Artificial Intelligence

Apr 2, 2026

Japanese AI startup Sakana AI has unveiled "Sakana Marlin," its first product for business customers. The system works autonomously: give it a topic, and it researches on its own for up to eight hours, then delivers detailed reports and presentations. Sakana AI says the tool can produce professional strategy analyses that would normally take human teams several weeks.

Sample output from "Sakana Marlin": after autonomous research, the tool creates text reports and presentation slides on a given topic (here: AI trends in the financial sector). | Image: Sakana AI

Sakana Marlin combines the company's "AI Scientist," designed to resolve contradictions, with its previously introduced "AB-MCTS" method for strategic searches. Multiple AI models work together, and longer thinking time is meant to yield better results, the company says.

The company is looking for beta testers in finance, research, and business consulting. The beta is free, but requires registration (the form is in Japanese). The biggest weakness of automated reports like these is hard-to-spot AI errors, something the startup doesn't address in its announcement.

Comment Source: Sakana AI

Matthias Bastian

Apr 2, 2026

Short News

Microsoft has introduced MAI-Transcribe-1, a speech-to-text model supporting 25 languages that achieves the lowest word error rate of any model tested on the FLEURS benchmark, beating Scribe v2, Whisper-large-V3, GPT-Transcribe, and Gemini 3.1 Flash-Lite. The model is also built to handle tough recording conditions like background noise, poor audio quality, and overlapping speech, Microsoft says.

MAI-Transcribe-1 (green) leads in word error rate on the FLEURS benchmark in most of the 25 languages tested, outperforming Scribe v2, Gemini 3.1 Flash-Lite, Whisper-large-v3, and GPT-Transcribe. | Image: Microsoft

Microsoft is rolling out MAI-Transcribe-1 across Copilot Voice and Microsoft Teams. Developers can try it as a public preview through Microsoft Foundry and the Microsoft AI Playground. The model runs 2.5 times faster than Microsoft's previous Azure Fast offering and costs $0.36 per audio hour. Combined with MAI-Voice-1 and a language model, it can also power voice agents, Microsoft says.

Cohere and Mistral recently released open-source alternatives that perform at a similar level.

Comment Source: Microsoft

Maximilian Schreiner

Apr 2, 2026

AI in practice

Nvidia sets new MLPerf records with 288 GPUs while AMD and Intel focus on different battles

Maximilian Schreiner

Apr 2, 2026

Short News

Alibaba has released Qwen3.6-Plus, its third proprietary AI model in just a few days. The model is available through the Alibaba Cloud Model Studio API and offers a context window of one million tokens. According to the Qwen team, the focus is on significantly improved capabilities for agentic coding, including frontend development and complex code tasks.

In benchmarks published by Alibaba, the model partially outperforms Anthropic's older flagship model Claude 4.5 Opus, which was replaced by the stronger 4.6 Opus in December 2025. It's worth noting that some of these measurements were conducted by Alibaba itself.

Das neue Qwen3.6-Plus übertrifft das ältere 3.5-Modell und in einigen Fällen auch Opus. Allerdings erreicht das im Dezember 2025 veröffentlichte Opus 4.6 etwa im Terminal-Bench 2.0 65,4 Prozent und liegt damit vor Qwen3.6-Plus. | Bild: Alibaba — Qwen3.6-Plus outperforms the older 3.5 model and in some cases beats Opus. However, the Opus 4.6 released in December 2025 scores 65.4 percent on Terminal-Bench 2.0, putting it ahead of Qwen3.6-Plus. | Image: Alibaba

For a long time, Alibaba released its Qwen models as open source, but the company has recently changed course. The latest Qwen3.5-Omni is also not freely available. Alibaba wants to drive more revenue from enterprise customers with its proprietary models, as its cloud division faces intense competition from ByteDance.

According to Bloomberg, Alibaba is targeting $100 billion in AI revenue over the next five years. Qwen3.6-Plus will be integrated into the Qwen chatbot app and the company's new enterprise AI service Wukong.

Comment Source: Qwen

Maximilian Schreiner

Apr 2, 2026

AI research

AI models fail at robot control without human-designed building blocks but agentic scaffolding closes the gap

Matthias Bastian

Apr 2, 2026

AI research

GPT reasoning models have "line of sight" to AGI, says OpenAI's Greg Brockman

Maximilian Schreiner

Apr 2, 2026

Short News

Chinese chipmakers captured nearly 41 percent of China's AI accelerator server market in 2025, according to an IDC report seen by Reuters. IDC is a global market research firm specializing in the technology industry.

Nvidia remains the market leader with roughly 2.2 million cards shipped and a 55 percent market share, but the company is losing ground fast. In total, about 4 million AI accelerator cards were shipped in China, according to the report.

Chinese vendors shipped a combined 1.65 million cards. Huawei leads the domestic pack with about 812,000 chips, followed by Alibaba's chip unit T-Head at 265,000 cards. Baidu Kunlunxin and Cambricon are tied at 116,000 units each. AMD held just 4 percent of the market.

The shift is driven by tightened US export controls and Beijing's push for companies to rely more heavily on domestic chips.

Comment Source: Reuters

Matthias Bastian

Apr 1, 2026

Short News

Following the accidental leak of its AI coding tool's source code, Anthropic has had more than "8,000 copies and adaptations of the raw Claude Code instructions" removed from GitHub via a copyright request, the Wall Street Journal reports. One programmer already used AI tools to rewrite the code in different languages, keeping it available despite takedowns. This shows just how damaging a code leak is in the age of AI: once it's out, it spreads faster than anyone can contain it.

The code contains valuable techniques Anthropic uses to control its AI models as coding agents—the "harness"—including a "dreaming" function for task consolidation. Competitors now have a blueprint to replicate Claude Code's capabilities, weakening Anthropic's edge in an already cutthroat market.

The timing is particularly bad: the company is planning an IPO at a $380 billion valuation, and this kind of leak is unlikely to sit well with investors. It also comes just days after a separate leak about Anthropic's new AI model Mythos, also caused by human error within the company's content management system.

Comment Source: WSJ

Matthias Bastian

Apr 1, 2026

AI research

Google Deepmind study exposes six "traps" that can easily hijack autonomous AI agents in the wild

AI agents are expected to browse the web on their own, handle emails, and carry out transactions. But the very environment they operate in can be weaponized against them. Researchers at Google Deepmind have put together the first systematic catalog of how websites, documents, and APIs can be used to manipulate, deceive, and hijack autonomous agents, and they’ve identified six main categories of attack.

Read full article

Comment

Maximilian Schreiner

Apr 1, 2026

AI and society

EU bars AI-generated content from official communications, according to Politico

Politico reports that the Commission, Parliament, and Council have barred their press teams from using fully AI-generated content. Experts see a missed opportunity.

Read full article

Comment