Independent journal of China AIPrimary sources / editorial judgment
China AI DeskNumbered issues / primary sources

Dated field log

Last checked 22 July 2026 03 signals

A 22 July checkpoint.
Not a current daily feed.

This edition preserves the 22 July 2026 checkpoint. It is not a claim about current availability or deadlines. Each brief separates the source date, the editorial check date, and the action recorded at that time.

01Build

Deadline · source dated 24 Apr · checked 22 Jul

At the 22 July check, two days remained before DeepSeek’s scheduled legacy-ID cutoff.

DeepSeek’s official API change log said deepseek-chat and deepseek-reasoner would be discontinued on 24 July 2026. During the documented transition, those names pointed to the non-thinking and thinking modes of deepseek-v4-flash.

This was an application-contract change, not merely a naming cleanup. The recorded action was to inspect hard-coded IDs, gateway aliases, fallback chains, monitoring labels, cost attribution, and tests that treated the two legacy names as separate model families. A request succeeding at that checkpoint did not prove that the application had migrated to the supported contract.

Action recordedMove explicit model selection and reasoning-mode assumptions onto the supported V4 contract.
DeepSeek API change log Read the full Issue 03 analysis
02Model

Release state · official page checked 22 Jul

At the 22 July check, Kimi K3 was usable through the API while full weights remained future-dated.

Moonshot’s official Kimi K3 page said the full model weights would be released by 27 July 2026. The same page provided the kimi-k3 API call and said access unlocked after a minimum one-dollar top-up.

The two facts supported different decisions at that checkpoint. Product teams could test response behavior, tool use, latency, pricing, and integration through the hosted service. Self-hosting teams did not yet have the released artifact set, final license terms, measured memory profile, serving support, and independent reproduction needed for a deployment judgment.

Action recordedEvaluate the hosted model; wait for artifacts and terms before sizing a self-hosted deployment.
Kimi K3 official documentation Read the later 27 July weight-state record in Issue 09
03Tools

Workflow · source dated 16 Jul · checked 22 Jul

Qwen Code’s 16 July update moved learned skills and verified review into the product layer.

Qwen Code’s 16 July update covers releases 0.19.9 through 0.19.11. It introduces a /learn command that turns a directory, URL, conversation, or text into a reusable skill; review suggestions can be posted as inline comments; and later review steps verify findings and reverse-audit the result.

The recorded editorial signal was larger than an isolated feature list. Competition was moving from model quality alone toward the reliability of the surrounding workflow: reusable knowledge, review discipline, multi-workspace operation, artifacts, recovery, and human approval.

Action recordedCompare review accuracy, learned-skill portability, and failure recovery—not only benchmark claims.
Qwen Code 16 July update Read the full Issue 04 analysis

Field-log method

Short does not mean source-light.

The dated field log records a source date, a check date, the reading at that checkpoint, and the action it changed. Long-form interpretation remains in numbered issues.