All AI news

AIReflection AI says its Beam model was pretrained in 4 weeks on 24T high-quality tokens, giving it innate coding capabilities. The company credits MoE stability improvements and large-scale data curation and deduplication for a base model it claims outperforms open-source base models of the same class. It presents this strong reasoning foundation as what makes sustained reinforcement learning gains possible.

AIReflection is set to release Beam, its first open-weight AI model, aiming to become a Western counterpart to DeepSeek. The source says Beam is trained from scratch for coding, reasoning, and AI agents, with benchmarks placing it alongside the strongest open models and more efficient token economics. Reflection has raised $4.6 billion from investors including Nvidia, Sequoia, and Lightspeed, and the interview covers its monetization plans for open-weight models.

AIGoogle for Developers highlights Art Codec, an interactive web experience built by @soumyadesign in Google AI Studio. It lets users explore classic masterpieces through dynamic hotspots, motifs, and visual techniques.

AIDeveloper @mwjoffe combined Antigravity with Lyria RealTime to create a 3D meadow in the browser that generates its adaptive soundtrack live. The demo showcases dynamic game audio, with music generated in real time as users interact with the scene.

AISuno announces albums as a new feature for artists to present cohesive collections on the platform. The post introduces artists behind some of the first albums released on Suno, linking to a blog highlighting them.

AICognition introduces Dreaming, a feature in which Devin builds a memory graph of how a user likes to work across sessions. At night, Devin self-improves this memory by removing stale records and discovering latent information. Cognition also says it is creating an open-source standard called Agent Memory Repo, linked in the post.

AILiquid AI's d1 picks the next move by reading game screenshots, without text versions of the board. In Tetris, adding the screen raised cleared lines from 70 to 81, and in Wordle it solved 12/12 games in an average of 3.8 guesses.

AILiquid AI released d1 with vision support, accepting images, text, or both as inputs. In tests on six real applications, d1 matched or beat GPT-6.1 Sol on four while costing 19x to 200x less than both GPT-6.1 Sol and Claude Opus 5.5. It returns probabilities for yes/no, choice, or score questions in one forward pass, with text decisions in 200 to 300 ms.

AILiquid AI's vision-enabled decision model d1 inspects parts directly from camera images and is described as the best such model currently on the market. It reaches 85% to 97% accuracy across four VisA inspection tasks covering circuit boards, candles, cashews, and chewing gum. It understands each task from a short description without task-specific training.

AINous Research says Hermes Agent can handle far more than flight booking and points to a documentation page listing 326 use cases from real users. The post gives no further details about specific tasks or results.

AILiveKit Agents pairs MAI-Transcribe-2-Streaming for speech-to-text, Gemma 4 on LiveKit Inference for reasoning and tool calls, and MAI-Voice-2.1-Flash for speech output in a demo support call. The post links separate speech-to-text and text-to-speech resources for developers.

AICursor is enabling an option to replace its built-in system prompt with a custom one, while rules, skills, and tool schemas still load. The feature is being rolled out account by account rather than to all users at once.

AICursor now supports MCP annotations such as readOnlyHint and destructiveHint on custom tools. These let the model distinguish a lookup from a delete before calling a tool.

AIBackground subagents now return their results to the parent as a follow-up turn on the same run. The stream() and wait() calls carry through until every subagent finishes.

AICursor announced that developers can steer Cursor SDK agents while they run using run.steer(), which adds a message to the next turn. If a subagent is mid-task, it moves to the background and continues working.

AITogether AI has announced Together Link, which lets developers run frontier open models inside their favorite coding harness. The product includes spending tracking and an Auto router that selects low-cost models for quick fixes and more capable models for harder tasks.

AIUpstage's Solar Mini 4 is free on Nous Portal for the next two weeks. The model has 3B active parameters out of 35B total and a 512K context window. It scores 24 on the Artificial Analysis Intelligence Index, above models with roughly 10x the active parameters.

AIOpenAI is rolling out text watermarking for EU AI Act compliance, with opt-in access for API customers globally on select models starting today. Watermarking stays off by default in the API, while an invisible watermark will be added to eligible ChatGPT and Codex text in the European Union over the coming weeks. Access to the text watermark detector is initially limited to approved researchers and expert organizations, and the image and audio verification tools remain publicly accessible.

AINico Caprez and Madison Huang, Jensen Huang's daughter, are described as a power couple inside Nvidia. The post says they serve as consiglieres to Jensen Huang and have been entrusted with key initiatives, including turning GPUs into a financial asset.

AIHugging Face says a capture proxy lets reinforcement learning train open models inside unmodified coding harnesses such as Claude Code, Codex, and OpenCode. The proxy records the exact token IDs and logprobs vLLM samples and hands them to TRL for training. On LFM2.5-2.6B, training in four harnesses at once raised OpenCode results from 34% to 58%, while SFT on 3,189 Qwen3.8-27B rollouts plateaued at 47.5%.

Why it matters: The capture proxy lets models train inside real coding harnesses without reimplementing them, with measured gains and a comparison against SFT on the same data.

AIDatabricks has made IP Functions generally available, letting users parse, validate, and join IPv4 and IPv6 addresses and CIDR blocks with built-in SQL functions optimized in Photon. In benchmarks versus another leading cloud data warehouse, CIDR joins ran up to 3.1x faster and cost up to 6.4x less. The functions support its Security Lakehouse vision for threat detection, investigation, and network analytics on one governed copy of data.

AIElevenCreative is launching The Search, a $100,000 competition to find the world's catchiest ad, with a jingle people keep humming the next day as the core criterion. First place wins $50,000, and 11 winners each get a one-on-one session with the ElevenLabs Creative Production Team.

AICohere's North 2 lets users create reusable agents, automations, and multi-agent orchestrations for completing tasks. Agents now retain context across sessions rather than starting fresh each time.