Agents & Manager Articles
Multi-agent development with Antigravity — Manager Surface leadership, orchestration patterns, loop prevention, and monitoring unattended runs.
Your Antigravity Sandbox Isolates Multi-Agents Less Than You Think — Notes on Containing the Blast Radius
An Antigravity sandbox gives you the feeling of isolation, but isolation leaks through three real gaps: shared volumes, over-broad allowed domains, and approval fatigue. Field notes on plugging the leaks, containing the blast radius by design, and proving isolation holds with tests.
How to Orchestrate Multiple Agents: Drawing the Line Between Parallel and Serial Work
Antigravity 2.0 brings true parallel execution across multiple agents. But making everything parallel does not make it faster. Which work should fan out in parallel, and which should stay serial? This is an orchestration design that does not fall apart, viewed through dependencies and contention.
Parallel or Keep It Serial: The Break-Even Point When Orchestrating Multiple Agents
Should you run agents in parallel or keep them serial? A simple way to estimate the break-even between coordination cost and saved wall-clock time, plus how I actually split parallel vs serial across four scheduled sites.
Handing Visual Regression to a Parallel Agent in Antigravity 2.0
A design for running a dedicated headless visual-regression agent alongside your main implementation agent using Antigravity 2.0's parallel orchestration — with a working harness and the reproducibility traps I hit in production.
When Your Antigravity Agent's Usage Ledger Quietly Drifts From Stripe's Bill — Field Notes on Idempotency, Late Events, and Reconciliation
Usage-based billing for Antigravity agents fails silently when your internal usage ledger and Stripe's Meter Events aggregation drift apart. Field notes on idempotency keys, absorbing late events, the 35-day window, and a daily reconciliation job.
Three Boundaries I Draw Before Handing Work to an Antigravity 2.0 Agent
What to hand a background agent, and what to keep in your own hands. The three boundaries I actually drew while running solo-dev automation in parallel, and how to encode them so the lines hold.
Spending Less on Failure Without Swallowing It: A Retry Budget for Agents Built Around Gemini 3.5 Flash
A design that separates an agent's retries from quietly swallowing errors: classify the failure first, then retry within a budget. Grounded in the speed and price of Gemini 3.5 Flash, with per-task caps, logging, and a weekly tightening routine.
Stopping Parallel Agents from Clobbering the Same File
When you run several agents at once in Antigravity 2.0, two of them can write to the same file and one set of changes silently disappears. Here is how to design write arbitration inside a shared workspace.
After Generating Several Candidates, Which One Do You Adopt? Designing Best-of-N That Arbitrates by Verification
With Gemini 3.5 Flash's speed, generating several implementations of the same task has become practical. The hard part is no longer generation but arbitration. Here is the design and TypeScript implementation of a Best-of-N arbiter that picks the winner using verifiable signals only — not majority vote, not self-reported confidence.
Accounting for Which Agent Spent What: A Cost Attribution Design by Task
Your month-end bill is one number, but running multiple agents on Gemini 3.5 Flash hides which task ate the cost. Separate from a budget guard, I share a cost-attribution accounting design that maps usage to per-task and per-site cost, with a solo-operator implementation and numbers.
Tracing Parallel Agents After the Fact: Observability with Structured Logs and Spans
Running multiple agents in parallel on the Antigravity 2.0 desktop makes it impossible to tell which one is doing what. I share an observability design that drops tangled print debugging for run_ids and spans you can trace afterward, with a solo-operator implementation and numbers.
Making Managed Agent Batches Safe to Re-run: Idempotency and Checkpoints
Running overnight batches on the Antigravity 2.0 Managed Agents API makes recovery from partial failure unavoidable. Starting from a duplicate-post incident, I share the implementation of idempotency keys, a checkpoint store, and resume logic, with real numbers from solo operations.
When Your Antigravity Agent Eval Gate Keeps Flickering — Build Notes on Pass/Fail That Survives Non-Determinism
Same code, yet the eval passes in the morning and fails by noon. The first thing that breaks when you put agent evaluation into CI on Antigravity is the stability of the verdict. Here's how I separate noise from real regression and lock down pass/fail in code.
