The Polylog AI Intelligence Brief

Morning Edition · Wednesday, August 12, 2026

Markets

Nvidia Enlists Six Asset Managers to Turn AI Compute Into a Financeable Asset

Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR signed agreements with Nvidia to build independent financing platforms targeting more than $500 billion in third-party capital, with Nvidia retaining some of the credit risk.

4 sources
Corroborated

Tech

OpenAI Puts Its Offensive Cyber Model on Amazon Bedrock Behind a Vetting Program

Daybreak Red serves GPT-5.6-Cyber, a model OpenAI says completes 95% of requests involving exploit chains and privilege escalation, the same category of request its general-purpose model refuses in 98% of cases.

3 sources
Corroborated

Tech

Google Reports Its Medical AI Matched Doctors in Simulated Video Consultations

In a randomized examination with 30 primary care physicians, 15 patient actors and 100 scenarios, evaluators rated the video version of AMIE at or above physicians on diagnosis and management.

3 sources
Corroborated

Tech

Nvidia Releases a 31.6-Billion-Parameter Open Model That Activates 3.6 Billion Per Token

Nemotron 3.5 Lightning uses a hybrid Mamba-Transformer design, runs on a single H100, and scores 24 on the Artificial Analysis Intelligence Index against 15 for its predecessor.

4 sources

Tech

Meta Opens the Weights of Muse Glimmer and Says Its Flagship Coding Model Will Follow

The 30-billion-parameter multimodal model ships under Apache 2.0 at under 20 gigabytes, days after Meta launched Muse Spark 1.2 and a terminal coding agent built on it.

4 sources
Corroborated

Tech

Anthropic Begins Watermarking All Claude Text Worldwide Under an EU Transparency Code

The marking happens at the model level, so it follows output through the interface, the application programming interface and Claude Code, days after researchers published a general method for stripping watermarks from artificial intelligence (AI) images.

4 sources
Corroborated

Tech

Anthropic Makes Automatic Approval the Default in Claude Code on 14 August

A classifier will screen each tool call for irreversible or out-of-bounds actions instead of prompting the user, and Anthropic has stopped billing Pro, Max and Team subscribers for the extra tokens it consumes.

4 sources

Tech

OpenAI Extends ChatGPT Advertising to Five More Countries

Ads now appear for free and Go users in the United Kingdom, Mexico, Brazil, Japan and South Korea, following an initial rollout across the United States, Canada, Australia and New Zealand.

3 sources

Tech

Meta's Open Vision Models Cut Beamline Analysis From a Month to Fifteen Minutes at Berkeley Lab

Segment Anything 3 and DINOv3 run on 300 A100 accelerators at the National Energy Research Scientific Computing Center, inside a Department of Energy project that keeps prepublication data off commercial clouds.

3 sources
Corroborated

Tech

New Preprint Attacks the Arithmetic of Low-Precision Formats, Not Just Their Storage

CurveFP proposes a codebook family whose products stay inside the format, targeting the error that microscaling four-bit types introduce during multiplication rather than during rounding.

3 sources

Tech

An Unannounced Image Model Carrying OpenAI Provenance Signals Appears in Blind Testing

Testers say images from a model listed as mona-lisa-1 on Arena are recognized by OpenAI's own verification tool, and show markedly less of the synthetic skin texture that identifies current generators.

3 sources
Plausible

World

Singapore Runs Computer Vision at Industrial Scale to Suppress Its Dengue Mosquitoes

Google's Debug unit expanded its Singapore site in May to automate sex-sorting and release, supporting a national programme the environment agency credits with 80% to 90% suppression of Aedes aegypti.

3 sources
Corroborated