I Asked the Agent to Pick Tests From My Diff and It Said "None" Nine Times Out of Ten
Change-based test selection returned zero tests for most edits. Measuring how far an import graph actually reaches across three repositories, and rebuilding the selection contract around it.
Seven Characters Don't Only Point at Commits — Measuring When Your Agent's Recorded Hashes Stop Resolving
Short commit hashes recorded by an agent quietly stop resolving as a repository grows. Measuring where ambiguity begins, fixing the recording side, and building a batch verification gate.
Three Tools Named read_file: Catching MCP Name Collisions Before Startup
Bundle enough MCP servers and tool names collide quietly. Here is what a real 5-server, 21-tool setup measured at 43% collision, plus a Python preflight that catches them before startup and assigns deterministic aliases.
The More I Wrote in AGENTS.md, the Less Got Followed — Measuring Adherence and Cutting Rules
The rules in my AGENTS.md were being ignored — not from precedence conflicts or load failures, just plain ignored. Here is how I turned rules into checkable predicates, measured adherence over three weeks, and cut the file in half.
When Your Agent's Commits Pick Up Junk Files: Fixing It With Staging Scope and a Message Convention
Agents tend to run git add -A, sweeping .bak files and caches into your history, and leave a one-word message. Here is how a staging allowlist, a preflight, and a fill-in message template stop it.
Where Context Should Live — Routing It Across Persistent, Per-Task, and In-the-Moment Layers
Antigravity 2.0's /btw lets you inject context mid-task. But unless you separate what belongs in a persistent rule, a task instruction, and a passing aside, your agent's accuracy actually drops. Here is a design for three context layers.
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.
The Built-in Guide Skill Is Only Advice — Pair It With a Gate That Mechanically Rejects Antigravity's Output
The v2.2.1 built-in Guide skill raises how often the agent complies, but it is still probabilistic advice. Here is the design for a deterministic gate that reliably stops the violations that slip through, with working code and measured results.
Prune an Antigravity plan before you approve it
Instead of approving a Planning-mode plan wholesale, cut the one risky step and keep the rest. A field-tested look at partial plan editing from a solo developer's desk.
Precedence for Nested AGENTS.md: A Merge Design for Many Projects in One Workspace
Put several projects in one workspace, each with its own AGENTS.md, and which instruction the agent follows turns ambiguous. Root and per-project rules quietly collide; one wins, or both blend. Taking 'closer is stronger' as the base rule, this designs a merge that distinguishes overriding from appending, with working Python and field notes.
This Week at Antigravity Lab: Five Posts on Deciding How Much to Hand Over to Agents
Editor's notes on five posts: sizing agent tasks by what you can review, delegating dependency updates by risk tier, a StoreKit 2 migration case study, and a Japanese UI setup guide.
Prompt Caching and Context Strategy for Antigravity Agents — Cutting 60-80% Off Monthly API Costs in Long-Running Production
The longer you keep agents running, the more the monthly invoice quietly piles up. Running Antigravity agents alongside an AdMob-monetized indie app business (50M cumulative downloads), I managed to cut API costs by 60-80% by rebuilding prompt caching and context strategy. This article shares the three-layer cache, context compression, and TTL design I now run in production — with the code and numbers behind them.