About

I'm Asif. I help people and solve problems. Right now that means building software with AI at a venture firm: the CRM, the agents, the plumbing that lets a small team punch above its weight.

I'm the only engineer in the building, so most of the interesting problems get solved alone. I ship something almost every day and write down what happened, out loud, caveats and all. This blog is that same habit pointed outward. The frontier stuff I'm working on doesn't have a lot of places to go, so it comes here.

How I work, if it's useful: I turn specific mistakes into specific rules. I lead with the number, not the vibe. When something's still rough I say so in the parentheses instead of hiding it. I'd rather ship a v1 and watch it in production than plan the perfect version. And I try to carry less. Observe widely, act narrowly. The interesting work is where attention compounds, not where it's spread thin.

The thing I can't stop thinking about: whether open-weight models catch frontier coding on hardware you'd actually buy. When a box under your desk runs an Opus-class coder, a lot changes. So I track it closely (Qwen, DeepSeek, the MoE efficiency curve, tokens-per-task over benchmark scores), even though I run the closed frontier in production because today that's still the right call.

Outside the terminal: dad of three, which is the real constraint the rest bends around. Studio Ghibli on repeat. And I think the most interesting open question in all of this is the human one. What happens to how we think when we start handing the thinking off.