Generating Multilingual Release Notes with the Managed Antigravity Agent via the Gemini API
A hands-on record of building a pipeline that turns git commit logs into multilingual App Store and Google Play release notes using the Managed Antigravity Agent, now in public preview through the Gemini API.
Bundling Nightly Tasks With agy Async Jobs — A fan-out, poll, join Design
The Go-based Antigravity CLI (agy) can detach jobs and run them asynchronously. Here is a fan-out, poll, join design for firing many long-running tasks at once, collecting their job IDs, and waiting for completion — drawn from an actual nightly batch.
Treating the Managed Agent as a Cost-Capped Throwaway Worker: Isolating Untrusted Input from Production
How to use the Managed Antigravity Agent, now in Gemini API public preview, as a throwaway worker that is born and discarded per request. Cost caps, isolation, and idempotency with implementation steps.
Receiving Managed Agent Async Jobs Through a Propose, Verify, Adopt Pipeline
The Managed Antigravity Agent, now in public preview via the Gemini API, autonomously plans, executes, and verifies inside a sandbox. Here is a design for catching its async deliverables through three stages — propose, verify, adopt — before they reach production, with implementation code and operational pitfalls.
Stop Letting Antigravity Agents Self-Report 'Done' — Completion Contracts and External Verification
An Antigravity agent reporting 'done' when the work was not actually finished is a failure mode I kept hitting. Moving the completion decision out of the agent and into code fixed it. Here is the contract, a three-layer verifier, and how it holds up under unattended, scheduled runs.
Calling a Managed Antigravity Agent from the Gemini API: Design Notes on the Preview Model
antigravity-preview-05-2026, now in public preview on the Gemini API, is a Managed Agent that plans, runs code, edits files, and browses the web autonomously inside a sandbox. Here is how it differs from rolling your own orchestration, and where to draw the line.
Containing Failure in Antigravity Multi-Agent Systems: Three Boundaries That Stop Cascades
Antigravity multi-agent setups run beautifully in isolation but cascade in production, where one small failure drags the whole orchestration down. These notes organize the fix around three boundaries—layered control, trust separation, and observability with idempotency—down to the TOML and the correlation-ID wrapper.
Designing Schema Evolution So Sub-Agent Handoffs Never Break
Put a typed contract at the boundary where a downstream agent receives an upstream agent's output, and learn how to evolve that schema without breaking existing flows — with validation code, a migration sequence, and the production symptoms to watch for.
When Managed Agents Run in the Cloud, How Do You Hand Them Credentials?
The Antigravity 2.0 Managed Agents API runs agents in the cloud, away from your machine. Convenient, but the credential handling that was trivial on your own laptop suddenly gets hard. Here is a design for not handing over long-lived tokens, but issuing them per run and expiring them quickly.
Making My Managed Agents Batch Survive a Crash Without Redoing Everything
Running a 200-item batch on the Managed Agents API kept torching tokens, because every mid-run failure restarted from item one. Here is the checkpoint-and-idempotency design I added so the batch resumes from where it died.
Before You Trust 'It's Fixed' — Make the Agent Confirm Your Live URL Actually Renders
The agent reported it had fixed the bug and deployed successfully, yet the production page was blank. To prevent the empty-body-with-200 trap, here is how to add a completion gate that makes Antigravity 2.0's Browser Sub-Agent open the live URL and confirm the main content selector is actually filled.
Designing Parallel Agent Changes So You Can Trace Them Later
Antigravity 2.0 became a control tower for many agents. Here is how to build an audit trail that lets you trace who changed what and why, designed from real operational failures.
Stopping an AI Agent from Skipping Quality Checks — A Two-Layer Push Gate with Antigravity CLI Hooks and git pre-push
An agent once judged my failing tests 'unrelated' and pushed anyway. Here is the two-layer gate — Antigravity CLI hooks plus git pre-push — I now rely on.
