News. Sourced and connected

Pony AI’s second-quarter results show faster robotaxi commercialization, but the company still reported substantial operating and net losses.

The chipmaker is extending its supply strategy beyond semiconductors to long-lived data-center sites, power, and buildings for frontier AI infrastructure.

The FDA’s discussion paper proposes evaluating generative AI medical devices across model capability and clinical-use risk, with controls proportionate to both.

A newly observed rapid-request restriction interrupted a ChatGPT Pro session after four prompts and blocked conversation access for several minutes.

The financing is evidence that securing autonomous software is becoming a separate control layer—not merely another feature inside identity or model evaluation.

A proposed U.S. testing framework has a clearer scope—and unresolved questions about transparency, consequences, autonomous tool use, and production security.

GPT-5.6 Sol is OpenAI’s new flagship reasoning model for difficult coding, research, science, cybersecurity, computer-use, and design work. The rollout is starting with eligible paid ChatGPT plans.

The release makes Instanciar easier to read, easier to configure, and clearer about profile health before a browser session launches.

GPT-5.5 pushed ChatGPT further into practical work execution: coding, research, tool use, data analysis, and longer multi-step tasks.

AutoQuill v0.13 redesigns the live typing session interface, improves sticky target feedback, adds faster profile and token access, and makes humanized typing controls easier to understand.

Instanciar v0.10 improves proxy visibility, browser identity controls, launch feedback, and everyday instance management for separated Chromium workflows.

Meta Muse Spark launches in Meta AI with Instant and Thinking-style modes, image understanding, and a wider rollout across Meta apps.

OpenAI’s GPT-5.4 rollout brought GPT-5.4 Thinking to ChatGPT paid tiers and expanded professional workflows across ChatGPT, the API, and Codex.

AutoQuill v0.12 adds sticky typing, more accurate WPM behavior, a run-state reset button, simulated breaks and pauses, and a cleaner Typing settings layout.

The acquisition was not simply a purchase of A/B-testing software. It combined an experimentation team and product leader with OpenAI’s application organization, while the commercial Statsig platform later moved into Amplitude’s product ecosystem.

GPT-5 simplified model selection for many users while expanding ChatGPT’s role as a work system for reasoning, coding, writing, and multimodal tasks.

AutoQuill v0.11 adds preset save/load, smarter inline commands, dynamic tokens, a wider key library, launch fixes, and smoother typing flow for the Windows auto typer.

Instanciar v0.09 adds a Windows 11-style dark interface, groups, multithreading, proxy health checks, command-bar actions, drag-and-drop organization, and smoother installation.

o3-Pro sharpened the reasoning-model side of ChatGPT before the GPT-5 era, trading speed for more careful work on complex prompts.

China has the policy support, supplier base, industrial-robot footprint, and EV factories to run humanoid pilots at unusual speed. The harder question is whether those pilots can beat specialized automation on reliability and cost.

The all-stock transaction formalized a close relationship between xAI and X, pairing Grok development with one of the largest real-time social platforms under Elon Musk’s control.

What began as a simple multi-image resizer now treats image preparation as a repeatable production pipeline: dimensions, formats, quality, filenames, metadata, previews, and exports in one local browser workflow.

The app formerly called RegexWizard has grown from a quick matcher into a browser workbench for inspecting patterns, testing replacements, and catching fragile assumptions before regex reaches production.