Sprint Contract System
Pre-work contracts co-signed by builder and evaluator. Eliminates close-enough shipping. 48 contracts shipped.
Working templates, operating disciplines, and reusable tools. Open a playbook to see the method, examples, and available files.
Pre-work contracts co-signed by builder and evaluator. Eliminates close-enough shipping. 48 contracts shipped.
Blameless postmortems with structural action items. The discipline that makes the same mistake stop happening twice.
The annotated skeleton a Claude Code operator needs before week two. Three examples, a framing essay, and the advice everyone skips: what not to put in.
The chief-of-staff + N-specialists scaffold. Routing table, authority levels, and an escalation protocol — the piece everyone skips until it bites them.
13-folder Obsidian template that scales across multiple ventures. The operating system around your AI agents.
The MEMORY.md index pattern. Four memory types, a staleness protocol, and the one rule that stops memory systems from becoming archives.
Authority levels, the mothership pattern, generation/evaluation separation. The structural decisions most multi-agent setups skip until they break.
Context window hygiene. What belongs in CLAUDE.md vs memory vs code. The compaction strategy that keeps sessions fast after month 3.
How agents, hooks, memory, sprint contracts, kanban, and the Telegram bridge connect into one operating system for a multi-venture portfolio.
Five Claude Code hooks, production-tested: session loading, config protection, eval audits, decision logs, and anchor checks. Not another promise.
Five gates against leaking secrets, source license, or patent rights. Written the day I pasted my session tokens into a chat.
Don't build computer vision for every sport. Start with one constrained game, get it referee-grade, then generalize. Three filters and a trust bar.
A small operator can run a research project beside two startups without it cannibalizing them, but only under specific conditions. Parley is the worked example.
Four levers for a side research arm: scope rails, monthly cadence, public publishing, and decision logs, with kill criteria written down in advance.
Question-first contracts, signer holdouts, multi-seed floors, failure modes first, and a Deaf-community checklist. A 45% you can trust beats an 85% you can't.
A shared recipe across architectures does not control for the recipe. It shows up as an architecture result. Four cheap controls keep the ranking honest.
Assets built on a deadline are often never checked adversarially. The cheapest insurance is one evening of AI falsification, before the asset feeds a decision.
Five phases: reproduce the as-built model, rebuild from first principles, map the divergence, budget the uncertainty, and ship a ranked measurement list.