Code with Claude 2026: agents that learn from your feedback
What was announced?
According to reporting by InfoQ, Anthropic’s Code with Claude 2026 event on May 6 centered on managed agents. Three upgrades led the list: multi-agent orchestration, an Outcomes feature for defining the success criteria an agent works toward, and a research preview called Dreaming, in which agents improve themselves using their own session history.
Claude Code, Anthropic’s developer tool, gained code review, remote agents, CI auto-fix, and routines. Rate limits doubled.
What does it mean for personal, on-device writing?
On the surface, nothing here is about writing. This is developer tooling. The signal is in the shape of the features.
Outcomes lets you define what success looks like, so the agent steers by your criteria instead of generic ones. Dreaming lets the tool improve from its own history with you. Code review keeps a human correcting the output. Read together, the developer world’s message is clear: a serious AI tool should get better the longer you use it, and your corrections are the training signal.
Expectations migrate. A developer who watches a coding agent absorb the team’s conventions will eventually ask why a writing tool resets to a generic voice every morning. Feedback-trained tools are becoming the norm, not the novelty.
Where does ILURA stand?
This is developer-tool news, not writing-app news. There is no feature gap for ILURA to close, and nothing in these announcements changes how ILURA works.
The relevance is the pattern. ILURA runs the same loop at personal scale. You set the criteria by choosing a role playbook, because how you write to a client is not how you write to your manager. You correct a draft, and the correction is stored as a learned preference inside that playbook. The next draft starts closer to you. That is training by correction — the consumer-sized version of what Anthropic is selling to engineering teams.
One difference is worth naming. Managed agents learn inside a vendor’s infrastructure, attached to a workspace and an account. ILURA’s learning lives on your iPhone, processed on-device by Apple Intelligence, with no account behind it. The same loop, with the memory in your pocket instead of a datacenter.
Read the signal through ILURA
Platform news matters when it changes what users expect from personal AI. ILURA reads these shifts through one lens: private agents trained by the user on iPhone.
- What becomes possible?
- What should stay user-controlled?
- What belongs on device?
Try it now
Put this to work on a real message.
Open ILURA, bring in a message you actually need to handle today, and get it done in your voice — free, on device, no account. It learns the preference, so the behavior carries to the next one.
Free to start · No account · Data Not CollectedQuick answers
- What did Anthropic announce at Code with Claude 2026?
- Per InfoQ's coverage, Anthropic announced managed-agent upgrades including multi-agent orchestration, Outcomes success criteria, and a Dreaming research preview, plus Claude Code features like code review, remote agents, CI auto-fix, and routines, alongside doubled rate limits.
- Does Code with Claude 2026 affect ILURA?
- No. This is developer-tool news, and ILURA is a personal writing app. The connection is the pattern: feedback-trained tools are becoming normal, and training by correction is exactly how ILURA learns your voice.