In partnership with
“AI is Going to Fundamentally Change…Everything”
That’s what NVIDIA CEO Jensen Huang just said about the AI boom, even calling it “the largest infrastructure buildout in human history.”
NVIDIA’s chips made this real-time revolution possible, but now it’s collaborating with Miso to unlock amazing new advances in robotics
Already a first-mover in the $1T fast-food industry, Miso’s AI-powered Flippy Fry Station robots have worked 200K+ hours for leading brands like White Castle, just surpassing 5M+ baskets of fried food.
And this latest NVIDIA collaboration unlocks up to 35% faster performance for Miso’s robots, which can cook perfect fried foods 24/7. In an industry experiencing 144% labor turnover, where speed is key, those gains can be game-changing.
There are 100K+ US fast-food locations in desperate need, a $4B/year revenue opportunity for Miso. And you can become an early-stage Miso shareholder today. Hurry to unlock up to 7% bonus stock.
This is a paid advertisement for Miso Robotics’ Regulation A offering. Please read the offering circular at invest.misorobotics.com.
|
ResearchAudio.io Anthropic Shipped What OpenClaw Couldn't SecureThree features this month. Several startup categories may not survive them. |
|
As of this week, you can send a Telegram message from your phone to a running Claude Code session on your workstation, and Claude will read your files, execute tasks against your codebase, and reply back through the same chat. No SSH tunnels. No VPNs. No ports exposed to the public internet. This is one of three features Anthropic shipped for Claude Code in March 2026. Together, they form something that looks less like a product update and more like a platform consolidation: absorbing capabilities that previously required third-party tools, open-source projects, or entire startups to provide. |
|
Claude Code's March 2026 Triple Launch Three features that absorb what took startups months to build |
|||||||||||||||||||
|
|||||||||||||||||||
|
|||||||||||||||||||
|
Source: Claude Code Docs, CVE-2026-25253, Kaspersky, Cisco AI Threat Research | ResearchAudio.io |
What Anthropic actually shipped
The three features serve different roles but share a common thread: they move Claude Code from a synchronous, terminal-bound tool into something closer to an always-on autonomous agent. Each one addresses a gap that previously required external tooling.
1. Channels: your terminal now has a chat bridge
A Channel is an MCP (Model Context Protocol) server that pushes events into a running Claude Code session. According to Anthropic's documentation, channels can be one-way (forwarding CI notifications or monitoring webhooks) or two-way, where Claude reads an inbound message and replies through the same platform.
The research preview supports Telegram and Discord as pre-built plugins. The architecture is notable: the MCP server runs locally on your machine and polls the messaging platform's API. No inbound ports are opened, no webhook endpoints are publicly reachable. The security model has two layers. Every channel plugin maintains a sender allowlist (specifically paired, approved user IDs can push messages). Additionally, the --channels flag controls which servers are active per session.
The practical result: you can kick off a long refactor at your workstation, walk away, and send instructions from your phone via Telegram. Claude processes the request using your full local environment (files, tools, git, MCP servers) and replies back. For Team and Enterprise organizations, admins must explicitly enable channels through managed settings. During the preview, plugins on Anthropic's curated allowlist can register as channels.
|
Key Detail: Developers can build custom channels for any system using the Channels reference. Any MCP server that implements the channel protocol can push events into Claude Code. Slack, CI pipelines, PagerDuty: if it sends a webhook, it can become a channel. |
2. /loop: Claude Code becomes a background worker
The /loop command brings cron-style scheduling to Claude Code. Pass an interval and a prompt, and Claude creates a background job that fires repeatedly while the session stays open. According to Anthropic's documentation, a session can hold up to 50 scheduled tasks. Supported intervals use standard time units: seconds (rounded to nearest minute), minutes, hours, and days.
/loop 2h /review-pr 1234
# Checks the PR every 2 hours, auto-fixes build issues
/loop 5m check if the deployment finished
# Polls deployment status every 5 minutes
There are two flavors. CLI session-scoped tasks (via /loop) die when you exit. Desktop scheduled tasks are persistent: they survive restarts, run on a visual schedule, and fire as long as the desktop app is open. The underlying tools are CronCreate, CronDelete, and CronList.
