All Dispatches.

Half a decade of building software for people who notice. A note on discipline, size, and why we still answer the phone.

The edge no longer comes from model access — it comes from proprietary context. Essential reading for operating partners rebuilding portfolio data systems.

The engineering philosophy behind systems that institutional clients depend on: correctness as the floor, not the ceiling.

For law firms doing cross-border work: why generic legal AI fails and what jurisdiction-specific models can do that semantic search cannot.

Why we serve institutions where mistakes compound. A note on who we build for, and why precision isn't optional.

Why we build AI systems as tools, not oracles. On the difference between augmentation and automation in institutional software.

Why fragmented data systems fail PE firms at scale, and what unified intelligence architecture looks like in practice.

The tools law firms use for discovery haven't kept pace with the volume of digital evidence. A technical overview of what comes next.

Based on conversations with 40+ institutional LPs: what they look for, what signals trust, and what makes them walk away.

Why institutional systems must maintain complete provenance. On the difference between what happened and what the data says happened.

Why institutional software is won or lost in the last 10% of implementation. A meditation on craft, edge cases, and finishing.

How we built a system that doesn't just find clauses, but understands relationships between obligations, parties, and dates.

PostgreSQL, server-rendered HTML, and why we keep reaching for tools that were already stable when we started the firm.

Why we use Postgres as our primary datastore, message queue, full-text search, job scheduler, and time-series database.

Why data quality problems multiply exponentially over time, and what PE firms can do about it before the next fund raise.