Independent AI news, with every claim linked to its primary source
Independent AI news, with every claim linked to its primary source
What happened in AI, and what it means. Model releases, company moves, funding and regulation.
Google launched AI study tools across Search and Gemini on August 19, 2026, with practice quizzes for nine standardized exams and a free year of a paid AI plan for eligible college students.
Read More
Stripe confirmed its OpenRouter deal on August 19 without disclosing terms. The NYT put it at $7.5 billion, close to six times the $1.3 billion valuation OpenRouter raised at in May.
Read More
AWS published ADOP, a reference architecture where Claude Code sub-agents on Amazon Bedrock generate ETL, quality checks and Cedar policies at build time, so production runs deterministic code with no model in the loop.
Read More
AI accounting startup Rillet has raised a $100 million Series C at a $1 billion valuation. Iconiq led the round,
Read More
OpenAI’s new plugin lets ChatGPT read, search and send Apple Messages on the Mac, in exchange for Full Disk Access and a stack of permissions.
Read More
Together AI ran 904 DeepSWE rollouts: GLM-5.3 tied Claude Fable 5 on first-attempt accuracy at $3.99 a rollout against $21.63, and a GLM-first cascade beat GPT-5.6 Sol.
Read More
TikTok and ByteDance settle the DOJ’s COPPA lawsuit for $400 million: $300 million now and $100 million once the Musical.ly consent decree is vacated.
Read More
Starcloud added $250 million to its Series A at a $2.3 billion valuation, with Nvidia and Cisco joining as it races to book rocket launches before Falcon 9 winds down.
Read More
Tesla is recalling about 3 million vehicles in China over door handles and driver monitoring, the biggest slice of a record recall of nearly 4.3 million cars from 11 automakers.
Read More
A Rust multi-agent system built to turn LLM hallucination into testable hypotheses lost to plain self-reflection across 180 runs and six ablation conditions.
Read More
Diversion decoding steers a model away from its own answer and scores how hard it resists. UT Dallas and NIST report 78.49% AUROC on Llama 2 13B using two generated answers instead of ten samples.
Read More
Two IIT Delhi researchers find that a single safety instruction makes aligned vision-language models refuse questions they had just answered from the same image, and that suppressing one activation direction brings the answers back.
Read More