Logo
FrontierNews.ai

Anthropic's Code with Claude Event Is About to Reshape How Developers Build AI Agents

Anthropic is hosting a major developer conference on May 6 in San Francisco, and the company is widely expected to announce several significant product updates that could change how engineers build and deploy AI agents. Based on internal signals and Anthropic's historical launch patterns, the keynote will likely include a new Claude Sonnet model release, persistent agent tooling, and announcements about expanded access to specialized Claude variants.

What Are the Five Most Likely Announcements at Code with Claude?

Anthropic has established a predictable rhythm for developer events: the first 30 minutes of the keynote focuses on product news, followed by platform demonstrations and partner announcements. Based on recent engineering signals and release schedules, five specific launches are on the watchlist.

  • Claude Sonnet 4.8 Release: Release date trackers have pointed to a May 2 to May 15 window for the past two weeks, making Tuesday's keynote the most likely venue. The new version is expected to show improvements in coding-agent tasks, tool-use accuracy, and long-horizon task completion, with pricing likely holding at $3 per million input tokens and $15 per million output tokens.
  • KAIROS Persistent Agents Demo: Persistent-agent tooling has appeared in Claude Code's package metadata under the codename "KAIROS" for the past two weeks. A public demo would signal that agents can maintain state across multiple sessions and resume long-running tasks on new conversations, a capability that would significantly expand what developers can build.
  • Project Glasswing Partner Expansion: Following recent news that the Trump administration opposed Anthropic's plan to expand Project Glasswing to roughly 70 partner organizations, the company is likely to address this on stage, either confirming the expansion path or naming new partners. Glasswing provides access to specialized Claude variants at higher pricing tiers ($25 to $125 per million tokens).
  • Cowork Mode General Availability: The multi-agent coordination feature has been in beta for months, and developer conferences are the standard venue for moving features from beta to production. A general availability announcement would make the feature available to all Claude Code users immediately.
  • Claude Code Version 2.2 Update: A minor version bump is highly likely given the pace of recent point releases. Expected improvements include better session resumption, improved long-context handling, an updated skills command interface, or enhancements to Model Context Protocol (MCP) integrations.

Why Should Developers Care About These Announcements?

Each of these launches addresses specific pain points in how developers currently build with Claude. A new Sonnet release typically brings measurable improvements in coding accuracy and reasoning, which directly affects the quality of AI agents that developers can deploy to production. The KAIROS persistent agents feature would eliminate the need to restart agent tasks from scratch when conversations end, a major limitation for long-running automation workflows.

The Glasswing expansion matters because it signals Anthropic's commitment to serving enterprise and government customers with specialized model variants, which could influence how organizations choose between Claude and competing AI platforms. Cowork mode moving to general availability would make multi-agent coordination available to all developers, not just beta testers, potentially unlocking new use cases for complex workflows that require multiple agents working together.

How to Prepare for the Code with Claude Livestream

If you're planning to watch the keynote live or follow along on social media, preparation matters. The event runs for 90 minutes, with the first 30 minutes dedicated to product announcements. Here's how to stay organized during the high-velocity news window.

  • Set Up Your Live-Blog Rig: Configure a screenshot hotkey for fast clipboard capture (Cmd+Shift+5 on macOS), open the Anthropic blog in a separate browser window ready to refresh, and draft posts in a notes file before publishing. Don't trust live composition during high-news moments; pre-draft and publish instead.
  • Plan Your Social Media Strategy: Post the agenda and livestream link 30 minutes before the keynote starts. During the first 30 minutes of product announcements, plan for one screenshot every 60 to 90 seconds. Pre-write a "headlines so far" comment that you update every five minutes. During the slower partner and demo window (minutes 30 to 90), three to five posts per ten minutes is sufficient.
  • Choose Your Session Focus: Three parallel tracks will run simultaneously covering Claude Code workflows, agent skills, and MCP-server patterns. If you plan to live-blog the sessions, pick one track and go deep rather than trying to cover all three at once, which produces shallow coverage.
  • Know the Timeline: The keynote runs 90 minutes. All five sessions will be recapped on the Anthropic blog within 48 hours, and the official YouTube upload typically follows within five to seven days. If you can't watch live, waiting for the blog recap Tuesday evening is a viable alternative.
  • Decide Your Depth Level: Solo developers should watch the keynote and skip the parallel sessions, then read about whichever launch matters most to their stack in depth on Wednesday. Engineering managers and tech leads should skip the live stream entirely, wait for the Anthropic blog recap Tuesday evening, then send their team a "what changed for us" message on Wednesday.

The London follow-on event is scheduled for May 19, with Tokyo's date confirmed on the announcement page but not yet publicly detailed.

What Should You Do If Claude Sonnet 4.8 Launches?

If the new Sonnet release lands as expected, the smart approach is to avoid immediately swapping production traffic to the new model. New Sonnet releases typically have a 24 to 48 hour stability window where edge cases surface. The recommended workflow is to pin your existing model in production for the first three days, run your evaluation tests against version 4.8 in parallel, then gradually ramp up traffic to the new version once you've confirmed stability.

This cautious approach protects production systems while still allowing teams to benefit from performance improvements. Given that Anthropic has maintained consistent pricing since Sonnet 4.5 was released at a developer event last September, cost shouldn't be a barrier to upgrading once you've validated the new version's performance on your specific use cases.

The Code with Claude event represents a significant moment for the AI developer community. Whether you're building AI agents, managing engineering teams, or evaluating which AI platform to standardize on, the announcements expected Tuesday will likely influence those decisions for the next six months.