China Publishes Draft E-Commerce Law Amendments, Expanding Platform Regulation
China's State Administration for Market Regulation and the Ministry of Commerce have released draft amendments to the E-Commerce Law and are soliciting public comments. The drafts broaden the scope of application beyond platforms and businesses, overhauling rules around the digital economy. For platform responsibility, the amendments indicate plans to introduce new regulatory tools in addition to existing measures such as fixed fines and business suspensions.
China's tech regulation has unfolded along multiple fronts since 2020 — antitrust, data security, and more — but these amendments appear to dig deeper into the everyday economic activity of e-commerce. As AI technology becomes increasingly integrated into platforms, how much consideration will be given to algorithmic pricing and its impact on recommendation systems remains to be seen as the public comment period unfolds.
Argentina Plans to Require Human Involvement in AI-Run Companies
According to Reuters, Argentina is considering a framework that would allow companies to be run autonomously by AI, but humans are not expected to be excluded entirely. While embracing AI-driven business models, the country is moving toward requiring some form of human involvement to ensure clear legal accountability and oversight.
Legislation for AI-run companies has few precedents worldwide, and comparisons with Delaware corporate law and the UK's limited company law are drawing attention. Argentina's initiative is a notable example of an emerging economy attempting to lead corporate law for the AI era.
Midjourney Seeks Discovery to Uncover Studios' AI Usage
According to Variety, Midjourney is pursuing discovery (evidence disclosure) proceedings in its copyright lawsuit against major Hollywood studios to reveal the extent to which the studios themselves use AI tools. Midjourney is searching for evidence that the studios make use of AI-generated content with their own copyrighted works — something that could flip the offense and defense in the debate over copyright infringement in AI training data.
The case shows that generative AI companies are not limited to playing defense against lawsuits from the creative industries; they can also go on the offensive, probing the other side's actual AI usage. Depending on how the litigation unfolds, it could influence AI usage guidelines across the entire industry.
Using Qwen3.6-27b Locally for Game AI Development
A post on Reddit's LocalLLaMA community reported on implementing an A* pathfinding algorithm in a Java-based game using the Qwen3.6-27b-mtp-q8 model. The developer ran Claude Code locally, had the model build a test suite, and ran an "automated test marathon" for roughly 12 hours — automatically running the game while repeatedly monitoring logs, fixing code, and restarting in real time. In the end, the NPCs were able to climb over obstacles and chase targets while avoiding gaps.
It's a fascinating example showing that even a locally run 27b model can deliver practical code generation with proper loop design and iterative verification. That said, it comes with the caveat that it works "most of the time," so full reliability would require further validation.
IKEA's Use of AI Makes Employees 10x More Valuable
According to a video that gained traction on Hacker News, IKEA has boosted employee productivity tenfold by embedding AI tools into employee workflows. What's drawing attention is the approach of using AI not to "replace" employees but to make them "more valuable." As an AI adoption case in the retail and furniture industry, it appears to be delivering results in areas such as customer support, inventory management, and design proposals.
A Hands-On Report on Loop Engineering — and How It Differs from cron
An article titled "Trying Loop Engineering on a Real Work Task" was published on Qiita. Loop engineering is a technique of continuously feeding prompts to a coding agent so that it completes tasks autonomously. It's a concept that has drawn community attention since June 2026, sparked by a post from Peter Steinberger, the creator of OpenClaw.
The article sets up loop engineering on an actual work task and breaks down how it differs from traditional cron-based scheduled execution. While cron "runs commands at fixed times," the essential difference cited is that with loop engineering, "the agent carries out tasks continuously while retaining context, and recovers autonomously when errors occur."
AI Hiring Interviews: Toward a Design That "Supports" Rather Than "Replaces"
An article published on Qiita introduces an implementer's account of AI company クリスタルメソッド developing an AI assistant for hiring interviews. It summarizes why they chose a design that assists interviewers rather than replacing them entirely with AI, along with the insights gained along the way. When it comes to AI use cases, choosing "support" versus "replacement" looks set to remain an important design decision in engineering.