Which local LLM actually runs well on my Mac?
Jul 1, 2026
Open-weight models are having a real moment, so I got curious about where they stand on my own hardware. I don't run them in Hindsight or my daily coding agents yet - this is exploration. Other projects already benchmark local models, but I wanted numbers for my exact laptop, so I built my own.
How I code from anywhere at any time
Jun 16, 2026
I work across many repos and many workspaces in parallel. Terminals and IDEs got in the way, so I built vipershell: a browser-based dev environment I reach from my phone over Tailscale.
We're in the jQuery era of AI. Who'll be the React of agents?
May 27, 2026
Agent harnesses today look like jQuery in 2006: pragmatic abstractions over a primitive whose surface keeps expanding. Nobody has built the React of agents yet.
Memory is the bottleneck for self-driving agents
May 8, 2026
Self-driving agents are bottlenecked by memory architecture, not autonomy. Forkability is the missing property - and the design that delivers it.
Why 10 million tokens is the only memory benchmark that matters
Apr 2, 2026
Hindsight is #1 on BEAM at the 10M token tier. At that scale, context-stuffing dies and only real memory architecture survives.
