Fable was unavailable as a selectable model in both claude.ai and Claude Code for roughly 30 minutes. The issue is resolved; users may need to restart Claude Code and reselect Fable via /model if it no longer shows as the default.
Moving up the AI adoption curve requires giving Claude end-to-end self-verification, enabling auto mode for permissions, defaulting on automated code and security review, and using multi-agent interfaces (Agent view in CLI, Desktop, iOS, Android, Tag). Higher levels unlock via /loop, /batch, dynamic workflows, and worktree isolation for subagents. The point isn't any single feature but the combination of the right features with the right guardrails to automate entire classes of work teams can trust.
Claude Code team member trq212 argues that building quick prototypes of mockups, schemas, data models, and proofs of concept is the most efficient approach to development, letting you validate direction before burning tokens on output you might not want.
Users who had extra usage turned on during the ~30-minute Fable model outage may have been overcharged. Affected accounts are being refunded the full amount, plus an additional credit equal to the overcharge.
A common pattern keeps emerging: one engineer is dramatically outperforming with Claude while the rest of the org hasn't caught up. bcherny mapped out four stages of AI adoption teams tend to move through, linking to a full breakdown.
Each stage of AI adoption requires breaking down specific bottlenecks and building new guardrails. Spending more tokens without addressing those constraints won't move a team forward.
Usage dashboards track activity, not value. The better measurement question is whether you would have done this engineering work anyway, and what it would have cost in manual hours. The real payoff arrives when maintenance happens in the background and teams can focus on building things that were previously out of reach. bcherny says Anthropic is at step 3, pushing toward 4; he personally just hit level 4.
Tracking AI ROI means asking whether the engineering work would have happened anyway and estimating what it would have cost in manual hours. That gap is your return.
Asked whether Claude Code can handle its own setup configuration, bcherny confirmed it can. Just ask it to.
A user expressed frustration with how Anthropic communicates and handles feedback, saying their enthusiasm for the company has faded. bcherny asked what's going on and said he's listening.
A user reported that a page isn't scrollable on Safari for Mac. bcherny is looking into it and shared a doc link as a temporary workaround.
A user reported that a link redirects to the Claude homepage instead of loading correctly on Android. bcherny is investigating and shared a doc link in the meantime.
Starting July 20, Claude Fable 5 will be included in all Max and Team Premium plans at 50% of limits. Pro and Team Standard users keep access via usage credits and will receive a one-time $100 credit to offset demand constraints.
Platform leads sat down to discuss what shipped over the past six months in Claude's agentic API surface and what patterns are emerging in production use. The conversation covers the new APIs built to help developers quickly build and deploy agents.
Claude Managed Agents now supports agent-to-agent communication, letting multiple agents coordinate and delegate tasks. Agents can run different models, prompts, and tools while sharing sandboxes or vault credentials.
Claude Managed Agents introduces an outcomes mode for long-horizon autonomous work: give Claude a rubric and it loops until the rubric is satisfied, self-correcting along the way.
