Skip to content
IFAI

Vibe coding

Industry tag968 stories

  1. Open sourceReach 50

    Moonshot AI drops open-source Kimi K2 with agentic superpowers

    Moonshot AI just open-sourced Kimi K2, a 1-trillion parameter MoE model with 32B activated parameters engineered specifically for coding and autonomous agent workflows. It scored 53.7% on LiveCodeBench, comfortably beating DeepSeek-V3 and GPT-4.1, powered by a new MuonClip optimizer that eliminates training instability. Why it matters: High-end agentic intelligence is rapidly shifting from pricey closed APIs to open, dirt-cheap alternatives.

  2. Open sourceReach 46

    Reka drops Flash 3.1 with near-lossless 3.5-bit compression

    Reka just open-sourced Flash 3.1, a 21B model with major coding upgrades, alongside its new Reka Quant library. The team managed to squeeze the model down to 3.5 bits for llama.cpp with minimal accuracy loss, giving developers a compact agent base that punches well above its weight.

  3. Open sourceReach 48

    Mistral and All Hands drop Devstral coding models

    Mistral AI just teamed up with All Hands AI to roll out the Devstral models built specifically for coding agents. The 24B Devstral Small 1.1 is fully open under Apache 2.0 and hits 53.6% on SWE-Bench Verified, while Devstral Medium reaches 61.6% on the benchmark via API at roughly a quarter of the price of frontier competitors.

  4. AI modelsReach 46

    OpenAI reportedly ready to launch open-weight model next week

    OpenAI is preparing to release an open-weight language model as soon as next week, according to sources. Described as similar to o3 mini with advanced reasoning capabilities, it will be available on Azure, Hugging Face, and rival cloud providers, marking OpenAI's first open-weight model since GPT-2 in 2019.

  5. AI toolsReach 38

    Deepgram wants you to code with your voice

    Deepgram just launched Saga, a voice-powered operating system designed to embed directly into coding environments like Cursor and Windsurf. Developers can speak out ideas to generate prompts or manage Jira and Slack tickets without touching the keyboard. Why it matters: It turns speech into the primary interface for software development, killing the quiet productivity tax of constant tab-switching.

  6. AI toolsReach 42

    Cursor apologizes for confusing Pro plan price changes

    Cursor maker Anysphere just apologized after shifting its $20-per-month Pro tier to API-rate usage, which caught subscribers off guard with unexpected charges and quickly depleted balances on newer models like Claude. CEO Michael Truell promised refunds to affected users, highlighting the growing cost pressures AI coding tools face from frontier model providers.

  7. AI modelsReach 52

    Elon Musk drops Grok 4

    Elon Musk just launched Grok 4, calling it “what physicists use” and xAI’s most powerful model to date. Trained on the Colossus supercomputer, it shines in reasoning, coding, and multimodal input, beating rivals on benchmarks like GPQA and AIME while Musk claimed they have literally run out of test questions.

  8. AI toolsReach 42

    Cursor Agent goes mobile

    Cursor just launched Cursor Agent on web and mobile, letting developers kick off coding tasks from any browser or PWA. Agents can fix bugs, create pull requests in the background, and trigger directly through Slack mentions, keeping everything in sync when you return to your desktop editor.

  10. RoboticsReach 48

    NVIDIA drops early preview of Isaac Sim 5.0 and Isaac Lab 2.2

    NVIDIA just released early developer previews of Isaac Sim 5.0 and Isaac Lab 2.2 on GitHub. The updates give robotics builders physics-based simulation tools to train and validate robots, adding open-source access, synthetic data generation extensions, and new robot models.

  11. IndustryReach 42

    OpenAI weighs antitrust claims against Microsoft as rift deepens

    OpenAI executives have reportedly considered publicly accusing Microsoft of anticompetitive behavior and asking federal regulators to review their contract. The tension escalated over OpenAI’s $3 billion acquisition of Windsurf, as OpenAI refuses to let Microsoft tap into the startup’s coding IP for GitHub Copilot. OpenAI wants to loosen Microsoft’s grip on its compute, but it still needs its partner's sign-off to complete a for-profit transition.

  12. AI toolsReach 41

    Cursor drops $200 Ultra plan for coding power users

    Cursor just launched a new $200/month Ultra plan packing 20x more usage than Pro, tailored for heavy coders who want predictable bills instead of pay-as-you-go spikes. The company is also upgrading Pro by shifting from request caps to compute limits, unlocking unlimited tool calls and unlimited access to its Auto model.

  13. AI toolsReach 38

    Groq lands on Hugging Face for blazing-fast inference

    Groq just officially became a supported Inference Provider on the Hugging Face Hub. Developers can now run open-source models like Llama 4 directly through Groq's low-latency LPU chips using HF's web UI or Python and JS SDKs. You can bill through Hugging Face without added markups or plug in your own Groq API key.

  14. AI toolsReach 36

    rabbit launches intern, a general-purpose AI agent

    rabbit just dropped its first software-only product, rabbit intern, an AI agent that builds full websites, research reports, and games from a single prompt. It swaps confusing token credits for straightforward task-based pricing, and anyone can test three tasks for free without an r1 device.

  16. AI toolsReach 39

    Google supercharges Gemini Code Assist with Gemini 2.5 and 1M context

    Google just upgraded Gemini Code Assist with its flagship Gemini 2.5 models, boosting reasoning and code generation. Developers can now feed entire folders into chat via a 1M token context window, set project-wide rules, create custom shortcut commands, and switch between multiple saved chat sessions. Personal users can try it for free in VS Code or JetBrains IDEs.

  17. ResearchReach 35

    Sakana AI turns text prompts into custom LoRA adapters

    Sakana AI just unveiled Text-to-LoRA, a hypernetwork that generates task-specific LLM adapters directly from plain English prompts. Instead of gathering datasets and burning compute on fine-tuning, it creates LoRAs in a single step and even generalizes to unseen tasks. The code is open source ahead of ICML 2025.

  18. AI modelsReach 51

    Google upgrades Gemini 2.5 Pro preview ahead of GA

    Google just rolled out an upgraded preview of Gemini 2.5 Pro before its general availability in a couple of weeks. The update gained 24 Elo points on LMArena to hit 1470, improves formatting and creative writing, and introduces thinking budgets in Google AI Studio to let developers manage latency and cost.

  19. Open sourceReach 44

    Stanford drops Tokasaurus, an inference engine built for raw throughput

    Stanford researchers just released Tokasaurus, an open-source LLM inference engine built specifically for high-throughput batch workloads like synthetic data generation. By slashing CPU overhead, dynamically grouping shared prefixes, and optimizing pipeline parallelism for non-NVLink GPUs, it beats vLLM and SGLang by over 3x on key benchmarks.

  20. AI toolsReach 40

    Cursor 1.0 adds code reviews and background multitasking

    Cursor just launched Cursor 1.0, giving the AI editor built-in code review, the ability to remember past mistakes, and background multitasking for dozens of jobs at once. It means developers can offload tedious fixes and refactoring to the background without getting blocked.