The latest developments, launches, and updates in artificial intelligence. Every story is a short summary in our own words — never copied — with why it matters and a link to the original source.
OpenAI ships GPT-Live-1.1 voice model in its API, with cheaper standby billing for idle calls
OpenAI made GPT-Live-1.1 available in its API as a full-duplex voice model that listens and speaks in one stream, with listed improvements in multilingual accents, transcription, and task completion over GPT-Live-1. A new standby state pauses inference during silence and bills idle WebRTC time at $0.02 per minute versus $0.05 active.
Why it matters: Cheaper idle time makes always-on voice agents — support lines, receptionists, phone assistants — noticeably less expensive to run.
Who should care: Voice-agent developers and customer-support teams
AI stock rout deepens: Nasdaq falls 5.3% on Nov 20, erasing its 2026 gains
U.S. markets sold off sharply on Friday, November 20: the Nasdaq Composite fell 5.28% to 22,617.73, wiping out its gains for 2026, while the S&P 500 dropped 3.70% and entered correction territory. AI names led the decline — Nvidia fell about 13.8% following reports of an SEC examination of its accounting.
Why it matters: The selloff signals investors are re-pricing AI spending and valuations, which could affect funding and pricing across the AI tool market.
Who should care: Investors and anyone following AI company valuations
Thinking Machines releases Inkling 2 Mini, an open-weight model built to run on a laptop
Thinking Machines Lab released Inkling 2 Mini with open weights under Apache 2.0. It is a mixture-of-experts model with 26.4B total and 3.4B active parameters that re-runs its middle layers three times per token. Its roughly 20 GB footprint lets it run on laptops; it is on Hugging Face, Ollama, and the company's Tinker platform.
Why it matters: It suggests capable agentic models no longer require cloud GPUs, which lowers the cost and privacy barriers for local AI tools.
Who should care: Developers and anyone running AI locally
Perplexity's Computer agent now works inside Chrome, Edge, Brave and Arc via an extension
Perplexity announced that its Computer agent now runs directly inside Chromium-based browsers — Chrome, Edge, Brave, Arc and others — through the Perplexity extension. Users hand off a task from the side panel and watch the agent work across their open tabs and logged-in sites, instead of installing Perplexity's separate Comet browser.
Why it matters: Agents are moving into the browser people already use, which removes the biggest adoption friction for browser automation tools.
Who should care: Knowledge workers and anyone testing AI browser agents
METR: AI models' autonomous work horizons keep doubling roughly every 3.6 months, but research judgment lags
METR published its Q4 capability update (Time Horizon 2), evaluating 15 frontier model checkpoints. It found that models' 50%-success task horizons on verifiable coding tasks are doubling about every 3.6 months, while performance on open-ended research judgment remains much weaker. METR says current evidence cannot rule out drastic AI R&D acceleration within a couple of years.
Why it matters: If the doubling trend holds, the length of work agents can finish unsupervised keeps compounding — a planning input for every team adopting them.
Who should care: AI developers, researchers, and policy watchers
OpenAI says new safeguards cut reasoning-extraction attacks on GPT-6.2 Astra from 54% to 0.2%
OpenAI published research on adversarial distillation — rivals extracting a model's hidden reasoning to train their own. In OpenAI's internal red-team tests, attacks pulling protected reasoning succeeded 54.0% of the time against GPT-6 Astra but 0.2% against GPT-6.2 Astra, after OpenAI tied protected reasoning to the conversation history that produced it. The figures are OpenAI's own evaluations.
Why it matters: Protecting hidden reasoning is becoming a front line in the model-copying fight, and it may shape how much capability leaks between labs.
Who should care: AI developers, security teams, and model providers