About
I’m Amit Agrawal. I’ve spent years in strategic GTM, sales, and partnerships roles inside complex organizations — carrying real accountability for revenue, retention, and team capacity in environments where knowledge is fragmented and coordination is expensive. I also bring a strong technical background from my time as a solutions architect, where I designed and built large-scale datacenter clusters, private clouds, and public cloud systems.
The constant across all of it has been the same orientation: seeing a real operational problem, diagnosing the underlying mechanism, and building the thing that should exist. On the infrastructure side that meant clusters and clouds at scale. On the customer side it meant translating those systems into outcomes that actually mattered.
AI collapsed the activation energy between seeing the pattern and shipping a working system. Builder isn’t a new identity I adopted when coding got easier. It’s the operating mode I’ve always been in. What changed is that the cost of going from diagnosis to a production-grade mechanism (agents, memory, distribution, observability, cost discipline) dropped by orders of magnitude. I wrote about this in Builder Is an Operating Mode, Not a Job Title.
This site is where I document the patterns that emerge when you run agents as real daily infrastructure — not demos or productivity theater. The perspective comes from having been accountable for both the technical build of large systems and the customer outcomes they were supposed to support and compound.
How I work
I move from specific problem to working system fast. Conviction first, then build. Proof over slides. Most of what I write starts as something I needed for myself (cost discipline in an experimentation account, graceful lifecycle for stateful workloads, making agent patterns actually reusable), then gets generalized once it proves itself under real conditions.
If a workflow or tool isn’t AI-native, I usually won’t invest in it. The friction and the ceiling are both too high.
What this site covers
- AI-native work patterns — how knowledge work actually changes when agents are real teammates (fleets over single agents, markdown as source of truth, llms.txt as distribution layer, cognitive modes that matter).
- Agent infrastructure that compounds — skills, memory, handoffs, observability, cost discipline, and institutional memory. Running real workloads sustainably instead of demos.
- The gap between what the tools promise and what they actually deliver when real coordination, real stakes, and real cost constraints are involved.
- Patterns that generalize across platforms, employers, and specific implementations.
The through-line is always what actually works when you’re responsible for outcomes that don’t care about the tool du jour. I care about the unglamorous parts: graceful lifecycle management, observability without compliance theater, moving static secrets out of paid services, and making patterns reusable so individual craft becomes collective capability.
See also How this site is built for the full infrastructure choices, syndication pipelines, and the ongoing “building in the open” experiment.
Disclaimer
All views and content here are entirely my own. Nothing represents my employer or any affiliated entity.