Technical posts from the FutureSearch team.
We built a pipeline that scans 18 community sources every morning, classifies opportunities with a 13-question rubric, and drafts responses. 2-3% signal rate. The hard part isn't running it - it's knowing what to look for.
Docker-based MCP servers leave behind zombie containers because the Docker daemon keeps them alive after Claude Code exits. Switching from docker run to uvx eliminates the problem entirely.
We publish the 300-line Claude Code review skill we use in production, covering five key rules: trusting your types, deferring to linters, avoiding over-engineering, skipping backwards compatibility, and encoding domain knowledge.
We built a Claude Code skill that reviews our AI agent traces and catches issues we'd miss ourselves. Here's how it works, and why it only became possible now.
Anthropic's new effort parameter for Claude 4.6 controls more than reasoning depth. Unlike thinking_level (Gemini) or reasoning.effort (OpenAI), which primarily set an upper bound on how much the model thinks, effort also governs how lazily the model behaves in general.
We run Claude Code in Kubernetes for long-running marketing CronJobs. This originally sounded like a terrible idea, but after running it for a few months, we think it's a genuinely valid engineering approach - for the right kind of work.
Instead of traditional orchestration platforms, some of our pipelines are markdown files that Claude Code executes directly. Skills define the phases, subagents do the work, and results are plain old files stored in GitHub.
Instead of dumping thousands of rows into the MCP tool response, split the audience: content for the model (text summary), structuredContent for the user (interactive widget at zero token cost), and a download URL for the sandbox.
OpenAI's responses and chat/completions API endpoints have differences that make no logical sense.
Inlining data in MCP tool calls eats the LLM's context window. We show how to use presigned URLs so Claude can upload files directly to your server, keeping the context clean with a 36-character artifact ID.
LLM APIs look interchangeable on paper. In practice, they diverge in subtle ways that break your code. We document the provider-specific quirks we've hit while running thousands of LLM calls per day across Anthropic, Google, and OpenAI.