Safety guardrails are deliberate. Recurring tasks automatically expire 3 days after creation (one final run, then self-delete). Developers can disable scheduling entirely with CLAUDE_CODE_DISABLE_CRON=1. All times are interpreted in local timezone.
|
Key Detail: Anthropic developer Thariq Shihipar demonstrated configuring Claude Code to check error logs every few hours and automatically open pull requests for fixable bugs, with no manual intervention. Claude Code creator Boris Cherny added workflows for overnight PR monitoring with self-fixing and automated morning Slack summaries. |
3. Release channels: enterprise-grade version control
Claude Code now provides two release channels: latest (the default, receiving features immediately) and stable (approximately one week old, with automatic filtering of problematic versions). Configure via /config or by setting autoUpdatesChannel in settings.json.
For enterprise deployments, organizations can enforce a consistent channel across teams using managed settings. They can also pin specific versions or disable the auto-updater entirely with DISABLE_AUTOUPDATER=1. The /doctor command shows available versions across both channels.
The OpenClaw comparison is unavoidable
OpenClaw (formerly Clawdbot, Moltbot) is the open-source AI agent with 257,000+ GitHub stars that lets users control an AI assistant through WhatsApp, Telegram, Slack, and other messaging platforms. It went viral in January 2026, and its remote messaging capability is the exact use case that Claude Code Channels now addresses natively.
The security contrast is stark. In early 2026, security researchers discovered CVE-2026-25253, a critical remote code execution vulnerability in OpenClaw with a CVSS score of 8.8. The flaw exploited a WebSocket origin header bypass, allowing attackers to execute arbitrary code on exposed instances. At disclosure, researchers found over 135,000 exposed OpenClaw instances on the public internet, with more than 50,000 directly vulnerable to this exploit.
A security audit of ClawHub (OpenClaw's community skills registry) found that approximately 12% of skills contained malicious code, including data exfiltration payloads and prompt injection attacks. Kaspersky documented the risks in detail, and Cisco's AI Threat Research team published findings showing that OpenClaw's broad system permissions (file access, shell commands, browser control, email) create a wide attack surface. Palo Alto Networks described OpenClaw as "the potential biggest insider threat of 2026."
|
8.8
CVSS Score (RCE)
|
135K+
Exposed Instances
|
~12%
Malicious ClawHub Skills
|
0
Claude Code Exposed Ports
|
Claude Code's Channels deliver the same cross-device messaging capability that made OpenClaw compelling, built natively with Anthropic's security model and no external server requirement. The sender allowlist, curated plugin allowlist, and zero inbound port architecture make it a fundamentally different security posture.
The startup categories under pressure
These three features collectively absorb functionality from multiple product categories. Remote AI agent control (what OpenClaw popularized) is now native to Claude Code via Channels. AI task scheduling (products like runCLAUDErun, a dedicated macOS app for scheduling Claude Code tasks) is absorbed by the /loop command and Desktop scheduled tasks. Version management tooling for AI coding assistants is handled by release channels with enterprise managed settings.
The pattern is consistent with how platform companies consolidate: wait for the ecosystem to prove the category, then build the capability natively with tighter integration, better security, and zero infrastructure overhead. The third-party tools still work, and OpenClaw serves use cases beyond coding (email, calendar, smart home). But for developers already paying for Claude Code, the value proposition of external scheduling tools and remote control wrappers narrows considerably this month.
|
Insight 1: Channels are built on MCP, not a proprietary protocol. This means any system that sends webhooks (CI, monitoring, incident management) can push events into a Claude Code session. The Telegram and Discord plugins are starting points, not the ceiling. |
|
Insight 2: The 3-day auto-expiry on /loop tasks is a deliberate safety boundary. Anthropic is granting autonomy incrementally: enough to automate overnight PR monitoring, not enough to create unbounded processes that run indefinitely. Developers who need persistent scheduling are pointed to Desktop scheduled tasks or GitHub Actions. |
|
Insight 3: The stable release channel solves a real production pain point. Claude Code ships updates frequently (sometimes multiple times per week). For teams running Claude Code in CI pipelines or production workflows, the ability to stay on a version that is approximately one week old and automatically skips major regressions reduces operational risk. |
The question for third-party tool builders is no longer whether Anthropic will build these capabilities, but how quickly. For developers, the calculus is simpler: three features that shipped this month replace tools that previously required separate installations, configurations, and security audits. The code runs on your machine, the messages flow through your approved channels, and the version you run is under your control.
|
ResearchAudio.io Sources: Claude Code Channels | Scheduled Tasks | Release Channels Security: Kaspersky | Cisco AI Threat Research | CVE-2026-25253 |
