• sidequest 2026-08-08

  Mapping a fake lending app to the Agentic Development Lifecycle

  Somewhere around the third or fourth session building Cashew Capital, I learned there's already a name for the process I'd been stumbling through: the ADLC. Here's how the work maps to it — and the gaps I'm leaving open on purpose.

• sidequest 2026-08-07

  Hardening a fake lending assistant: RAG and tool calls

  Cashew Advisor could talk convincingly about loans. It couldn't guarantee its facts or its math were actually right — until I added retrieval and tool calling to fix both, in that order.

• sidequest 2026-08-03

  Picking a model with evidence instead of a guess

  Before I let Cashew Advisor run on whatever model I happened to pick first, I built a small harness to test cost, latency, and safety across a few options — and found the cheapest one wasn't quietly worse.

• sidequest 2026-07-29

  Building a fake fintech company to learn AI-assisted infrastructure work

  I'm a sales engineer, not a developer. Over a few sessions with Cursor and Claude, I stood up a fictional AI lending platform on real AWS infrastructure — and learned more from what broke than from what worked.

• shiptalk 2026-07-25

  Three months into running deal reviews: what I'm learning about listening first

  I'm not a veteran at this yet, but three months of running deal reviews for my SE team has already taught me something about the order you say things in.

• shiptalk 2026-07-22

  Two things I learned from a week of QSRs

  Notes from a week of QSRs: what to do when a customer skips the POV, and a technique for finding the ask that actually moves a deal.

• shiptalk 2026-07-20

  Building technical champions is an SE job, not a side effect

  Champions don't happen by accident. A few patterns on identifying the right person early, and why relying on just one of them is the quiet way deals die.

• wanderlog 2026-07-18

  Planning Alaska like an SE planning a POV: dependencies, risk, and two kids

  11 days, Boston to Anchorage and back, two young kids — and it turns out planning a family trip isn't that different from running a proof of concept.

• shiptalk 2026-06-23

  Three prompts every SE should have ready before a QSR

  Copy-paste prompts for deal prep, qualification gaps, and champion health checks — built to work with whatever notes or CRM export you've already got.

• wanderlog 2026-06-15

  Six months I didn't plan for

  A trip home for a wedding turned into six months — old friendships, new ones, a changed city, and the family who made it feel easy.

• sidequest 2026-06-10

  How this site went from zero to live in one night

  Node, git, DNS, and a lot of trial and error — the real, unglamorous path from an empty folder to a working personal website.

• shiptalk 2026-06-01

  Five rooms, three cities, one trip to India from Boston I'm still thinking about

  Eight weeks back in India for work — GenAI talks, sales enablement, and the conversations that didn't end when the sessions did.