Mahir Kalra

Stoked to share what the @supertraceai team has been grinding on for the last year! 🚀🚀 fierce-network.com/cloud/supertra… Thanks @DGoovaerts & @FierceNetwork for the feature!

Ask a general-purpose LLM to diagnose a BGP route leak and it will give you a confident, well-structured answer that would almost certainly make the problem worse. It knows the theory, but it has never seen your routing table, your topology, or your policies.

Your dashboard tells you the link is degraded. It does not tell you what to do about it. That answer lives in a runbook from two years ago, a Slack thread from last time, and the head of your most senior engineer. None of it is in your observability tool.

We’re excited to announce today that @mastra has raised a $22M Series A led by Spark Capital. This brings our total capital raised to $35M:

It’s so exciting to see @supertraceai get the press it deserves ! 🚀🚀 He’s incredible product for network operators. fierce-network.com/cloud/supertra… Thanks @DGoovaerts & @FierceNetwork for the feature!

@abhiaiyer Ask and you shall receive 😊

This is so dope!

Wake up babe, new ST website just dropped. supertrace.ai

@abhiaiyer and team crushing it 💪

We’re Temple: the anti-survey company. We spent 2025 working with some of the top brands in the world to answer one question: how can they see what their customers see? From store execution to field marketing to product packaging, brands have been forced to rely on slow & inconsistent customer surveys for decades. Now, we’re super excited to redefine the category with our AI Mystery Shoppers, powered by best-in-class visual intelligence.

@chrisckchang @cursor_ai Daddy satya already has @GitHubCopilot Maybe aws since Sonnet is (currently) the best model?

@mahirk_ @cursor_ai And then daddy satya will buy them.

With the amount of usage data @cursor_ai is probably collecting I wouldn't be surprised if they beat Cognition/Devin to a true AI SWE.

@chrisckchang What about LLM input/output tokens?

I draw the line at tokens tho. No tokens for me

Apple Intelligence’s ‘aha’ moment will be on CarPlay

We're dealing with both 🙃

Dealing with the unpredictability of LLMs in code reminds me a lot of network programming. You end up having to implement retries, back off, and sometimes have to overlay transaction identification for idempotency. I wouldn't be surprised if this emerges as a formal layer.

I would love a cursor "spotify wrapped" at the end of the year.

@BhardwajRick ⚽️⚽️⚽️

Football will be the biggest sport in the world by 2035

Here's my lessons learned while building a product around AI agents for the past year: @cpdough/building-ai-agents-lessons-learned-over-the-past-year-41dc4725d8e5" target="_blank" rel="nofollow noopener">medium.com/@cpdough/build…

tl;dr: o1-preview solves an agent problem that no other model has been able to... saying NO bakeoff: gpt-4o vs o1-preview from @OpenAI task: SQL generation, but for a question that is not answerable using the provided table schemas (find orders that were cancelled because of out of stocks) gpt-4o: comes up with a SQL query anyway that will run and return an inaccurate answer o1-preview: thinks (for a long time) and then accurately determines that it is not possible to answer the question using the provided tables