeks not because the idea was weak because they never built it into something strong enough to last

@SahuTapan_ Are you still using claude or switched to codex?

Switched back to Visual Studio Code. Cursor looked cool… until my M4 started heating like a toaster. If your editor needs more resources than your app, it’s not productivity, It’s overhead.

one thing i’m learning most systems don’t break because of complexity they break because something simple was assumed and never questioned again

@HelloPrerona MCP

@aksh_twt well said!

You're still copy-pasting between tabs like it's 2019. MCP exists. Just saying. And what I love most about it is it’s not some developer thing. I can use it in plain english. MCP lets Claude reach outside the chat window and actually touch other tools. Your CRM. Your outreach platform. LinkedIn. So when people say "Claude + MCP + LinkedIn" they mean: Claude finds the right person, writes the message, and tells the tool to send it. You describe who you want. Claude handles research, Personalization, Execution. You watch. Old way: Find lead. Write message. Open tool. Send. Check reply. Repeat. You're the connective tissue. You're the unpaid intern between every single step. This collapses that into one conversation. SalesRobot doesn't have a native MCP yet but we have an API you can feed into Claude Code. It pulls up all the endpoints and starts running campaigns. Last night I used it to reach everyone who commented on my LinkedIn posts in the last two weeks. Drop "MCP" and I'll show you the full Claude Code session.

one thing i’m thinking about a lot while building a cli tool the setup is only useful if people can trust it good defaults, clear choices, and less hidden magic matter just as much as speed

@JenilLaheri 😅

@aksh_twt I think it's story for each project. 😅😅

started with what looked like a small bug in one part of the flow turned out a lot more of the system was relying on that behavior than anyone realized nothing dramatic on the surface but once that assumption broke, everything around it started acting weird

vercel got breached through a third-party ai tool an employee was using the tool got compromised. oauth did the rest. your security is only as tight as the last "allow all" you clicked