✔️ Anthropic has launched a certification program for engineers.
Anthropic has introduced a new initiative — the Claude Certified Architect program, which includes a technical exam for developers and solution architects using Claude API, Claude Code, Agent SDK, and MCP.
The exam consists of 60 multiple-choice questions, with a two-hour time limit. Retakes are not permitted, and external materials cannot be used. Results are available within two business days and are presented across five areas: agent architecture, Claude Code tuning, prompt engineering, MCP integration, and context management.
Currently, certification is only available to employees of companies in Anthropic’s partner network. The first 5,000 participants can take the test for free; afterward, the cost per attempt will be $99. Successful candidates will receive a digital badge for professional platforms.
Program website: anthropic.skilljar.com
✔️ NVIDIA formed a coalition to develop open models.
NVIDIA announced the creation of the Nemotron Coalition — an alliance of various companies collaborating on open-source model development. Participants include Mistral AI, Black Forest Labs, Cursor, LangChain, Perplexity, Reflection AI, Sarvam, and Thinking Machines Lab.
The main idea is to pool efforts to develop a shared foundational model, which members can then fine-tune for specific tasks such as industry-specific needs, language nuances, or regional requirements. Each participant retains independence and continues developing their own products.
The first joint project will involve creating a base model by NVIDIA in partnership with Mistral AI. Other companies will provide data, expertise, and evaluation systems. The model will be trained on NVIDIA DGX Cloud and later released as part of the NVIDIA Nemotron 4 family.
Source: nvidianews.nvidia.com
✔️ Mistral released an open agent model for formal proofs based on Lean 4.
The project called Leanstral is the first open-source agent designed for working with formal proof systems in Lean 4. It is based on a general-purpose model with 119 billion parameters, with six billion active. Thanks to its sparse architecture, it operates quickly and cost-effectively.
The model was trained on real repositories of formal proofs rather than on Olympiad-style problems. On the FLTEval benchmark, it scored 26.3 points in two attempts — approximately 2.6 points higher than Claude Sonnet, at a cost of about $36 (compared to Claude’s $549). Although Claude Opus still outperforms with a score of 39.6, its cost is around $1650 — roughly 46 times more expensive.
The model is accessible via the Mistral Vibe platform, API, or on HuggingFace under the Apache 2.0 license.
Source: mistral.ai
✔️ Alibaba consolidated all its AI initiatives into a single unit.
The largest Chinese tech giant established a new division called Alibaba Token Hub, led by CEO Eddie Wu. This center combines projects like Qwen, consumer applications, the DingTalk enterprise messenger, and Quark devices.
The main goal of this new structure is to accelerate the transition from research to ready-to-use products and to monetize AI across the company. Soon, Alibaba plans to introduce an enterprise AI agent based on Qwen for business clients; it will be integrated with platforms like Taobao and Alipay.
Source: bloomberg.com
✔️ Britannica filed a lawsuit against OpenAI for copyright infringement.
Britannica, together with Merriam-Webster, has filed a lawsuit in Manhattan. The plaintiffs allege that OpenAI trained its models on over one hundred thousand encyclopedia articles and dictionary definitions without permission from rights holders. They claim GPT-4 memorized large portions of Britannica’s content and reproduces it nearly verbatim upon user request.
Additionally, Britannica accuses OpenAI of trademark infringement: it is alleged that ChatGPT creates a false impression of Britannica’s endorsement by referencing its materials in inaccurate responses. The company seeks damages and an injunction against using its data without authorization.
Created with n8n:
https://cutt.ly/n8n
Created with syllaby:
https://cutt.ly/syllaby