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 night my agent shipped nothing: giving generation agents an abstain outcome
When you score a background generation agent by how much it produces, the quality gate quietly loosens over time. Here is a three-valued ACCEPT / ABSTAIN / REJECT design that counts a zero-artifact run as a success, with the code and the measurements from running it.
Leasing Runtime Environments to Parallel Agents: Port Ranges, DB Schemas, and Dev Servers, One Per Agent
When Antigravity 2.0 runs agents truly in parallel, they collide on the same port, database, and dev server. Here is a lease-based design that hands each agent its own isolated environment, with implementation and measured numbers.
One Space in a Path, and Nine Commands Reported Success While Counting the Wrong Place
A single space in a workspace name sends agent-written commands somewhere else, quietly. Measurements across eleven unquoted-path forms, and the entry-point script that closes the boundary in one cd.
Your UI Is in Japanese. Your Commit Log Isn't.
Setting your editor's display language does nothing for the language your agents write into the repository. Here is what six weeks of unattended runs actually produced, how to pin output language per audience, and a gate that catches the drift mechanically.
Calling Local LLMs from Antigravity — Ollama and LM Studio Integration in Practice
Running local LLMs from Antigravity via Ollama or LM Studio: a real benchmark harness, how to confirm the model is actually on the GPU, a monthly breakeven model, and a wrapper that forces JSON output.
Localizing Antigravity 2.0 to Japanese — The Three-Layer Setup for Menus, Commands, and AI Responses
Installing the Japanese Language Pack in Antigravity 2.0 only translates part of the UI. To get a clean Japanese experience — menus, command palette, AI responses — you need to configure three independent layers. Here is the exact setup I run as an indie developer at Dolice, plus which parts I deliberately leave in English.
Is the $100 AI Ultra Tier Worth It Solo? Measure the Break-Even from Limits and Parallelism
Whether the $100/month AI Ultra tier (5x the Pro limit) is worth it for an indie developer, framed as a break-even from how often you hit the cap and the effective throughput of parallel agents, with a calculator script.
Running Antigravity Behind a Proxy
Keep Antigravity working behind corporate proxies, SSL inspection, and hotel Wi-Fi: layer-by-layer diagnosis, certificate trust, NO_PROXY design, and connection profiles with a local LLM fallback.
Cutting Down 'Plausible but Wrong' RAG Answers — A Retrieval Evaluation Harness for Gemma 4 and Antigravity
Replace gut feeling with recall@5, MRR and faithfulness scores — a 30-question golden dataset and a small Python harness for evaluating a local Gemma 4 RAG stack.
Capping Parallel Agents With a Token Budget — Designing a Guard That Stops Runaway Cost
Running many agents in parallel quietly inflates your token bill. This is not about shrinking prompts — it is about a governance layer that meters spend in real time and cuts it off at a budget. Full design and TypeScript implementation, drawn from my own overnight jobs.
Five Weeks Letting Antigravity's Background Agent Refresh App Store Screenshots Across Four Apps
An operations log from running Antigravity Background Agent for five weeks to keep App Store screenshots for four wallpaper apps in sync across eight locales — including the caption-width validation step that cut manual fixes from 18 to 2.