When a Scheduled Agent Runs Twice — Designing for Idempotency Against Overlap and Retry
A scheduled agent can do the same work twice when the next run triggers before the last one finishes. Here is a design with an overlap lock and an idempotency guard that survives mid-run failures, drawn from a double-publish incident I ran into in production.
Instruction Drift in Scheduled Agents — A Three-Layer Design for Keeping Definitions, Docs, and Reality Aligned
Scheduled agents keep logging success even after their instructions diverge from reality. Here is the three-layer drift-detection design — definition, documentation, reality — I built after silent failures in my own operations.
Running Multiple Agents on One Repo Breaks It — Isolating Work Areas with Worktrees
When you run several Antigravity 2.0 agents against the same repository, you hit index.lock collisions and half-staged commits. Here is an isolation design using git worktree and projects that gives each agent its own work area, drawn from an incident I ran into in production.
Building Idempotent Scheduled Agents with the Antigravity SDK
Scheduling an Antigravity SDK agent is almost a one-liner. The hard part is making it idempotent — so a double trigger never runs the job twice, a crash never drops a day, and the result always converges to one. Here is how I build idempotent scheduled agents, learned from the maintenance jobs I run as an indie developer.
When an AI Agent's git push Reports Success but Nothing Reaches the Remote
Why agent-automated git pushes fail silently (a missing identity plus a no-op push), with three fixes: explicit config, SHA verification, and the GitHub REST API.
Handing Dependency Updates to Antigravity Agents — Risk Tiers, Verification, and Rollback
How far can you trust Antigravity agents with dependency updates? A four-tier risk model that corrects semver optimism, worktree-isolated lots, a fixed verification script, and a rollback-first ledger — the operations design I settled on while maintaining multiple apps.
Running Gemini's Managed Agents API: Where Cloud Execution Ends and My Local Agents Begin
A hands-on record of launching Gemini's Managed Agents (public preview) from Python — polling, artifact retrieval, and a cost guard — plus five criteria I use to decide what stays on my local CLI agents.
Size Antigravity Agent Tasks by What You Can Review — a Practical Cure for Rework
A one-line request cost me forty minutes of agent time and a Monday rewrite. Here is the sizing rule I switched to — tasks I can review in fifteen minutes — with the three actual briefs, acceptance criteria phrasing, and file-boundary rules for worktree parallelism.
Scheduling an autonomous agent fleet to run 6 sites solo — a timetable that avoids collisions and spam flags
A real example of the autonomous-agent scheduling I built to run 6 sites and an app business in parallel, solo: off-peak distribution, a daily generation cap, and collision avoidance, drawn from the actual timetable and the reasoning behind it.
Choosing the Right Granularity for an Agent's Tools — Bundle or Split?
Should you split an agent's tools finely or bundle them coarsely? A single decision rule, an intentionally asymmetric two-tier design for destructive actions, and real numbers from running six apps.
Delegate the Undoable, Guard the Irreversible — Tiering Agent Autonomy by Reversibility
When you hand production work to an Antigravity agent, the thing that bites first isn't intelligence — it's whether the operation can be undone. Here is a design that sorts every operation into three reversibility tiers and routes each to auto-execution, checkpointed execution, or a human gate, with TypeScript implementations and real numbers from running six apps in parallel.
Keeping Secrets Out of Your Antigravity Agent's Output: Layered Defenses for Logs, Diffs, and PR Bodies
The three paths through which background agents leak secrets, and how to defend commit diffs, execution logs, and PR bodies with layered protection, drawn from running six apps and measured false-positive rates.
Teaching Antigravity's Agent to Keep Diffs Small: How a Month Changed the Way I Review
When an agent hands you one giant diff, a solo developer's review can't keep up. After a month of asking Antigravity to break work into small steps first, here's how the experience of reviewing changed — with the exact instructions I now use.
Rehearsing an Agent's Actions Before They Touch Production — Designing a Zero-Side-Effect Dry-Run Layer
Some accidents survive shadow mode and canaries: the very first time an agent touches an external API. This is the design and TypeScript implementation of a zero-side-effect dry-run layer you can bolt onto Antigravity's parallel agents, with the real numbers from running six sites autonomously.