The VP Geek Speaks

Vercel's Plugin for Coding Agents: Deployment Knowledge as Infrastructure

Vercel’s plugin for coding agents is one of those releases that sounds incremental until you notice what it is really doing: it turns deployment and edge-platform expertise into structured, reusable context that agents can invoke reliably.

According to Vercel’s announcement, the plugin bundles broad platform coverage (on the order of 47+ skills) and includes specialist agents aimed at deployment optimization scenarios, so agents are not guessing their way through framework-specific hosting details every time.

Trace Every Copilot Agent Commit Back to Its Session Logs

Agent-generated commits used to arrive like any other push: you saw the diff, but not the reasoning, tool calls, or missteps that produced it. In March 2026, GitHub tightened that story.

Copilot coding agent commits can now include an Agent-Logs-Url trailer that points reviewers back to the full session logs for that change. GitHub also highlighted live monitoring of Copilot coding agent logs through integrations such as Raycast.

This is not a flashy model upgrade. It is infrastructure for accountability.

GitHub Says Copilot's Coding Agent Starts Work 50% Faster. Here's Why That Changes the Math

In a March 2026 changelog update, GitHub reported that the Copilot coding agent starts work roughly 50% faster, with optimizations to the cloud-based development environments agents use to spin up and begin executing on a repository.

That sounds like a performance tweak. It is also a shift in how teams should think about agent economics.

Cold Start Was a Hidden Tax

For any agent that runs in an isolated or remote environment, time-to-first-action is not just latency. It is friction that shapes behavior:

The MCP Security Problem Is Really a Least-Privilege Problem

The most important security story around agent infrastructure right now is not a single CVE. It is the growing realization that MCP security is immature by default.

As of March 2026, reporting around MCP security points to more than 30 CVEs, thousands of publicly reachable servers with weak or no authentication, and a still-evolving roadmap for making the protocol more production-ready. One especially concerning figure: roughly 36% of observed MCP servers reportedly accept connections without meaningful authentication.

Why NIST's AI Agent Standards Initiative Matters Right Now

One of the most consequential AI stories this month is not a product launch. It is the NIST AI Agent Standards Initiative.

NIST launched the effort through its Center for AI Standards and Innovation to focus on security, interoperability, and identity for AI agents. The initiative is structured around three pillars: industry-led standards development, open protocol support, and security research. It already has concrete deadlines attached, including a March security request for input and an April identity concept paper.

Gemini CLI Conductor Turns Review into a Structured Report

Google’s automated review update for Gemini CLI Conductor is worth paying attention to for a simple reason: it treats AI review as a structured verification step, not as another free-form chat.

Conductor’s new review mode evaluates generated code across multiple explicit dimensions:

• code quality
• plan compliance
• style and guideline adherence
• test validation
• security review

The output is a categorized report by severity, with exact file references and a path to launch follow-up work. That is an important product choice.

Claude Code Review and the New Economics of Verification

Anthropic’s new Claude Code Review feature is one of the clearest signs yet that the economics of AI development are shifting from generation toward verification.

The March launch is aimed at Teams and Enterprise customers and uses multiple specialized review agents to examine pull requests in parallel, verify findings, and rank issues by severity. Anthropic says reviews typically take around 20 minutes, cost roughly $15-$25 per PR, and increased substantive feedback from 16% of PRs to 54% internally. For large pull requests over 1,000 lines, 84% reportedly received findings.

JetBrains Air and the Case for the Agent-Native IDE

JetBrains Air, launched in public preview in early March, is one of the more interesting answers yet to a question the AI tooling market keeps circling: do we really want agents bolted onto traditional editors, or do we eventually need environments designed around them from the start?

Air is betting on the second path.

Why Air Is Interesting

Most current AI coding tools still inherit the shape of the pre-AI IDE. There is a primary editor, maybe a chat pane, maybe an agent sidebar, and the user is still clearly the central operator of a mostly traditional workspace.

Cursor Automations and the Shift from Prompting to Policy

One of the most current product shifts in AI development tooling is Cursor Automations, which turns coding agents into event-driven workflows instead of one-off assistants. The feature can trigger work from commits, Slack messages, timers, and operational events, then route the agent through review, checks, and deployment-style steps with humans only stepping in at key points.

That may sound like just another convenience layer. It is not. It reflects a deeper change in how teams are thinking about AI tooling.

VS Code's New Agent Features Show What 'Practical' Actually Means

One of the better AI tooling posts of the month came from Microsoft itself: “Making agents practical for real-world development.” That framing is useful because it captures what the market is moving toward. The interesting releases are no longer just about whether an agent can generate code. They are about whether the agent can survive contact with a messy, real workflow.

VS Code 1.110 is a good example of that shift. The March release adds native browser control for agents, better session memory, context compaction for long conversations, installable agent extensions, and a real-time Agent Debug panel. None of those features are flashy in isolation. Together, they show what “practical” now means in agentic development.