AI — Page 25

Annie Sexton's interactive guide explains how token probabilities become compression costs; ngrok separately operates an AI Gateway for routing model traffic.

Michael Intrator's AI cloud reported $2.58 billion in quarterly revenue and a $626 million loss while its backlog reached $104 billion.

The 138 billion-parameter mixture-of-experts model adds image input and arrives 10 weeks after Microsoft's first in-house Copilot coding model.

The roster entries point to multimodal and web-enabled variants, but their labels do not establish that Grok 4.6 is behind them.

The open-source Mac, Windows and Linux app combines local inference, training and connections for Claude Code and Codex.

The 30-billion-parameter open-weight model activates 3 billion parameters and permits commercial use under Nvidia's OpenMDW license.

The agent pipeline plans terrain, builds regional assets and refines scenes in Blender, using four Nvidia H20 GPUs in Tencent's tests.

An outside tester received preview access hours after Cursor users found the unreleased model in its production picker.

Hansson's reversal shows how knowledge workers move from skepticism to agent-supervised output once AI can handle assignments, a transition journalism is likely to follow.

The Beijing AI company reset GLM Coding Plan limits, but its self-reported ZCode milestone does not distinguish registered, active or paying users.

Nick Dobos's objection exposes a policy gap: EU guidance excludes source code, while Anthropic says its marks cover all generated text.

The Shanghai AI company asked viewers to identify a subject in inaccessible media, leaving no verifiable connection to H3 or another product.

The tool would scan messages on Pangram's servers, apply AI, Mixed or Human labels, and automate where flagged mail goes.

In an interview, Meta's founder offered no release date or independent evaluations as the company pushes 2026 capital spending toward $145 billion.

The policy covers Claude's API and cloud partners, while generated files will carry C2PA provenance metadata where supported.

The model rose eight places from MAI-Image-2.5 and led Arena's 3D category, narrowing OpenAI's lead to 45 points.

Henry Ndubuaku and Roman Shemet designed Needle 2 to turn speech into typed actions on CPUs, with Pebble already running Needle locally.

The 10-to-12-family launch turns a liberal arts campus into a test site for TimeBack, 2HR's two-hour academic model.

CEO Paul Edmondson told staff the plan combines the Info Sentience acquisition with a new product called Cutter Studio.

Meta is publishing downloadable Muse Glimmer weights under Apache 2.0, giving developers control over a multimodal agent model that can run within 24 GB or 32 GB memory envelopes.

Microsoft is seeking TSMC capacity for more than 300,000 chips in 2027 after deploying Maia 200 in only two U.S. data centers.

Anthropic says an unreleased research model increased the fraction of zeta-function zeros known to lie on the critical line, but the supplied post excerpt ends before stating either figure.

The AI lab promised new power support, efficient cooling and usage disclosures as Texas audits projects seeking grid connections.

Stoa Markets' YC S26 founders want firm bids and transaction data to replace GPU deals conducted through calls, spreadsheets, and group chats.