Qwen Code released two small versions in a row, focusing on "smarter proxies, smoother editing, and more stable integration". Add sub-agents and Todo Write, and see the "Welcome Back" project summary when restarting, and customize the caching policy. It also brings fewer retries, better login processing, terminal stress testing, optimized read/write for large projects, enhanced IDE+shell integration, improved MCP and OAuth, upgraded memory and session management, and multilingual documentation.
1. Key functions and experience upgrades
1. In terms of productivity function, the new
sub-agent breaks down the big task into small goals that can be paralleled; Todo Write records tasks and progress, forming a closed loop of "planning-execution-review"; Open the project again to see the "Welcome Back" summary; The cache policy can be customized to balance stability and freshness.
2. Performance and development experience
Editing is smoother and avoids proxy loops; Built-in terminal can do pressure testing; fewer retries and more stable logins; Large repository files are read faster; IDE is more tightly integrated with the shell; MCP and OAuth support is improved; More reliable memory and session management; Multilingual documentation is updated in place.
2. Three steps to implement your team process
1. Get started as soon
as you start using the"Welcome Back" summary to review yesterday's context, pull up the Todo Write list, split it into code, test, script and other subtasks by sub-agent, and clarify the acceptance criteria and submission nodes.
2. Acceleration of large projects
Enable more aggressive caching policies for dependencies and set short-term caching in modules with frequent changes. Integrate with the IDE with optimized file reading to reduce stuttering and unnecessary scanning.
3. Observable and steady state
Run an end-to-end link with terminal stress test to record the suggestion acceptance rate, revocation rate, and time-consuming; Deposit failure samples and session snapshots into the knowledge base to form playable improvement clues.
3. Selection and usage suggestions
1. When to enable sub-agents
Give priority to complex tasks such as cross-file reconstruction, end-to-end scaffolding, and data pipeline construction. Minor patches to a single file can continue to use the master proxy to avoid additional routing costs.
2. Trade-offs of caching strategy
Theevaluation and regression stages are conservative to ensure reproducibility. the exploration and reconstruction stage is more aggressive, pursuing response speed; The critical path is set to "force refresh" switch.
3. Integration and permissions
IDE+shell+MCP unified minimum permissions, OAuth credentials are managed by environment; Set secondary confirmation and log trace for high-risk operations (write, delete, and publish).
Frequently Asked Questions (Q&A)
Q: How can sub-agents actually improve AI coding efficiency?
A: Sub-agents break down complex tasks into parallel sub-targets, each holding local context and tool permissions, reducing back and forth jumps and invalid attempts, and making the overall delivery more controllable.
Q: How is Todo Write used in team collaboration?
A: Break down the requirements into checkable items, bind the submission and test results, and combine them with the "Welcome Back" summary to achieve cross-day continuation to avoid context loss.
Q: What does customizable caching mean for large warehouses?
A: Different caching strategies can be set for hot directories and stable dependencies, taking into account speed and consistency. Change more module short caches, and rely on locked module long caches.
Q: What problems can be found in terminal stress testing?
A: It can expose time-consuming bottlenecks and error-prone points in long links, such as authentication jitter, slow file read/write, improper retry strategy, etc., and provide a quantitative basis for optimization.
