Search “GitHub”

English translation is being prepared.

inclusionAI open-sources the Ming-3.0-asr-tiny speech recognition model repository

inclusionAI has open-sourced the Ming-3.0-asr-tiny speech recognition model repository on GitHub. The model combines the Ling-tiny-v3 language model with an audio encoder, supporting Chinese-English transcription, dialect recognition, and Context ASR. Dialect evaluation covers 15 test sets, including Hunan, Minnan, Cantonese, and Sichuan-Chongqing; Context ASR can recognize proper nouns through hotwords in prompts or prior conversation. The inference stack uses Transformers 5.12.1, fla-core 0.5.2, and PyTorch 2.11.0, and provides an SGLang accelerated deployment example. There are currently no subsequent development facts.

The SGLang repository introduces its open-source inference framework

Ant inclusionAI has launched the sglang_ling_v3 repository on GitHub, whose contents are based on the open-source inference framework SGLang. The repository's README introduces that SGLang targets large language models, vision-language models, and diffusion models, and is optimized for agent workloads, large-scale serving, and RL rollout. The README also shows that SGLang supports GPU, TPU, NPU, CPU, and Apple Silicon, and includes a built-in image and video generation engine, SGLang Diffusion. The repository has currently been released, with no further development activity yet.

Zihao launches a Rime engine-based iPhone input method on the App Store

On October 20, 2026, the author released Zihao, an iPhone input method. The input method is now available on the App Store, is based on the Rime (Zhongzhou Yun) engine, and comes with Wusong Pinyin built in. According to a report by Shaoshupai the same day, its basic typing features are free and ad-free, and typed content is processed only on the phone; it supports importing Rime configuration folders, ZIP files, YAML files, or GitHub repositories and updating them automatically, and the lexicon can be synced to iCloud Drive in Rime format and merged with Mac Squirrel. The author says that in simulated tests, the one-pass accuracy rate for sentences with 2 wrong letters rose from 36% to 56%, and for 3 wrong letters from 26% to 53%. There are currently no subsequent developments.

heise online launches three-day hands-on RAG workshop

heise online's iX is launching a three-day online interactive workshop, "Implementing AI Correctly with Your Own Data: Evaluating and Applying RAG, Embeddings and Retrieval Architectures," which explains RAG architectures and technical foundations and covers both cloud and on-premise deployment. The course is taught by Gregor Biswanger, a Microsoft Azure AI MVP and GitHub Copilot collaborator, and covers prompt optimization, integrating the OpenAI and Azure OpenAI APIs, and developing and testing self-built AI agents. The workshop runs online daily from 09:00 to 17:00 and is aimed at developers who want to get started with RAG systems; it is suitable for both beginners and advanced participants. There are currently no subsequent developments.

Tessl Code Review supports reading the full PR and publishing reviews

Tessl Code Review now supports reading the full pull request rather than only viewing an isolated diff, and publishes a review result based on it. According to a tutorial published on the Tessl product and engineering blog on October 19, 2026, users can set up the agentic code review for a GitHub repository in about 5 minutes: it scores against the standards the team defines in the repository and publishes the review result. Reviews can be set to run only when a PR is opened or requested, or after every update; the handling can be comment-only, or approve/request changes, and supports automatic approval once all Tessl discussion threads are resolved. The tutorial also notes that the relevant changes only take effect for new review runs. There are currently no subsequent developments.

Author uses Codex and DeepSeek-V4.1-Flash-1017 to generate a four-quadrant sticky note app

On October 25, 2026, an author used the Chinese-localized version of Codex, DeepSeek-V4.1-Flash-1017, and the Hermes agent to generate a four-quadrant sticky note app from a single prompt, and automatically uploaded and published it to GitHub. According to the same report, the finished product supports Chinese and English, a time capsule, free resizing of sticky notes, and local saving, and is packaged as a single-file portable app; the first version generated a static index.html in about six to seven minutes, and subsequent iterations changed it to a sticky note format, with the whole process costing a little over four yuan. There are currently no subsequent independent development facts.

