Startup land is wild. Customers can't log in, you're freaking out… then realise it's your auth provider who is also freaking out and shipping an SDK fix. chaos all the way down. 😂 Love it.

Be useful. It's that simple for me. Amongst my peers, I aim to be the useful friend.

How much quicker is Python 3.14? That's all I care about.

Managers have been vibe coding forever

Employees know what good AI looks like. So if Enterprise gives them something too far below SOTA, that's guaranteed pilot failure.

How to know if you're moving fast enough with your product ... 👇🏼

I love experiements and Deepseek's OCR is the latest of them. What if the memory problem what a visual proble all along? Coming to think of it, when I think long and hard at things I tend to visualise them in the back of my mind

LangChain and LangGraph 1.0, look awesome. Have a feeling I'm sticking with langGraph but a lot of exciting things on the new langchain

I love customer discovery, u start with a set of unknowns, and after a number of conversations, clear winner features/insights ... etc seem so glaring

🚀 𝗪𝗲’𝗿𝗲 𝗛𝗶𝗿𝗶𝗻𝗴 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀! Our team at 𝗥𝗲𝗱 𝗛𝗮𝘁 𝗔𝗜 is growing, and we’re looking for passionate engineers (at different levels) to help us push the boundaries of 𝗔𝗜 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 𝗮𝗻𝗱 𝗺𝗼𝗱𝗲𝗹 𝘀𝗲𝗿𝘃𝗶𝗻𝗴. We’re

Chatgpt caught in the act.

"AI-first" used to mean "data-first". Now it's more or less is just "llm-first"

It's a big mistake to skip technical signal when hiring a senior engineer. Pedigree ≠ proficiency

Too much shiny object syndrome in crypto. Makes it really hard to sift through the noise if you aren't in there full-time.

Nothing better than chatting with product users.

If you think "senior engineer interview" = no coding test, you're making a big mistake.

What does "AI-native" or "AI-first"even mean 🤔?

78% of candidates say they'd accept less pay to work somewhere mission-driven. People will trade cash for meaning when the math still works. Just don't use "mission" to underpay.

Supported language lists are becoming less and less important. We must keep getting these models as linguistically diverse as possible.

If you're going to do a live coding interview, having AI tab completion makes you look way better than going to stack overflow or chatgpt.