When the Android CLI Got 3x Faster and Cut Tokens by ~70%, the Right Move Was More Verification Per Change — Not More Parallelism
Reading that the Android CLI agent runs ~3x faster while using ~70% fewer tokens, my first instinct was to ask how many runs to parallelize. But a faster agent doesn't change how much work ships — it changes where the queue forms. This walks through why, sizes the new bottleneck (review and verification gates) with Little's Law, enforces a WIP cap with a working Python admission controller, and reinvests the freed budget into depth per change — with measured results.
Rolling Back a Half-Finished Agent: Compensating Transactions for Partial Failure
When you let an Antigravity agent run work that spans several external systems, a failure in the middle leaves the world half-rewritten. Retrying doesn't fix that. Here is how to fold it back safely with compensating transactions (the Saga pattern), with TypeScript and real operational numbers.
Flow Control for Autonomous Agents: Backpressure and Queues That Keep Production Alive
Run several Antigravity agents at once and the problem stops being how smart they are and becomes how little your downstream can absorb. Here is a flow-control design — bounded queue, semaphore, token bucket, backpressure, dead-letter — with TypeScript and real numbers.
Build for the Day the Agent Breaks Something: Keeping Blast Radius Small
Once you let an Antigravity agent touch production, the problem stops being how smart it is and becomes how much it wrecks when it slips. Here is a four-layer containment design that shrinks blast radius, with TypeScript and real numbers.