Cursor releases Agent SDK

On October 20, 2026, Cursor released the Agent SDK, the framework it uses to build hosted agents such as Bugbot, Security Reviewer, and Rollouts. Developers can define agents in code and deploy them to Cursor infrastructure with a single command, agent-sdk deploy. Agents are organized as small software projects: prompts are written in Markdown, tools are defined as TypeScript functions, and agents can respond to PR events from GitHub, GitLab, and Bitbucket, with built-in evals and event replay; they can also be packaged to run as a single Node process on your own infrastructure. There have been no further developments so far.

Microsoft launches security controls for out-of-control AI agents

On October 19, 2026, Microsoft launched security controls for out-of-control AI agents. According to SiliconANGLE, the feature progressively reduces an agent's permissions when its risk score rises: the Discovery tool can identify 60 categories of local agents on employee devices, including Claude Code, Codex, and GitHub Copilot, and assigns each agent a separate identity through Entra Agent ID; Dynamic Agent Risk continuously updates the risk score based on identity, endpoint, cloud, and network signals, and Dynamic Conditional Access uses it to first revoke write permissions, then require human approval, and at the highest level directly isolate or stop the runtime. These features are grouped under Project Perception and Microsoft 365 E7, and Microsoft has not announced availability or pricing. On the same day, Satya Nadella said that he had previously written an article arguing that models should be treated as an internal risk and that security controls should be placed outside the model, and that today Microsoft is putting this principle into practice by extending Project Perception, adding capabilities to continuously assess agent risk, adaptively adjust permissions, and contain threats.

Ant Ling Open-Sources the Ming-3.0 Speech Series, Five Models in Total

On October 20, 2026, Ant Ling (inclusionAI) open-sourced the Ming-3.0 speech series on GitHub, offering three series—speech synthesis, audio understanding and dialogue, and speech recognition—for a total of five models. Among them, Ming-3.0-tts includes a 0.5B dense version and a tiny MoE version based on Ling V3 with 7.9B total parameters and 1.3B activated parameters, supporting voice cloning, text-designed timbres, and multilingual synthesis; Ming-3.0-audio-flash uses a 126B-parameter MoE architecture with 6.5B activated parameters, and there is also a tiny version with 9.3B parameters and 2.7B activated parameters, covering multi-dialect accent recognition, long audio understanding, and end-to-end speech dialogue, under the MIT license. On the same day, vLLM announced that vLLM-Omni would provide support on the day of release, and Ant Ling thanked it and said deployment would be more convenient; atomic.chat will also integrate Ming-3.0 into local AI workflows. Artificial Analysis published an evaluation stating that Ming-3.0-tts ranks first among open-weight models in the controlled speech TTS Arena, and its pronunciation robustness of 93.2% ties for first among 31 models. On October 22, a blogger said they planned to trial-run the 0.5B version locally; on October 23, Artificial Analysis said it achieved top-tier TTS quality with 0.5B parameters.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

OpenAI plans to release hundreds of mathematical results

On October 7, 2026, WIRED reported that OpenAI had told relevant parties it planned to release solutions to hundreds of unsolved mathematical problems all at once on GitHub on Tuesday, but the release time had not yet been determined at that point. Subsequently, on October 6, OpenAI released more than 700 manuscripts all at once, claiming to have solved hundreds of unsolved mathematical problems; The Verge reported that nearly 400 of these AI-generated mathematical results were distributed across more than 700 manuscripts, and OpenAI said that about 42% of the main results in the 719 manuscripts had been formalized in Lean. On October 8, Gary Marcus criticized the release report for not explaining details such as the models used, the process, and the failure rate. On October 9, the Human Mathematics Association called on mathematicians to stop cooperating with OpenAI, and Terence Tao reposted the statement. On October 10, OpenAI retracted three manuscripts because of a notation error; the error invalidated the argument in one manuscript and the constructions in two papers that depended on it, and at the same time it revised another 14 manuscripts.

Birla AI Labs releases time series foundation model Yug-1

Birla AI Labs has released the time series foundation model Yug-1. The model is used to predict outcomes, identify patterns, and detect anomalies in interconnected datasets. According to Analytics India Magazine, Yug-1 outperforms other first-generation TSMFs on the GiftEval benchmark and is also competitive with newer-generation models, taking less than six months from training to deployment. Yug-1 has been made available to developers and researchers for testing on GitHub and Hugging Face. There are currently no subsequent development facts.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.