Thoughts on building tools for a safer, more efficient relationship with artificial intelligence.

• August 21, 2026

  Engram 0.1: the memory vault ships

  The local-first memory vault is on GitHub Releases — Linux, macOS, Windows, source-available under the FSL, with a sync relay that never sees your plaintext.

• August 21, 2026

  Engram memory, measured: 100% recall, and what it costs

  We ran the Engram vault head to head against wholesale CLAUDE.md injection: 100% recall at both scales, 100% unsolicited tool adoption — and an honest cost crossover around 110-150 memories.

• August 22, 2026

  Drilling our own security, and a console that grew up

  This week: org-scoped API keys, audit-first tier gating, per-org usage metering — and a round of drills that attacked our own infrastructure, including a two-machine pairing test.

• August 16, 2026

  Audit first, enforce second: enforcement modes per organization

  Every organization on the console now starts in audit mode — full visibility, zero blocking. Enforce mode is one admin flip away, gated by plan tier and attributed by API key.

• August 15, 2026

  The relay: sync without the cloud seeing your memory

  Engram's public sync relay is live: your vaults stay on your devices, and the relay forwards encrypted blobs it can never read.

• August 12, 2026

  The guardrail, measured: 15µs, 37,700 checks/sec, zero false-ALLOWs

  We built a reproducible benchmark and ran it to the wall. The numbers were better than we claimed — and one finding surprised us.

• July 15, 2026

  Why deterministic beats probabilistic for security boundaries

  When the cost of being wrong is catastrophic, 99.9% accuracy isn't good enough.

• July 10, 2026

  The carbon cost of 'just ask the LLM'

  Every unnecessary model invocation burns electricity. We measured how much — and built an alternative.