Most common failure modes for agents during QA
How we set up every coding agent to QA its own work: an isolated environment per change, realistic data, a written playbook, and a preflight.
How to maintain review quality with 10x PR volume
Four check categories, three outcomes per finding, and why a “required” AI review can report a pass when the model behind it is down.
Install Mesmer for Cursor: one engineer, a fleet, or a Build
Mesmer's Cursor integration installs with one shell command: once per machine for an individual engineer, as a recurring Rippling script policy plus a configuration profile for a fleet, or inside a Cloud Agent Build via environment.json. Capture every repository, no per-project opt-in, no per-engineer opt-out.
Most common agent errors
Five common ways teams produce wrong or stale context for their agents, and four principles that keep it from happening.
You may be spending more tokens on context gathering than writing code
Five context-gathering patterns that burn tokens in agentic eng orgs, and the fix for each: scoped instruction files, context firewalls, zero-token polling.
Golden standard for agent readiness: treat agents as new hires
We build harnesses for engineering orgs, and this is the metric we've seen is reliable at predicting increases in speed and quality developing with agents, proven through 8 months of experimentation including +2000 engineers across +100 companies. Measure the onboarding of an agent as if it were a new hire, looking at time from access to a deployed first feature, with no help.
The playbook behind teams that 2x'd productivity in Q1
Hundreds of teams have connected their code and project data to Mesmer, and the ones that moved 2x+ faster in Q1 share a playbook. Humans stopped writing code directly, agents get triggered by alerts rather than by people, review moved off the human critical path in four stages, codebases were reshaped for retrieval, and pods shrank to two or three engineers per bet.
Life after DORA: eng metrics for agentic teams
DORA still works as a health check but it is a lagging indicator you cannot optimize toward. SPACE is worse off — its volume components are game-able and it pins system problems on individuals. What replaces them: org-level measurement, volume used as a cross-reference rather than a target, quality graded on every merge, review coverage weighted by effort, and two per-person metrics.