All Articles
Tracing Why an Agent Wrote That Line Six Months Ago — Commit Granularity and Provenance Trailers
When an agent packs 14 files and 800 lines into a single commit, git blame tells you nothing six months later. Here is how I split commits at intent boundaries, recorded provenance as machine-readable Git trailers, and built a one-command path from a blamed line back to the design decision behind it.
When Your Agent's Files Vanish Into .gitignore — A Pre-Commit Detection Gate
When an agent writes files that match .gitignore, the diff review looks perfect but nothing lands in the commit. Here is a gate script that catches ignored build output before you push, plus the tuning it needs.
Measuring the Rework Rate of What You Delegate to Agents: Drawing Delegation Boundaries with Numbers, Not Instinct
How much should you hand to an agent? I drew that line by instinct for a long time. Here is a practical way to compute a per-category rework rate from your git history and redraw the delegation boundary with numbers, with working code.
When Parallel Agents Ran the Same Task Twice and Quietly Doubled the Bill — Field Notes on Measuring and Stopping Duplicates
The bill for our parallel agents came in about 1.9x higher than expected — because multiple workers were running the same task twice. These are field notes on measuring the duplication, stopping it with idempotency keys, and attributing cost per task.
Never Miss a Managed Agent Completion: Pairing a Serverless Receiver with Polling Reconciliation
A cloud Managed Agent can finish while you are not watching, and the webhook that should tell you can quietly fail. Here is a serverless receiver on Cloudflare Workers paired with polling reconciliation, and a state machine that recovers every completion within minutes.
Protecting Your Agent Stack's Known-Good State with a Single Lockfile — Change-Budget Design for an Era of Simultaneously Moving Parts
When the IDE build, CLI, model, and dependencies all move at once, you can no longer tell which one caused a regression. Here is a change-budget design that pins your known-good state to one lockfile, with working code and operational logs.
Make the Self-Debugging Agent Walk the Logged-In and Post-Paywall Screens
By default, Antigravity 2.0's real-browser self-debug only sees the logged-out free view and reports success. To catch billing regressions, inject an authenticated session and paid state into the agent's browser and force coverage with assertions.
When Nobody Reads Your AI Code Reviewer Anymore — Field Notes on Measuring Actioned-Rate
Our production AI code-review agent quietly went hollow over six months. When the team started silently resolving every comment, we instrumented actioned-rate and false-positive rate to bring it back. These are the field notes.
Where to Put Evidence and Approval When Your Agent Self-Debugs in a Real Browser
Antigravity 2.0 launches a real Chrome mid-build, clicking buttons and taking screenshots to self-heal. It is fast, but shipping that as-is is risky. Here is how to capture evidence and draw the approval boundary.
Turning Last Night's Failed Runs into Tomorrow's Prevention — Designing a Postmortem Feedback Loop
Stop letting unattended failures end at a notification. A concrete design for classifying failures and feeding fixes back into Guide skills, gates, and schedules, with measured recurrence rates.
When the Tech-Debt Score Drops but the Same Files Keep Breaking — Field Notes on Instrumenting Fan-in and Churn
Letting Antigravity's architecture agent score technical debt is not enough — bugs often recur in the same files after refactoring. Here is how we instrumented the fan-in times churn that static complexity misses, and reconciled the score against real incidents.
It Worked Interactively but Went Silent Overnight — Making an Antigravity Agent Behave the Same in the Desktop and the CLI
An agent that runs perfectly in the Antigravity desktop app but does nothing when you schedule it through the CLI. This walks through absorbing the gap between interactive and unattended runs across four points — approvals, context, secrets, and runtime — with working code and a preflight check, so one definition behaves identically on both.