Articles
Kyutai teaches a speech model to do arithmetic out loud
Kyutai released two open-weight speech-to-speech models that reason through math without transcribing audio first, posting a large jump on spoken GSM8K.
Editor's Picks
Hugging Face ships a $399 duck robot you can retrain
Hugging Face unveiled Microduck, a $399 open source duck robot that learns tricks through reinforcement…
Keenable scores $26M from Accel for an agent-first web index
A Yandex search veteran's startup raised $26M to build a web index tuned for AI…
ChatGPT Work signs into your web accounts all by itself
ChatGPT Work can now store login cookies in its cloud browser and sign into your…
NVIDIA doubles edge AI speed with Jetson Orin Nano 2 module
NVIDIA's Jetson Orin Nano 2 doubles entry-level edge inference and cuts power draw 40 percent.
Latest Articles
Classroom test finds students without AI finish last
A law professor split his course into three groups for two years and found the cohort without ChatGPT placed last…
Fly connectome wired into a small model brings no measured gain
A developer attached all 166,700 nodes of the fruit fly brain to a frozen 1.2B model, and the project's own…
AWS opens the agent inbox behind its internal Pizza Bot tool
The company rebuilt an internal tool that served more than 2,000 Amazon staff as a self-hosted inbox where agents drop…
Princeton proposal loops decoder state across every token
A technical report argues that letting each token see the previous one's final hidden state, rather than only cached attention…
Cognition post-trains SWE-2 on Kimi K3 to undercut Fable 5.1
The Devin maker tuned Moonshot's open model with reinforcement learning and says the result lands within a point of Anthropic's…
GPT-6 Astra tops two agent tests from drone piloting to vending
OpenAI's newest model beat a human baseline on every drone-control subtask and tripled a rival's simulated vending income in fresh…
Redis moves the cache outside the model to cut LLM bills
LangCache stores finished answers rather than model state, so a paraphrased question can skip inference altogether.
NVIDIA hands factory allocations to a planner built with Palantir
NVIDIA's operations team now uses a Palantir Foundry command center and its own cuOpt solver to decide weekly how parts…
