AI news of 25 October 2026 · Robots Versus Humans

Nvidia open-sources Flora to make robot video labeling 125x cheaper than GPT-6 Astra

Nvidia releases Flora tool that labels robot videos 125 times cheaper than prior method, enabling faster robot training.

Nvidia's new data engine Flora labels one hour of robot video for 72 cents instead of $88 with GPT-6 Astra, while scoring slightly higher on quality. The open system, built on three fine-tuned Qwen3.5 models, catches errors in robot actions far more reliably than any competitor but sometimes falls short on descriptive accuracy. Nvidia is positioning Flora as a key piece of its strategy to control the entire robotics value chain.

The Decoder 16h ago 4 min read

AI agents

Anthropic says one AI agent beats many subagents now that models can handle complexity alone

Anthropic research shows single AI agents can replace multiple sub-agents with faster performance and lower complexity.

The Decoder 17h ago

Machine learning, surveillance

Samsung running a TensorFlow lite model on 960x540 image to track you Opens at the publisher in a new tab

Samsung deploys TensorFlow lite model for on-device tracking on low-resolution displays.

Hacker News 15h ago

self-improving AI

Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027 Opens at the publisher in a new tab

Self-improving AI benchmark shows rapid progress, moving timeline forward to 2027 according to Anthropic co-founder.

The Decoder 17h ago

Mobile agents

OpenMobile-2 proves open-source smartphone agents can rival proprietary models

OpenMobile-2 shows open-source smartphone agents rival proprietary models in app task performance.

The Decoder 19h ago

OpenAI, Anthropic

OpenAI offers discounts up to 90 percent to poach Anthropic's biggest customers

OpenAI undercuts Anthropic with discounts up to 90 percent, trading margins for market dominance.

The Decoder 19h ago

AI geopolitics

Ex-Google CEO Eric Schmidt calls OpenAI and Anthropic modern colonial trading companies

Schmidt and historians frame AI development as great power competition with other nations as dependencies.

The Decoder 18h ago

Generative AI

Canva’s first AI image model plans designs before generating them

Canva launches proprietary image generation model with improved customization and lower costs for users.

The Verge 3d ago also at TechCrunch, The Decoder

speech synthesis

Block’s new voice model sounds eerily human Opens at the publisher in a new tab

Block releases voice model that runs on MacBook M4 Max with human-like output.

TechCrunch 1d ago

Dessa, speech AI

Dessa: the most authentic full-duplex speech model ever built Opens at the publisher in a new tab

Dessa builds full-duplex speech model described as highly authentic for conversational AI applications.

Hacker News 2d ago

Your future hasn’t been written yet. No one’s has.
So make it a good one.

Nvidia open-sources Flora to make robot video labeling 125x cheaper than GPT-6 Astra

Anthropic says one AI agent beats many subagents now that models can handle complexity alone

Samsung running a TensorFlow lite model on 960x540 image to track you Opens at the publisher in a new tab

Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027 Opens at the publisher in a new tab

OpenMobile-2 proves open-source smartphone agents can rival proprietary models

OpenAI offers discounts up to 90 percent to poach Anthropic's biggest customers

Ex-Google CEO Eric Schmidt calls OpenAI and Anthropic modern colonial trading companies

Canva’s first AI image model plans designs before generating them

Block’s new voice model sounds eerily human Opens at the publisher in a new tab

Dessa: the most authentic full-duplex speech model ever built Opens at the publisher in a new tab

Latest AI News

Choose sides Choose whose side you're on

Install it as an app

Ready for more?

Spread the word

PWA successfully installed.
Take that, Apple! 😉