Lead Story · AI News

Qwen Cuts Translation Lag to 2.3s With New Interleave Architecture

Alibaba’s Qwen team just dropped Qwen3.8-LiveTranslate, a model that doesn’t just translate speech—it keeps up with it. By shifting from traditional batch processing to a new Interleave architecture, the system slashes average lagging (LAAL…

Continue reading

Qwen Cuts Translation Lag to 2.3s With New Interleave Architecture

From the wire

Google’s Gemini autonomously hacked three firms during security tests

Meta’s Muse admits it can’t explain how it reads your texts

Vals raises $40M from a16z to fix broken AI benchmarks

Save $200 on TechCrunch Disrupt 2026 Tickets Before Sept. 25 Deadline

Research

Build a Vector DB in 10 Steps: No GPU, Just NumPy

MIT Editors: AI Won't Wipe Us Out, But It Will Kill Some

Microsoft and OpenAI staff warn AI is eating the web

SerpApi cuts AI search tokens by 74% with Markdown output

Anthropic Staff Warn AI Could Kill Humans Within a Decade

Kill 5 Models With One: BGE-M3 Embeddings Unify Multilingual Text Classification

Trending

AI Regulation Surges as Wild West Era Ends in Sept 2026

Nvidia Buys Hugging Face as AI Public Interest Fades

AI Trends: Builders Race Ahead as Public Hype Fades

AI's Reckoning: Self-Improvement Faces Audits and Null Tests

AI Trends: From Hype to Infrastructure

AI Consolidation: Upskilling Surge Meets Agent Reliability Push

Startups

Trump Calls AI Safety a 'Hoax,' Plans New AI Force and Czar

Flock Safety offers buyouts to 1,500 staff amid surveillance backlash

Google’s Gemini autonomously hacked three firms during security tests

Vals raises $40M from a16z to fix broken AI benchmarks

Save $200 on TechCrunch Disrupt 2026 Tickets Before Sept. 25 Deadline

Andrew Yang Cites Viral AI Bot Myth; OpenAI Scientist Warns of Heat-Sensor Hacks

The reference desk

How inference cost actually breaks down

Speculative decoding, KV cache, and more

Model vs. model, benchmarked

The sharpest AI news, curated daily.